Flexible Distribution Systems Through the Application of Multi Back-To-Back Converters : Concept, Implementation and Experimental Verification

Flexible distribution systems through the application of multi back-to-back converters : concept, implementation and experimental verification

Citation for published version (APA): Graaff, de, R. A. A. (2010). Flexible distribution systems through the application of multi back-to-back converters : concept, implementation and experimental verification. Technische Universiteit Eindhoven. https://doi.org/10.6100/IR673052

DOI: 10.6100/IR673052

Document status and date: Published: 01/01/2010

Document Version: Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers. Link to publication

General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please follow below link for the End User Agreement: www.tue.nl/taverne

Take down policy If you believe that this document breaches copyright please contact us at: [email protected] providing details and we will investigate your claim.

Download date: 06. Oct. 2021

Flexible distribution systems through the application of multi back-to-back converters: Concept, implementation and experimental veriﬁcation

PROEFSCHRIFT

ter verkrijging van de graad van doctor aan de Technische Universiteit Eindhoven, op gezag van de rector magniﬁcus, prof.dr.ir. C.J. van Duijn, voor een commissie aangewezen door het College voor promoties in het openbaar te verdedigen op woensdag 26 mei 2010 om 14.00 uur

door

Roald Antonius Adrianus de Graaﬀ

geboren te Waalwijk Dit proefschrift is goedgekeurd door de promotor: prof.ir. W.L. Kling

Copromotor: dr. J.L. Duarte

Copyright © 2010 R.A.A. de Graaﬀ All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic, mechanical, including photocopy, recording, or any information storage and retrieval system, without the prior written permission of the copyright owner.

The work leading to this thesis was supported by KEMA and the IOP-EMVT program of SenterNovem.

CIP-DATA LIBRARY TECHNISCHE UNIVERSITEIT EIND- HOVEN Graaff, Roald A.A. de Flexible distribution systems through the application of multi back-to-back converters: Concept, implementation and experimental verification / by Roald Antonius Adrianus de Graaff. - Eindhoven: Technische Universiteit Eindhoven, 2010. Proefschrift. - ISBN 978-90-386-2220-0 NUR 959 Trefw.: Elektriciteitsdistributie / Vermogenselektronica / Ver- mogenssturing / Besturing elektriciteitsdistributie / Span- ningskwaliteit / Spanningsregeling Subject headings: Power distribution / Power electronics / Load flow control / Power distribution control / Power quality / Voltage control To Susana To my parents Promotor: prof.ir. W.L. Kling, TU/e

Copromotor: dr. J.L. Duarte, TU/e

Core committee: prof.dr.ir. R.W. De Doncker, RWTH Aachen University prof.dr. E. Lomonova, TU/e prof.dr. J.A. Pe¸cas Lopes, University of Porto

Other members: prof.dr.ir. J.H. Blom (reserve), TU/e dr.ir. F. van Overbeeke, EMforce prof.dr. A.G. Tijhuis (chairman), TU/e ir. P.T.M. Vaessen, KEMA Contents

List of Figures v

List of Tables ix

Abstract xi

Samenvatting xv

1 Introduction 1 1.1 Changes in electrical power generation ...... 2 1.2 Changes in the organization of power systems ...... 4 1.3 Consequences for the distribution network ...... 5 1.4 Ongoing research ...... 6 1.4.1 Communication and automation ...... 8 1.4.2 Load control ...... 8 1.4.3 Generation control ...... 9 1.4.4 Storage ...... 10 1.4.5 Power electronics ...... 10 1.4.6 Active distribution networks ...... 11 1.5 Research objective ...... 11 1.6 Research questions ...... 12 1.7 Research approach ...... 12 1.8 IOP-EMVT programme ...... 13 1.9 Outline of the thesis ...... 15 1.10 Publications ...... 16

2 Distribution systems 17 2.1 The network operator’s role ...... 17 2.2 Network topology and redundancy ...... 19 2.2.1 Network topology ...... 19 2.2.2 Redundancy ...... 20 2.3 Power quality aspects ...... 21 2.3.1 Steady state voltage amplitude ...... 21

i ii Contents

2.3.2 Flicker ...... 23 2.3.3 Voltage dips ...... 24 2.3.4 Phase angle jumps ...... 25 2.3.5 Power frequency ...... 26 2.4 Voltage control ...... 26 2.5 Conclusion ...... 27

3 FACTS in distribution systems 29 3.1 Principles of power flow control ...... 29 3.1.1 Power flow in overhead line ...... 30 3.1.2 Power flow in underground cable ...... 35 3.2 FACTS technologies ...... 36 3.2.1 Solid-state switching devices ...... 36 3.2.2 Converter topologies and switching strategies . . . 37 3.2.3 Mechanical switches ...... 39 3.3 FACTS and D-FACTS applications ...... 40 3.3.1 Shunt FACTS and D-FACTS devices ...... 41 3.3.2 Series FACTS and D-FACTS devices ...... 45 3.3.3 Mixed form FACTS and D-FACTS devices . . . . 46 3.4 Conclusion ...... 52

4 Functional concept of the Intelligent Node 55 4.1 Facilitating increased loading ...... 56 4.1.1 Controlled sharing of redundancy ...... 56 4.1.2 Controlled power exchange between grid areas . . 60 4.2 Controlling voltage proﬁles ...... 61 4.2.1 Example application ...... 63 4.3 Voltage dip mitigation ...... 70 4.3.1 Example application ...... 71 4.4 Possible Intelligent Node topologies ...... 74 4.4.1 Power electronics controlled auto transformers . . 74 4.4.2 Power electronics controlled series impedances . . 75 4.4.3 Power electronics converters ...... 76 4.5 Conclusion ...... 77

5 Intelligent Node control and protection 79 5.1 Basic converter controls ...... 79 5.1.1 Controller discretization ...... 81 5.1.2 AC current control ...... 82 5.1.3 AC voltage control ...... 83 5.1.4 Active and reactive power control ...... 85 5.1.5 DC bus voltage control ...... 85 5.2 IN response to unplanned power system events ...... 86 5.2.1 Voltage dip mitigation by injecting reactive power 88 Contents iii

5.2.2 IN protection concept ...... 94 5.3 IN role in planned power system events ...... 96 5.3.1 Energization and de-energization ...... 96 5.3.2 Disconnecting grid areas ...... 97 5.3.3 Connecting grid areas ...... 103 5.4 Conclusion ...... 114

6 Laboratory-scale demonstration 115 6.1 Experimental set-up ...... 115 6.1.1 Converter control implementation ...... 117 6.1.2 Modeling of experimental set-up ...... 118 6.2 Basic converter step responses ...... 119 6.2.1 Changing power reference values ...... 120 6.2.2 Changing voltage reference value ...... 121 6.2.3 Changing load ...... 121 6.3 Transition from radial to meshed operation ...... 124 6.3.1 Synchronization ...... 124 6.3.2 Three-phase load-break switch closing ...... 124 6.3.3 Phase-by-phase load-break switch closing . . . . . 127 6.3.4 Ensuring load-break switch closing detection . . . 131 6.4 Transition from meshed to radial operation ...... 132 6.4.1 Three-phase load-break switch opening ...... 133 6.4.2 Phase-by-phase load-break switch opening . . . . . 134 6.5 Voltage dip and swell mitigation ...... 136 6.6 Conclusion ...... 137

7 Conclusions, thesis contribution and recommendations 139 7.1 Conclusions ...... 139 7.2 Thesis contribution ...... 143 7.3 Recommendations ...... 144

References 147

Abbreviations, symbols and notations 159

A DC current of AC/DC converter 163 A.1 DC link current in single phase voltage source converter . 163 A.2 DC link current in three-phase voltage source converter . 164

B Simulations and experimental results practical set-up 167

Acknowledgements 185

Curriculum vitae 187

List of Figures

1.1 Small and medium scale renewable energy sources...... 3 1.2 Google Timelines results for selected search terms...... 7 1.3 Structure of the IOP-EMVT Intelligent Power Systems project. 14

2.1 Types of medium and low voltage grids...... 19 2.2 Each feeder can supply the load of another feeder...... 20 2.3 Sequence of events during a short-circuit...... 22 2.4 Curve for Pst = 1 for rectangular voltage changes...... 24 2.5 Typical voltage coordination radial MV/LV network...... 27

3.1 Single-line and phasor diagram for overhead line...... 30 3.2 Power flow control using a series voltage source...... 32 3.3 Power flow control using a series impedance...... 33 3.4 Power flow control using a parallel device...... 34 3.5 Power flow in cable with series voltage source...... 35 3.6 Ratings of solid-state switching devices...... 37 3.7 Basic diagrams current and voltage source converters. . . . . 38 3.8 Topology and output voltage source converters...... 39 3.9 PWM switching technique applied to a single-switch topology. 39 3.10 Mechanical switches applied in FACTS...... 40 3.11 Basic (D-)FACTS connection methods...... 41 3.12 Single-line diagram and operating characteristic SVC. . . . . 42 3.13 Single-line diagram and operating characteristic STATCOM. 44 3.14 Single-line diagram and operating characteristic TSSC/TCSC. 47 3.15 Single-line diagrams SSSC and D-SSSC...... 47 3.16 Single-line diagrams UPFC and IPFC...... 48 3.17 Transfer switch...... 49 3.18 Single-line diagram and operating characteristic B2B device. 50

4.1 Sharing of redundancy...... 57 4.2 Sequence of events during short-circuit in a network with a 3-port IN...... 58

v vi List of Figures

4.3 Manual phase-by-phase opening/closing of MV switchgear. . 59 4.4 Traditional grid reinforcements to supply increasing load Ld-C. 60 4.5 IN application to supply increasing load Ld-C...... 61 4.6 Voltage profiles with increasing DG...... 62 4.7 Voltage profile ’bending’ with the application of a 2-port IN. 63 4.8 Medium voltage cable network with a 4-port IN...... 64 4.9 Graphical impression gradient method applied to 2D problem. 66 4.10 Flowchart for voltage profile optimization using Cauchy’s gradient method...... 67 4.11 Configuration I: Iteration path towards optimal IN settings. . 69 4.12 Voltage dip mitigation in cable network, IN without storage. 72 4.13 Medium voltage cable network with IN and reactors...... 73 4.14 IN consisting of multi-output PE controlled auto transformer. 75 4.15 IN consisting of multiple controlled impedances...... 76 4.16 IN consisting of multiple converters...... 77

5.1 Basic converter operating modes...... 80 5.2 Converter with control system...... 81 5.3 Transfer function proportional resonant current controller. . . 83 5.4 Transfer function proportional resonant voltage controller. . . 84 5.5 Negative-sequence filter αβ reference frame...... 86 5.6 Transfer function DC voltage controller...... 88 5.7 Calculation delay for single-phase r.m.s. detection method. . 89 5.8 Calculation delay for three-phase rectification method. . . . . 90 5.9 Calculation delay for αβ domain detection method...... 91 5.10 Comparison of P and PI voltage control methods...... 93 5.11 Proposed PI controller for voltage dip and swell mitigation. . 93 5.12 Grid during transition from meshed to radial operation. . . . 97 5.13 Frequency density and probability plots of frequency logs. . . 100 5.14 Control topology meshed to radial operation...... 101 5.15 Orthogonal system generator...... 102 5.16 Voltage control two-wire grid connection...... 102 5.17 Positive-sequence filter αβ reference frame...... 103 5.18 Grid during transition from radial to meshed operation. . . . 104 5.19 Control scheme to synchronize load-break switch voltages. . . 105 5.20 Probability of exceeding phase angle difference ...... 109 5.21 Generator supplying load through an impedance...... 110 5.22 Power control scheme after closing of load-break switch. . . . 112

6.1 Pictures of laboratory-scale set-up...... 116 6.2 Single-line diagram of laboratory-scale set-up...... 117 6.3 Example of interface implementation in ControlDesk software.119 6.4 Converter response to step function active power...... 120 6.5 Converter response to step function reactive power...... 121 List of Figures vii

6.6 Converter response to step function voltage amplitude. . . . . 122 6.7 Converter response to load step from 0 to 0.9 p.u...... 122 6.8 DC controller response to load step...... 123 6.9 Angle deviation while synchronization algorithm is active. . . 125 6.10 Three-phase closing of LB2, Area 2 without load...... 126 6.11 Three-phase closing of LB2, Area 2 has 0.9 p.u. load...... 127 6.12 Closing of LB2-B, Area 2 without load...... 128 − 6.13 Detection of LB2-B closing based on I and Iabc,rms. . . . . 129 6.14 Detection of LB2-B closing based on only I−...... 129 6.15 Closing LB2-C, island without load...... 130 6.16 Closing LB2-C, island with load...... 130 6.17 R.m.s. converter currents after closing of LB2-C...... 131 6.18 Required current directions to ensure detection...... 131 6.19 Frequency and voltages LB2 after three-phase opening LB2. . 133 6.20 Opening of LB2-C...... 134 6.21 Disconnecting phases A and B...... 135 6.22 Mitigation of 20 % balanced voltage dip...... 136 6.23 Mitigation of 50 % balanced voltage dip...... 137 6.24 Mitigation of 17 % balanced voltage swell...... 137

B.1 AC current and AC voltage controller responses...... 168 B.2 AC and DC voltage controller responses...... 169 B.3 Three-phase connection of unloaded Area 2...... 170 B.4 Three-phase connection of loaded Area 2...... 171 B.5 Connection of phase B of unloaded Area 2...... 172 B.6 Connection of phase B of loaded Area 2...... 173 B.7 Connection of phase C of unloaded Area 2...... 174 B.8 Connection of phase C of loaded Area 2...... 175 B.9 Three-phase disconnection of unloaded Area 2...... 176 B.10 Three-phase disconnection of loaded Area 2...... 177 B.11 Disconnection of phase C of unloaded Area 2...... 178 B.12 Disconnection of phase C of loaded Area 2...... 179 B.13 Disconnection of phase B of unloaded Area 2...... 180 B.14 Disconnection of phase B of loaded Area 2...... 181 B.15 Mitigation of 12, 15 and 20 % voltage dips...... 182 B.16 Mitigation of 30 and 50 % voltage dips and 17 % swell. . . . . 183

List of Tables

2.1 Classiﬁcation of voltage dip measurement results...... 25

4.1 Simulation results voltage proﬁles...... 70 4.2 Simulation results voltage dip mitigation example network. . 72

5.1 Converter and controller transfer functions...... 87

6.1 Electrical components practical set-up ...... 117 6.2 Converter controller parameters practical set-up ...... 118

Abstract

Flexible distribution systems through the application of multi back-to-back converters: Concept, implementation and experimental veriﬁcation

In recent years the planning and operation of electrical power systems has changed significantly. The unbundling of the formerly integrated generation, transmission, distribution and delivery companies and the growing penetration of DG is increasing the complexity and uncertainty in distribution network planning and operation. Due to the uncertainty, network investments that are done to anticipate load growth or the connection of expected new generators may turn out to be uneconomic. The complexity leads to a higher risk of failures. This stimulates the distribution network operator to consider flexible alternatives to traditional network reinforcements and flexible operation measures. This thesis concerns a power flow control device based on power electronics, called Intelligent Node (IN), which can provide such needs. The idea to control power flow by the application of power electronics is not new in itself. The existing applications are mainly aimed at transmission systems, with its high voltage and power levels and exis- tence of advanced measurement and control systems. In this thesis, an overview is given of existing applications. Successful experiences in the transmission network cannot be applied directly to the distribution system due to its different network topology and operation. For example, the low-inductive character of underground cables, as opposed to the mostly inductive impedance of overhead lines, requires adaptations on the methods to control voltages and power flow in the distribution system. Also the phase-by-phase operation of load-break switches is only found in distribution networks, and the availability of measurement data for all network nodes cannot be taken for granted. The IN consists of multiple converters interconnected on their DC

xi xii Abstract side, and thus the AC side voltages are decoupled. This topology has the ability to control power flow between its AC ports and can supply a radial network section with a controllable voltage. By having the ability to control the power flow it is possible to distribute redundancy over different feeders when needed. In the current practice, every feeder must be able to supply the full load of another feeder, and can therefore only be loaded up to around fifty percent of its power rating. Sharing the redundancy over more feeders allows the connection of loads and generation units beyond this limit. Since the AC voltages on the different IN ports are unrelated, the IN can connect networks with different voltage amplitudes, phase angles and/or frequencies, which makes it possible to also share redundancy in such situations. Controlling the power flow in a meshed network can also be used to optimize voltage profiles, and thus maximize the penetration level of distributed generation units in the network. Alternatively, the power flow can be optimized to reduce losses in the network. During a network disturbance, the IN can prevent spreading of this disturbance, support the disturbed network, temporarily supply part of the network as a radial network, and restore meshed operation after the disturbance. To allow the Intelligent Node (IN) to perform the described tasks, the IN converters need to be able to respond quickly to planned and unplanned events in the power system, such as load changes, short- circuits and the opening and closing of load-break switches. The ability of the converters to do so, depends, besides on their ratings, mainly on the controls that drive them. Furthermore, the protection system of the IN needs to prevent damage to the IN components due to over-currents and over-voltages. At the converter level two basic operating modes exist: power flow control and voltage control. The first operating mode is used in meshed network operation, and called P Q control mode. The converter controls its power exchange with the network by controlling its output current. In the second operating mode, called V control mode, the converter defines the amplitude, frequency and phase angle of the voltage on its AC port. The converter behaves as a voltage source with a fixed frequency and supplies or consumes the active and reactive power as required by the connected loads and generators of a radial network section. In the proposed IN concept, at least one of the converters of the IN is galvanically connected to the ’central grid’, and operates in P Q control mode, in order to supply the connected sections and to control the DC bus voltage. To fully utilize the capabilities of the interconnected converters, the IN control concept also includes specific detection schemes and addi- Abstract xiii tional control and protections, which can change the operating mode and set-points of the converters or shut down the IN in response to power system events. When the power system is in normal operation conditions and the converters are in P Q control mode, the IN controls the power flow in the meshed network according to centrally determined P and Q set-points. During a short-circuit, and the resulting voltage dip, the IN should no longer follow these set-points, but inject reactive power to mitigate the voltage dip. To do so, a control scheme was developed, which is only active when the network voltage is outside of a certain voltage band. Although it is assumed that meshed operation is the normal situation, it might be necessary to operate some parts of the network radially for some time. In order to perform maintenance or repair work on a certain network section, it can be necessary for instance to isolate it by opening the switches on each of its sides. In such a situation, the IN can supply a resulting radial part of the network, with the applicable converter in V control mode. To do so, the applicable IN converter must stop controlling the power flow and start controlling the voltage level instead, after detecting the change in the network. A control and detection scheme was developed to implement this functionality. After the maintenance or repair work has been finished, the radial part of the grid is to be reconnected to the rest of grid. To maintain IN operation and minimize voltage discontinuities after restoring meshed operation, it is necessary that the voltage of the radial network section is synchronized with the voltage of the rest of the network. Therefore, the voltage amplitude, frequency and phase angle are periodically measured at a remote location, and transmitted to the IN with a random but limited time delay. To determine the maximum remote measurement interval, the statistics of frequency variations in the public electricity network have been gathered through measurements. The maximum interval is determined as a function of acceptable phase angle difference between the networks. After detecting that the meshed network has been restored, the applicable converter must be able to change from V control mode to P Q control mode, without disconnecting from the grid or stopping operation. The operation of circuit breakers is in many networks performed simultaneously on all three phases. In other medium voltage networks, for example in the Netherlands, however, the phase-by-phase operation of load-break switches is common, given the wide-spread application of manually operated, compact, epoxy resin insulated, single-phase switchgear. Phase-by-phase connection and disconnection of grid areas requires a different IN behavior. The control and detection schemes were developed both for three-phase and for phase-by-phase switchgear operation. xiv Abstract

Existing back-to-back applications cannot make the described transitions without supply interruption, neither for three-phase nor for phase- by-phase switchgear operation. The developed control and detection schemes are implemented in a laboratory-scale set-up. The main components of this set-up are two 400 V, three-phase converters, connected on their DC sides, with the possibility to connect the AC sides to a radial network with a resistive load or to the public low voltage network. With this set-up, experiments are performed, focusing on the connection and disconnection of network areas and on voltage sag and swell mitigation. The experimental veriﬁcation of the connection and disconnection control and detection schemes, as well as the voltage dip and swell mitigation implementation, shows a successful implementation of the concept. Samenvatting

In het recente verleden zijn de planning en de bedrijfsvoering van elek- triciteitsnetten wezenlijk veranderd. De ontbundeling van de voorheen ge¨ıntegreerde productie-, transport-, distributie- en leveringsbedrijven en de toenemende penetratie van decentrale opwekking veroorzaakt een toename in de complexiteit en onzekerheid in de planning en bedrijfsvoering van het elektriciteitsdistributienet. Tengevolge van de onzekerheid, kunnen investeringen in het net, die gedaan worden om op belas- tingtoename te anticiperen of op het aansluiten van nieuwe opwekkers, oneconomisch blijken. De complexiteit leidt tot een hogere kans op storingen. Dit stimuleert de distributienetbeheerder tot het overwegen van ﬂexibele alternatieven voor de traditionele netverzwaringen en ﬂexi- bele beheersmaatregelen. Dit proefschrift betreft een apparaat voor vermogenssturing dat is gebaseerd op vermogenselektronica, genaamd een Intelligent Knooppunt (IN), dat in zulke behoeften kan voorzien. Het idee om vermogens te sturen met behulp van vermogenselektronica is op zich niet nieuw. De bestaande toepassingen zijn vooral gericht op transportnetten, met hun hoge spannings- en vermogensniveaus en het bestaan van geavanceerde meet- en regelsystemen. In dit proefschrift wordt een overzicht gegeven van bestaande toepassingen. Succesvolle ervaringen in het transportnet kunnen niet direct worden toegepast op het distributienet vanwege de afwijkende netopbouw en bedrijfsvoering. Bijvoorbeeld, het laaginductieve karakter van ondergrondse ka- bels, in tegenstelling tot de meestal inductieve impedantie van boven- grondse lijnen, vereist aanpassingen van de regelmethodes voor spanning en vermogensstroom in het distributienet. Ook het fase-voor-fase bedienen van lastschakelaars komt alleen voor in distributienetten, en de beschikbaarheid van meetgegevens voor alle netknooppunten kan niet als vanzelfsprekend worden aangenomen. De IN bestaat uit meerdere converters die aan hun DC-zijde zijn gekoppeld, en op die manier zijn de spanningen aan de AC-zijden ont- koppeld. Deze topologie heeft de mogelijkheid tot het regelen van de vermogensstroom tussen de AC-poorten en kan een radiaal netwerkdeel

xv xvi Samenvatting voeden met een regelbare spanning. Door de mogelijkheid tot het regelen van de vermogensstroom is het mogelijk om redundantie te verdelen over verschillende strengen, als dat nodig is. In de huidige praktijk moet iedere streng de volle belas- ting van een andere streng kunnen voeden, en kan daardoor slechts tot circa vijftig procent van zijn capaciteit belast worden. Het delen van redundantie over meerdere strengen maakt het mogelijk om belastingen en opwekeenheden aan te sluiten boven deze limiet. Aangezien de AC- spanningen op de verschillende IN-poorten onafhankelijk zijn, kan de IN netten met verschillende spanningsamplitude, -fasehoek en -frequentie koppelen, wat het mogelijk maakt om ook in zulke situaties redundantie te delen. Het sturen van de vermogensstroom in een vermaasd net kan ook ge- bruikt worden om spanningsprofielen te optimaliseren, en op die manier het penetratieniveau van decentrale opwekkers te maximaliseren. Ander- zijds kan vermogensstroomsturing toegepast worden voor het reduceren van verliezen in het net. Tijdens een verstoring in het net kan de IN de verspreiding van deze verstoring voorkomen, het gestoorde net ondersteunen, tijdelijk een netgedeelte als radiaal net voeden en na de verstoring de vermaasde bedrijfsvoering herstellen. Om het mogelijk te maken dat de IN de genoemde taken kan uitvoe- ren, moeten de IN-converters snel kunnen reageren op geplande en onge- plande gebeurtenissen in het elektriciteitssysteem, zoals belastingsveran- deringen, kortsluitingen en het openen en sluiten van lastschakelaars. De geschiktheid van de converters om dit te kunnen, hangt, behalve van hun vermogen, af van de regelingen die hen aansturen. Voorts dient het beveiligingssysteem van de IN schade aan de IN-componenten door overstromen en overspanningen te voorkomen. Op het converterniveau bestaan er twee basis-bedrijfstoestanden: vermogenssturing en spanningsregeling. De eerste bedrijfstoestand wordt toegepast in vermaasd netbedrijf en wordt de P Q bedrijfstoestand ge- noemd. De converter regelt de vermogensuitwisseling met het net door zijn uitgangsstroom te regelen. In de tweede bedrijfstoestand, de V bedrijfstoestand, regelt de converter de amplitude, frequentie en fasehoek van de spanning op zijn AC-poort. De converter gedraagt zich als een spanningsbron met een vaste frequentie en levert of absorbeert het actieve en blindvermogen zoals vereist is voor de op het radiale net aangesloten belastingen en opwekkers. In het voorgestelde IN-concept is tenminste éénvan de converters van de IN aangesloten op het ’cen- trale net’, en bevindt zich in de P Q bedrijfstoestand, om de aangesloten netdelen te voeden en om de DC-railspanning te regelen. Om de mogelijkheden van de gekoppelde converter ten volle te be- nutten, omvat het IN-regelconcept ook de specifieke detectiesystemen Samenvatting xvii en additionele regelingen en beveiligingen, die de bedrijfstoestand en de instelwaarden van de converters kunnen wijzigen of de IN uit kunnen schakelen, in reactie op gebeurtenissen in het elektriciteitssysteem. Wanneer het elektriciteitssysteem onder normale bedrijfsomstandig- heden verkeert en de converters in de P Q bedrijfstoestand zijn, regelt de IN de vermogensstroom in het vermaasde net op basis van centraal bepaalde P en Q instelwaarden. Gedurende een kortsluiting en de resulterende spanningsdip moet de IN niet langer deze instelwaarden volgen, maar blindvermogen injecte- ren om de spanningsdip te verminderen. Hiertoe is een regelsysteem ontwikkeld, dat alleen actief is als de netspanning buiten een bepaalde spanningsband komt. Hoewel de aanname is dat vermaasd bedrijf de normale toestand is, kan het noodzakelijk zijn om enkele delen van het net gedurende enige tijd radiaal te bedrijven. Om onderhouds- of reparatiewerkzaamheden uit te voeren aan een bepaald netdeel, kan het bijvoorbeeld nodig zijn om dit te isoleren door de schakelaars aan de beide zijden ervan te openen. In zo’n situatie kan de IN het resulterende radiale netdeel voeden, met de betreffende converter in de V bedrijfstoestand. Hiertoe dient de IN-converter te stoppen met het regelen van de vermogensstroom en daarvoor in de plaats de spanning te regelen, na detectie van de veran- dering in het net. Een regelen detectiesysteem is ontwikkeld om deze functionaliteit te implementeren. Nadat de onderhouds- of reparatiewerkzaamheden zijn afgerond dient het radiale netdeel opnieuw met de rest van het net verbonden te worden. Om de IN in bedrijf te houden en spanningsdiscontinu¨ıteiten te voorkomen na het herstellen van vermaasd bedrijf, is het noodzakelijk dat de spanning op het radiale netdeel wordt gesynchroniseerd met de spanning op de rest van het net. Hiertoe worden de spanningsamplitude, -frequentie en -fasehoek periodiek gemeten op een locatie op af- stand en naar de IN verzonden met een willekeurige, maar beperkte tijdsvertraging. Om het maximale tijdsinterval voor de meting op af- stand te bepalen, zijn statistische gegevens van de frequentievariaties in het openbare elektriciteitsnet verzameld door middel van metingen. Het maximale tijdsinterval is bepaald als functie van het toelaatbare fase- hoekverschil tussen de netten. Nadat gedetecteerd is dat het vermaasd bedrijf is hersteld, dient de betreffende converter van de V naar de P Q bedrijfstoestand te gaan, zonder de verbinding met het net te verbreken of de bedrijfsvoering te stoppen. Het bedienen van vermogensschakelaars wordt in veel netten tegelij- kertijd op alle drie de fasen uitgevoerd. Echter, in andere middenspan- ningsnetten, zoals bijvoorbeeld in Nederland, is de fase-voor-fase bediening van lastschakelaars gebruikelijk, vanwege de wijdverbreide toepassing van handbediende, compacte, gietharsge¨ısoleerde, enkelfasige xviii Samenvatting schakelaars. De fase-voor-fase koppeling en ontkoppeling van netten vereist een ander IN-gedrag. De regelen detectiesystemen zijn zowel voor driefasige als voor fase-voor-fase bediening van schakelaars ontwikkeld. Bestaande back-to-back toepassingen kunnen de beschreven over- gangen niet maken zonder leveringsonderbreking, noch voor de driefasige, noch voor de fase-voor-fase bediening van schakelaars. De ontwikkelde regelen detectiesystemen zijn ge¨ımplementeerd in een opstelling op laboratoriumschaal. De hoofdcomponenten van deze opstelling zijn twee 400 V driefasen-converters, gekoppeld aan hun DC zijden, met de mogelijkheid tot het koppelen van de AC zijden aan een radiaal net met een weerstandsbelasting of aan het openbare laagspan- ningsnet. Met deze opstelling zijn experimenten uitgevoerd met de nadruk op het koppelen en ontkoppelen van netdelen en op het verminderen van spanningsdips en spanningsverhogingen. De experimentele toetsing van de regelen detectiesystemen voor het koppelen en ontkoppelen, alsook van de toepassing van de vermin- dering van spanningsdips en spanningsverhogingen, laat een succesvolle implementatie van het concept zien. Chapter 1

Introduction

Nowadays society is more than ever dependent on energy, and thus demands for a high reliability of its energy supply. At the same time environmental concerns stimulate a reduced and more sustainable use of energy. The construction of energy-efficient buildings, more efficient transportation methods and the use of renewable energy sources (RES) such as wind and sun are examples of this. Some of these energy sources can be used directly, if they are available where and when needed in the form it is produced, but more often an intermediate energy conversion is required. Electricity is such an intermediate energy carrier. A characteristic of electrical energy is that, once the primary energy form has been converted into electricity, it can be transported and converted to most other forms of energy with a high efficiency. This makes electricity the enabler for many types of renewable developments such as electric vehicles, heat pumps, photovoltaic cells and wind turbines. Because of this, electrical power systems are and will continue to be a key factor to satisfy society’s energy needs. This role of enabler introduces however some challenges. The intermittent character of renewable energy sources and the unrestricted energy trading in the lib- eralized market increase the stress on the electricity networks. On all voltage levels, the power flow becomes less predictable and controllable. Also the balancing of supply and demand is becoming more complicated because of the fluctuating generation. Furthermore, increased efforts are needed to ensure sufficient power quality as experienced by consumers. To meet these challenges in a cost effective way, innovative ways of operating the networks are needed [1]. International research focuses on increasing the flexibility of electricity consumption, using storage to smooth out power fluctuations, and on controlling the power flow using new hardware and software technologies. This thesis is on the control of power flow in distribution networks through power electronics incorporated in the networks.

1 2 Chapter 1

In this introductory chapter, an overview is given of the various changes and the possible steps to be taken in power systems. The spe- ciﬁc research objectives, questions and approach of this thesis are given, as well as how this ﬁts in national and international research programs. The chapter ends with an outline of the thesis.

1.1 Changes in electrical power generation

In the early days of electrical power systems the grid consisted of small island grids with a local balance of load and generation. Later, to improve reliability in an efficient way, interconnections were made between these island grids, creating a larger power system. Due to economies of scale the power rating of individual power plants increased. These two trends caused an evolution of power systems into the current situation, which is characterized by large central power generation plants feeding into networks which span entire continents. In these networks, electrical energy is delivered from the higher voltage levels to the lower voltage levels where the loads are connected. On the primary energy source side, an ever existing concern is the increasing scarcity of fossil energy sources such as oil and gas. Fur- ther there has always been the awareness that being dependent from other countries for energy supply, makes a nation politically dependent, an unwanted situation, especially when the political bonds with those countries are unstable. Diversification of primary energy sources has always been the primary means to mitigate this dependence. As a result of the availability of primary energy sources the electrical power generation mix consists mainly of coal, gas, nuclear power, hydro power and oil. A strong driver to reconsider this generation mix is the growing concern over anthropogenic climate change [2]. The most important man- influenced gas to contribute to global warming is CO2 which is unbreak- ably connected to the use of fossil fuels. This concern has led to a global political intention to reduce the emission of greenhouse gases, written down in the Kyoto Protocol of 1997 [3], after which several international conferences were held on the implementation, the latest one was held in Copenhagen, Denmark [4]. One of the means to reach these targets is to use renewable energy sources, which emit no or less CO2, or which have a short CO2 life cycle. These include biogas, biomass, geother- mal electricity, large and small scale hydro, photovoltaic (PV) and wind power. Some examples of these generation technologies are shown in Figure 1.1. Wave and tidal energy are RES technologies that are still in an infant stage. The European Commission set an indicative objective for the contribution by RES to the European Unions gross inland energy Introduction 3

(a) Combined heat and power.

(b) Solar photovoltaic.

(c) Wind turbines. Figure 1.1: Small and medium scale renewable energy sources. consumption to 12% by 2010 [5] and to 20 % by 2020 [6]. A more specific target, the RES share in electrical energy use, was set to a value of 22% for the European Union as a whole [7]. These targets are planned to be achieved by direct financial support, subsidy schemes, tariff structures, investments in research, etc. Next to RES also combined heat and power systems (CHP) can contribute to a lower CO2 emission by increasing en- 4 Chapter 1 ergy efficiency through the combination of heating and electrical power generation, which can be applied in greenhouses, industrial processes and residential and commercial buildings. On the consumer side, one of the oldest trends in electrical power systems is the steadily increasing energy use per capita per year. While energy use is increasing, society also becomes more dependent on the availability of electrical power. This, and the energy cost becoming a relatively larger part of the expenses, results in an increased customer interest in the efficiency and reliability of electrical power supply. For some consumers this can be a reason to install their own generation units [8]. Because of the mentioned reasons (CO2 emission, energy efficiency, political dependence and fossil fuel scarcity), the penetration level of distributed and renewable energy sources has increased in recent years and is expected to continue to do so in the following years. Several of the mentioned generation techniques depend on energy sources that are distributed by nature, such as wind and sun, or are directly coupled to relatively small scale processes such as residential heating, greenhouses or industrial plants. For this reason, these generators are often also referred to by the term distributed generation (DG). DG units and partly also RES units are relatively small in size and in the 1980s it was already noticeable that the size of newly built generating units started to decrease [9]. In this thesis the term DG will be used rather than RES, since the distributed nature is more important than the environmental aspect when studying the impact of these generators on the electrical power system.

1.2 Changes in the organization of power systems

In the past the generation, transmission, distribution and delivery of electrical energy was centrally coordinated. This meant, for example, that plans for new power plants were developed in close cooperation with network planners. Also the active and reactive power output of power plants was coordinated with the network operator so that overloading and network instability were prevented and minimum losses were achieved. This made the planning and operation of electrical power systems relatively well surveyable. In the unbundling process, which is being implemented across Europe [10], the various tasks and responsibilities are split amongst different parties. For example, in the Nether- lands, the ’Elektriciteitswet’ (Electricity Act) [11] distinguishes the entities of supplier, producer, trader and network operator. For this thesis, the network operator is the most relevant entity. Two types of network operators exist: the distribution network operator (DNO) and the Introduction 5 transmission network operator. The DNO’s main task is to provide and operate a distribution network of sufficient capacity. The transmission network operator is responsible for doing the same for the transmission network and mostly also for providing system services such as matching supply and demand, which is why this role is generally referred to as the transmission system operator (TSO). To fulfill the task of matching supply and demand the TSO organizes a market place where suppliers can offer power generation patterns. In other countries, similar roles and responsibilities have been described.

1.3 Consequences for the distribution network

The consequences of the organizational changes and the changing generation mix for the DNO are both in the network planning phase and in the operational phase. The increasing DG penetration in distribution networks has several consequences, as described by, for example, [12]. The most relevant aspects for this thesis are discussed here shortly, and will be treated in more detail in the next chapter.

Power flow and voltage variations. The power production of many DG units fluctuates. This can, for example, be due to the natural variations of the primary energy source, in case of wind or solar power, or because the electricity production is coupled to the consumer’s heat demand, in case of combined-heat-power installations, or to the electricity market. In the network this results in a fluctuating power flow, and, especially in radially operated networks, in fluctuating voltages. The power flow would normally follow the load profile of the consumers connected to the network and it can be anticipated how to handle the resulting voltage deviations. However, with less predictable, fluctuating, power sources the voltage variations become rather stochastic. In the existing operation of distribution networks only a limited number of signals is measured and of those only a part is recorded. This makes it hard for the network operator to distinguish consumption and generation, and thus to determine accurately where, when and how much power is being consumed and produced. This uncertainty hampers the existing concept of voltage regulation. The so called compounding technique [13] measures voltage and current at the high voltage to medium voltage (HV/MV) transformer terminals and operates the transformer tap changer in such a way that a constant voltage is obtained in a location at some distance from the transformer. Thus, the voltages in the network are kept within a certain band. The lack of knowledge on consumed and produced power can cause more frequent, and non-optimal tap changer operations, result- 6 Chapter 1 ing in increased component ageing. And, much more important, feeders that connect large amounts of DG units in a network that for the rest mainly supplies loads on the other feeders, may experience excessive voltage increase, which cannot be mitigated using the mentioned compounding technique. With inhomogeneously distributed DG units, the compounding technique can even worsen instead of improve voltage profiles. In such situations, the compounding technique must be replaced with an alternative method that determines the optimal tap changer position not only from local measurements, but also from more detailed information on the power flow situation along the feeders. Voltage fluctuations due to changes of the power flow are most serious in radial networks [14]. The obvious way to mitigate these fluctuations is to convert radial to meshed networks. This however decreases reliability because of faults affecting larger network areas, and increases complexity of protection systems.

Efficient network operation versus obligation to connect. The network operator has the duty to provide for any new connection when requested, and to secure the transfer of power to and from that point of connection continuously, but also to operate the network efficiently. However, the planning of size, location and moment of coming online of DG units is done independently of network planning. Together with the already mentioned lack of accurate information on existing load and generation in the network, the planning of network expansions and reinforcements has become more difficult and risky. Using standard technologies, the network operator can choose to invest in advance and risk stranded investments, or to refuse new connections, or to accept new connections and take the risk of losing redundancy. None of these options is ideal and new innovative options are worth considering.

No more ﬁt and forget. Most studies conﬁrm, that the local penetration of distributed generation up to a level of about 20% of the maximum load can be absorbed by the electricity distribution network without major costs [15]. The penetration level in many networks is still below this limit, but this will change. In situations where more DG is being connected, measures are likely needed [8, 16].

1.4 Ongoing research

For several years there has been international research on what should be the right direction for power systems to develop in this context. The European Commission has initiated several research programmes on the topic of integration of DG and on network operation. To study fu- Introduction 7 ture system concepts programs like FENIX [17], MICROGRIDS and MOREMICROGRIDS [18, 19], and EU-DEEP [20, 21] were initiated. In Japan such research was performed within the FRIENDS project [22, 23]. More speciﬁcally targeted to the level of the distributed generation itself are the programs DISPOWER [24] and DGFACTS [25]. Several of these research programs are part of the IRED-cluster [26]. To converge the outcomes of all these programs and develop new steps ahead, the Eu- ropean Technology Platform for the Electricity Networks of the Future, in short ’ETP SmartGrids’ [27], was installed. Also in the United States of America research programs are ongoing, such as GRIDWISE [28], In- telliGrid [29], GridStat [30] and a number of projects under the Electric Distribution program [31]. As an illustration of the history of international attention for the main research topics, Figure 1.2 shows the

distributed generation electrical energy storage systems microgrids power electronics smartgrids

1950 1960 1970 1980 1990 2000 2010 t (y) Figure 1.2: Google Timelines results for selected search terms (vertical scale is different for each line). results of a comparison of Google Timeline results. The lines show the relative amount of internet content on the different topics which can be traced back to a certain year. Especially ”smartgrids” is a buzzword that is growing strongly at the moment of writing, while, for example, ”distributed generation” is already on its return. As a result of the (partly still ongoing) research a consensus is developing that the passive network operating must be transformed into active network management [32]. The academic and industrial efforts to implement this transition have resulted in a wide range of possible methods. These methods can be divided into two categories. The first 8 Chapter 1 category is based only on the use of communication and automation technology to control the network and the connected load and generation. These methods are considered to be the first steps towards smarter grids and experience is being built up world wide [33]. The second category uses, besides the technologies from the first category, also electrical power equipment with the ability to influence power flow and power quality. In the following an overview is given of the current status of the most important methods.

1.4.1 Communication and automation Communication technology applications are developing rapidly in several directions such as broadband wired and wireless internet services, satel- lite communication, power line communication etc. These technologies allow fast and synchronized exchange of measurement and control signals between, for example, a central control system, the on-load tap changer (OLTC) of the utility’s transformer, a consumer’s electrical hot water boiler and a producer’s wind turbine converter to name a few. Since the introduction of phasor measurement units (PMU) in the eighties of the twentieth century [34] it is possible to synchronize power systems measurement data with a high accuracy. The standardization of communication protocols is developing rapidly and is resulting in a family of standards around IEC61850 [35]. Also the development of smart meters already enables some DNOs to gain detailed information on consumption patterns [36]. Furthermore, smart meters have the potential to be the intermediary between the consumers’ installations and the distribution system operator. The mentioned developments in communication and automation are an important enabler for a more active role of loads and DG in the operation of power systems.

1.4.2 Load control Traditional demand-side management (DSM), which has been implemented internationally for many years, uses fixed, nation-wide tariff patterns. This reduces the power system peak load, losses, generation costs and prevents or postpones power system reinforcements. This method is effective because consumers, at least when aggregated in large groups, have a more or less constant and predictable response to tariff patterns when looking at yearly, weekly and daily consumption patterns. With the introduction of large amounts of intermittent and uncontrollable amounts of generation the current fixed tariff patterns may become an obsolete technique. In the extreme situation of a network with mainly uncontrollable generation, the generators can, by definition, no longer Introduction 9 be controlled to match the demand. Instead, the loads must be controlled to match the available supply. This is a paradigm shift that is very important for the further development of power systems. In the ADDRESS project [37] a commercial and technical framework is being developed to make this paradigm shift possible. Not all loads are suitable to be controlled. Mainly load types which are associated to any kind of storage are relatively well controllable without affecting the consumer’s comfort or process. Examples of such loads are freezers, refrigerators, water boilers, space heating and air conditioning and industrial processes which use heat, cold or compressed air. Here, electric vehicles must be mentioned as a type of load, with a high potential as far as controllability is concerned [38]. E-vehicles have an intrinsic large energy storage capacity (20 to 100 kWh), and a large power output (30 to 50 kW), when compared with typical power rating of low voltage equipment. Above, load control was only discussed to support the matching of supply and demand. Another application of load control, other than to match supply and demand, is to solve network problems [39, 40]. This application is discussed in paragraph 1.4.6 on active distribution networks.

1.4.3 Generation control Controlling the active and reactive power output of generators in response to system conditions is the basis for today’s power system stability. Two main controls exist: frequency regulation by controlling active power output and voltage regulation by controlling reactive power output. Currently, DG units based on non-controllable energy sources are excluded from the obligation to participate in frequency and voltage regulation. Therefore, most DG operators run their units at zero reactive power output, in order to maximize the active power output. In a situation where the majority of energy is supplied by intermittent sources, it becomes necessary to also control those generators. For generators with an intermittent availability of the primary energy resource, such as wind and solar power, the possibilities are however limited. Active and reactive power control possibilities can only be achieved when the DG units operate below their maximum active power output. To reach a cost effective power system operation, the ancillary services of active and reactive power control by DG could be offered in a market. Analogous to load control, also generation control can be used to solve network problems, in addition to the discussed balancing of supply and demand. Various methods are being developed to control voltage profiles [41–44], but also to avoid network overloading [45] or to create a local balance between demand and supply [46]. These applications 10 Chapter 1 can be used as a component of active distribution networks, see also paragraph 1.4.6.

1.4.4 Storage In any market, storage reduces price spikes due to temporary changes in supply and demand and increases the availability and reliability of a commodity. Currently, electrical storage is not used on a large scale, which makes it essential to have a momentary balance between supply and demand. Because of this, production must be capable of supplying the peak demand instantaneously, which is costly. One way to reduce the peak demand is to control loads, as discussed previously. Another way is to supply the peak load from temporary energy storage installations that charge during off-peak hours. With the advancement of DG, a part of generation units supply an amount of power, independent of the actual balance of supply and demand, which increases the stress on the rest of the electricity production plants, the network and therefore the market. Storage can help matching supply and demand by providing frequency support. Other ancillary services that storage can provide, are voltage regulation, flicker reduction and reliability improvement. Stor- age is also proposed to facilitate the implementation of DG, by solving local network constraints that occur due to the intermittent nature of the DG output [40, 47]. The most mature storage application today is pumped hydro, which is mainly used to counteract the effect of the fluctuating output of wind farms. The other mature, but used only very limited, technology with this capability is compressed air energy storage (CAES). Other storage technologies are various types of battery systems (mature: lead-acid, sodium sulphur), redox installations and flywheels [48]. A type of storage which may become interesting for distribution networks is formed by electric vehicles, as mentioned already in paragraph 1.4.2.

1.4.5 Power electronics A technological breakthrough that allows a flexible control of electrical power is the advancement in power electronics. The oldest power electronics devices for power systems are applied in transmission systems and are based on thyristors. The two main applications are the high voltage direct current (HVDC) technology, which is used to transmit power over large distances, and the static var compensator (SVC), which is used to control voltage by injecting reactive power. These devices are characterized by low switching frequencies and large filters. De- velopments of the solid state switching elements have resulted in higher voltage and current ratings and higher switching frequencies [49]. This Introduction 11 enables the use of converter based topologies instead of thyristor based devices [50]. For distribution systems the main applications concern the control of voltage and power flow but also the improvement of power quality. A future price reduction of the semiconductor switching elements is expected to strengthen the business case for such distribution system applications of power electronic devices [51, 52]. Also the development and standardization of modular products is needed to enable the large scale deployment of these devices in distribution systems [53].

1.4.6 Active distribution networks As part of a larger vision on the future of electrical power systems, the ETP SmartGrids has proposed the concept of active distribution networks [54]. This concept can be considered the integration of the topics discussed above: communication and automation, load and generation control, storage and power electronics. In the MicroGrids, MoreMicro- Grids and FRIENDS projects [18, 19, 22, 23] a forerunner of this concept was developed, developing a concept in which a certain network area is operated autonomously and that is in principle independent of the rest of the network. Such a microgrid may or may not be connected to a larger network. When connected to a larger network, the microgrid concept can be applied to reach an optimum situation, taking into account the market price of electricity, the cost of local production, energy losses and greenhouse gas emission [55]. During a blackout in the larger grid, the microgrid concept makes it possible to operate as an island, implementing black start capabilities, thus increasing security of supply [56]. In the current regulation scheme, it is the DNO’s role to provide suﬃ- cient network capacity to supply load and connect generators. Reducing loads or limiting generators to solve network problems is in clear contra- diction with this role. To inﬂuence supply and demand on a local scale requires the establishment of a local market [57]. This requires drastic changes in regulation, but even more so in the consumer’s behavior and their willingness to participate in such a scheme.

1.5 Research objective

Meshed distribution networks are mostly operated radially to prevent geographic spreading of disturbances, and to limit short-circuit currents. A consequence is that grid faults require time consuming operations to restore power delivery through other routes. Also, growing penetration levels of DG cause fluctuations in the power flow situation and voltage profiles. Operating networks in a meshed way mitigates these effects and reduces losses, but also makes the network more vulnerable, and more 12 Chapter 1 complicated to operate and protect. The interconnection of grid areas is sometimes not possible because of too large voltage differences. Having the ability to couple networks in a controlled manner would combine the benefits of radial and meshed networks: limited geographic spreading of disturbances, the ability to control power flow and voltages and the online availability of an alternative supply path. The objective of the here presented PhD research project is to investigate the use of a multi back-to-back converter to support the transition to active distribution networks. The investigation must explicitly not focus on finding the optimal power electronics topology, but rather on defining the role that such a device can play in a power distribution system and on the proof of principle. The possible tasks of this device are to actively control power flow and to maintain power quality and stability, both during normal operation and during fault conditions in medium voltage networks with distributed generation. The combination of this versatile converter system and the control and protection systems that define its functionality are in this thesis denoted as the Intelligent Node (IN).

1.6 Research questions

To contribute to this objective, the following questions need to be an- swered:

1. Which are the main beneﬁts of multi back-to-back converter devices in distribution systems?

2. How and under which conditions can these beneﬁts be achieved? What controls should be adopted?

3. Which are the speciﬁc aspects that distinguish the application of the multi back-to-back devices in distribution systems from similar applications in transmission systems?

4. Can this be demonstrated on a laboratory-scale set-up?

1.7 Research approach

The research approach consists of the following three steps.

Development of the Intelligent Node concept. To develop the IN concept, the various possible benefits are defined and evaluated. Based on this evaluation, the primary benefits are distinguished, which are those benefits that could be enough reason to implement this concept. Introduction 13

Also secondary benefits are discussed, which can be achieved once such a concept is applied, but which are considered to not be enough reason to choose for the concept. Based on the defined concept, the required functions are specified on a more detailed level.

Implementation in a practical set-up. The various functions are implemented into a control system for the Intelligent Node as a whole and for each of the converters. These controls are developed in Matlab Simulink. In the laboratory a practical set-up is realized, which consists of two low voltage converters, connected on their DC sides. The developed control system is implemented in this set-up.

Verification by experiments. Using the described set-up, the innovative aspects of the IN concept are experimentally verified. The experimental results are compared with results from calculations. Based on the experimental results, the effects of scaling the developed concept to medium voltage levels are analyzed.

1.8 IOP-EMVT programme

The research presented in this thesis was performed within the framework of the ’Intelligent Power Systems’ research project. The project is part of the IOP-EMVT programme (Innovation Oriented research Programme, Electro Magnetic Power Technology), which is financially supported by SenterNovem, an agency of the Dutch Ministry of Eco- nomic Affairs. The ’Intelligent Power Systems’ project is initiated by the Electrical Power Systems and Electrical Power Processing groups of Delft University of Technology and the Electrical Energy Systems and Control Systems groups of Eindhoven University of Technology. In total 10 PhD students, of which 9 have already finished their work, are involved in the project. The research focuses on the effects of the structural changes in generation and consumption that are taking place, like for instance the large-scale introduction of distributed (renewable) generators. Such a large-scale implementation of distributed generators leads to a gradual transition from the current vertically-operated power system, which is supported mainly by several big centralized generators, into a future horizontally-operated power system, having also a large number of small to medium-sized distributed (renewable) generators [58]. The project consists of four parts, which is illustrated in Figure 1.3. The first part, ’inherently stable transmission system’, investigates the influence of uncontrolled decentralized generation on the stability and dynamic behavior of the transmission network. As a consequence of the transition in the generation, less centralized plants will be con- 14 Chapter 1

Inherently Stable Transmission System

Self- Optimal Manageable Controlling Power Distribution Autonomous Quality Networks Networks

Figure 1.3: Structure of the IOP-EMVT Intelligent Power Systems project.

nected to the transmission network as more generation takes place in the distribution networks, whereas the remainder is possibly generated further away in neighboring systems. The investigated solutions include the control of centralized and decentralized power, the application of power electronic interfaces and monitoring of the system stability. The second part, ’manageable distribution networks’, focuses on the distribution network, which becomes ’active’. Technologies and strategies have to be developed that can operate the distribution network in different modes and support the operation and robustness of the network. The project investigates how the power electronic interfaces of decentralized generators or between network parts can be used to support the grid. Also the stability of the distribution network and the effect of the stochastic behavior of decentralized generators on the voltage level are investigated. The research described in this thesis is within this part of the programme. In the third part, ’self-controlling autonomous networks’, autonomous networks are considered. When the amount of power generated in a part of the distribution network is sufficient to supply the local loads, the network can be operated autonomously but actually remains connected to the rest of the grid for security reasons. The project investigates the control functions needed to operate the autonomous networks in an optimal and secure way. The interaction between the grid and the connected appliances has a large influence on the power quality. The fourth part, ’optimal power Introduction 15 quality’, analyzes all aspects of power quality. The goal is to provide elements for the discussion between polluter and grid operator who has to take measures to comply with the standards and grid codes. Setting up a power quality test lab is an integral part of the project.

1.9 Outline of the thesis

This introductory chapter is followed by the following chapters:

Chapter 2 Distribution systems The relevant aspects of distribution systems are discussed. This includes power quality considerations, redundancy aspects, and voltage control methods.

Chapter 3 FACTS in distribution systems After a description of power ﬂow control principles, an overview is given of the state-of-the-art of power electronic applications for power ﬂow control in electrical power systems.

Chapter 4 Functional concept of the Intelligent Node The proposed functional concept for the application of a multi back-to-back converter device as IN is given here. The applications are developed on a functional level, with the device at a black-box level. To satisfy the resulting requirements several technology options are analyzed, ending with the versatile multi back-to-back converter topology.

Chapter 5 Intelligent Node control and protection In this chapter, the developed applications are translated into control schemes for the individual converters. To ensure proper implementation of the concept, together with the control concept, also a protection concept is developed, which is tailored to the speciﬁc characteristics of the Intelligent Node.

Chapter 6 Laboratory-scale demonstration To demonstrate the concept and controls as described in the previous chapters, a three-phase laboratory-scale set-up is realized. In this set-up experiments are carried out, which focus on the innovative parts of the IN concept.

Chapter 7 Conclusions The thesis ends with general conclusions and recommendations for future research. 16 Chapter 1

1.10 Publications

The results of the research presented in this thesis have been presented in the following conferences and journals: • R.A.A. de Graaff, J.L. Duarte, W.L. Kling, P.T.M. Vaessen, ”Pha- se-by-Phase Connection and Disconnection of Grid Areas using Multi Back-to-Back Converters”, IEEE Tr. Power Delivery (submitted for review) • R.A.A. de Graaff, J.L. Duarte, W.L. Kling, P.T.M. Vaessen, ”Flex- ible Framework for the Operation of Distribution Networks - Syn- chronizing and Connecting Grid Areas using Multi Back-to-Back Converters”, European Tr. on Electrical Power (under review) • R. de Graaff, J. Duarte, W.L. Kling, P. Vaessen, ”Intelligent Nodes in Distribution Systems - Transition from Radial to Meshed Op- eration”, Proc. CIRED 2009, June 8-11, 2009, Prague, Czech Republic • R.A.A. de Graaff, J.M.A. Myrzik, W.L. Kling, J.H.R. Enslin, ”In- telligent Nodes in Distribution Systems - Optimizing Steady State Settings”, Proc. PowerTech 2007, July 1-5, 2007, Lausanne, Swit- serland • R.A.A. de Graaff, J.M.A. Myrzik, W.L. Kling, J.H.R. Enslin, ”Intelligent Nodes in Distribution systems - Operating Concept”, Proc. CIRED 2007, May 21-24 2007, Vienna, Austria • R.A.A. de Graaff, J.H.R. Enslin, ”Profitable, Plug and Play Dis- persed Generation: The Future?”, Leonardo Energy Digest, Vol. 3, No. 1, 2007 • R.A.A. de Graaff, J.M.A. Myrzik, W.L. Kling, J.H.R. Enslin, ”Se- ries Controllers in Distribution Systems - Facilitating Increased Loading and Higher DG Penetration”, Proc. Power Systems Con- ference and Exposition 2006, October 29 - November 1, 2006, At- lanta, Georgia, USA • R.A.A. de Graaff, J.M.A. Myrzik, W.L. Kling, J.H.R. Enslin, ”Se- ries Controllers in Distribution Systems - Facilitating Increased Loading”, Proc. Universities Power Engineering Conference 2006, September 6-8, 2006, Newcastle upon Tyne, UK • R.A.A. de Graaff, J.M.A. Myrzik, W.L. Kling, J.H.R. Enslin, ”Se- ries Controllers in distribution systems - A survey of benefits in relation to DG”, Proc. Conference on Future Power Systems 2005, November 16-18, 2005, Amsterdam, the Netherlands Chapter 2

Distribution systems

The overall, functional requirements for a properly designed and operated power system are [59]: 1. The system must be able to meet the continually changing load demand for active and reactive power, without overloading equipment. 2. The system should provide energy at minimum cost and with minimum ecological impact under market conditions. 3. The quality of power supply must meet certain minimum standards with regard to constancy of frequency, constancy of voltage and reliability The physical infrastructure of power systems is developed to meet these requirements. The operation of this infrastructure is based on market rules for matching supply and demand and uses supervisory control for the interconnected power system and separated operation of its subsys- tems. For the network two operation levels exist: transmission system and the distribution system. This thesis concerns distribution systems. In this chapter, those aspects of distribution systems are discussed that are relevant to the Intelligent Node concept and its implementation. First the legal framework that deﬁnes the network operator’s role is given, then reliability and redundancy considerations are treated, followed by power quality aspects and implemented voltage control methods. Finally, the chapter concludes with a summary of the challenges DNOs are facing.

2.1 The network operator’s role

In Europe the legal entity of the network operator is by law separated from the legal entities of the commercial roles of supplier, producer and

17 18 Chapter 2 trader [60]. As an example of the role and responsibility of the network operator, the Dutch situation is described here. In the Dutch Electric- ity Act [11] the responsibilities of the DNO are deﬁned, of which the following are the most relevant ones for this thesis. DNOs should:

• Operate and maintain the network in a certain region (voltage level < 50 kV).

• Ensure the safety and reliability of their networks and the transmission of electricity in the most eﬃcient way.

• Construct, repair, renew or expand their network, while taking into consideration measures regarding sustainable electricity, energy saving, demand management or distributed generation by which the need to replace or increase generation capacity can be prevented.

• Maintain enough spare capacity to transmit electricity. • Connect third parties to the network within a reasonable time (which is at most 18 weeks for connections up to 10 MVA).

The TSO has, in addition to these tasks for voltage levels > 50 kV, also the responsibility to take measures and perform system services needed to ensure the transmission of electricity through all networks in a safe and eﬃcient manner. Also the coordination of measures after large-scale black-outs is done by the TSO. The power threshold of 10 MVA is higher than the power rating of most DG units connected to distribution networks. In such situations, the DNO has therefore only a limited time to take adequate measures in its network, if needed. The connection of a single generator may not require any measures, but in areas which are favorable for DG multiple connection requests can be expected at very short notice. In such a situation, the DNO may need to take measures in the distribution network and this can easily need more time than this period [16]. Therefore, grid reinforcements must be made in advance, based on estimates of future connection requests. If these connection requests do not come, grid investments turn into stranded costs. This is against the DNO’s task to transmit electricity in the most eﬃcient way. It is important to note that in the current legal and regulatory framework the DNO cannot oblige loads and generators to support the operation of the distribution network. This control feature, also called active demand side participation [37], for the moment can only be applied for matching generation and loads by market parties [11, 60]. This can actually increase the coincidence of loads and thus, instead of solve constraints, cause additional stress in the distribution system. Distribution systems 19

2.2 Network topology and redundancy

The reliability of a distribution network depends on the reliability of its individual components, the topology of the network, the loadability, the protection system, the operating concept and on external factors such as digging activities and lightning [61]. Here, the choices made in the network topology, protection system and operating concept are discussed.

2.2.1 Network topology Three basic distribution network topologies exist: radial, ring and meshed [62], as shown in Figure 2.1. The radial topology is characterized by

MV/LV transformer LV connection MV cable LV cable

(a) Radial (b) Ring (c) Meshed

Figure 2.1: Types of medium and low voltage grids. only one possible supply path for each load, no redundancy exists. In the ring and meshed topologies at least two supply paths exist, which leads to higher reliability. Operating a network as a ring or meshed requires distance or zone protection and more switchgear to ensure that only the faulted section is switched oﬀ. Ensuring adequate settings for all network conditions for large numbers of distance protection relays complicates network operation. In radial networks the simpler maximum 20 Chapter 2 current-time protection can be applied. By opening switches, a ring or meshed network can be operated radially. This gives both the beneﬁt of a simple protection system and the availability of an alternative supply path (after switching over).

2.2.2 Redundancy The duration of the outage experienced due to a disturbance in a network (fault) is determined by the time needed to clear the fault, the availability of an alternative supply path and on the time needed to re- configure the network. In radial networks, no alternative supply path exists and the supply interruption time is equal to the time needed to repair the faulted component. In radially operated ring or meshed networks, there is an alternative supply path. To allow reconfiguration, sufficient capacity must be available on the alternative supply path. To provide this capacity, networks are normally not loaded up to their nominal rating. In the symmetric situation shown in Figure 2.2 this results

50% a

b 50%

c 50%+50%=100% Coupling busbar d Sum=200% Figure 2.2: Each feeder can supply the load of another feeder. in a maximum loading level of each feeder of 50 %. By operating the load-break switches (LBs) on the right, the diﬀerent feeders can be interconnected through the coupling busbar. Also the transformers are not loaded above half of their power rating. The sequence of events that follows a fault is illustrated and described in Figure 2.3. The shown steps are typical for underground cable networks, which are characterized by mostly permanent faults. In overhead line networks 50 to 80 % of faults are temporary [61], since they are caused by lightning. This allows faster restoration of supply on disconnected feeders. To achieve this, after step c), the circuit breaker of feeder c is reclosed automatically after some seconds. If the fault then still exists, after a delay the circuit breaker opens again and the sequence of events is continued with step d) and further. The reclosing process can be limited to one single attempt, or can be repeated several times. If the fault has disappeared, the Distribution systems 21 reclosing brings the network immediately back to situation a). This pro- cedure is called automatic reclosing. In this thesis, the Dutch situation is assumed, where the radially operated meshed medium voltage (MV) and radial low voltage (LV) networks consist entirely of underground cables [63], without automatic reclosing.

2.3 Power quality aspects

Electrical equipment and processes connected to the power system can only withstand limited deviations of the supply voltage from its nominal parameters. These deviations are quantiﬁed using several power quality indicators. The IN can improve the power quality level in a network, for example, by injecting reactive power in the network during a fault, by controlling the network voltage, or by compensating harmonic currents. During switching actions in the network, the IN plays an important role to ensure that the voltage stays within acceptable limits. In the following paragraphs, the power quality indicators that are relevant in this thesis are discussed: steady state voltage amplitude, ﬂicker, voltage dips, phase angle jumps and power frequency.

2.3.1 Steady state voltage amplitude The commonly used standard EN50160 [64] gives a characterization of expected values in normal operation, excluding ’abnormal operating conditions’. These ’abnormal operating conditions’ include, for example, ”a temporary supply arrangement to keep the network users supplied during condition arising as a result of a fault, maintenance and construction work or to minimize the extent and duration of a loss of supply”. Explic- itly it is stated that ”the voltage characteristics (...) are not intended to be used as electromagnetic compatibility (EMC) levels or user emission limits for conducted disturbances in public distribution networks”. Re- cently, the organization of European regulators ERGEG has proposed to transform the current characterization into compatibility limits [65–67] to protect customers interests better. Currently EN50160 states that in medium voltage systems 95 % of all 10 min mean r.m.s. measurements of the supply voltage within a week can be expected to be within 10% of the declared voltage. For low voltage networks the same applies± and additionally, all 10 min mean r.m.s. values of the supply voltage can be expected to be within the range of +10 % and 15 %. − 22 Chapter 2

a a) The power system is in b normal operation, all feeders are operated radially. c

a b) A permanent short- b circuit occurs on feeder c c.

a c) The circuit breaker of b feeder c opens. The loads on feeder c are unsupplied. c

a d) The fault is isolated b by opening the load-break switches of the aﬀected sec- c tion.

a e) The LBs of feeders b and b c at the coupling busbar are closed, the circuit breaker of c feeder c is reclosed. Feeder c is supplied through feeder b, all loads are supplied again.

a f) The line section is re- b paired.

a g) The two load-break b switches of the repaired section are closed and feed- c ers b and c are in meshed operation.

a h) By opening the two b switches at the coupling busbar the radial operation c of situation a) is restored. Figure 2.3: Sequence of events during a short-circuit. Distribution systems 23

Anticipating a future restriction of EN50160 following ERGEG’s recommendations, in this thesis the following is defined with respect to voltage magnitude limits: ”Network companies shall ensure that supply voltage variations (r.m.s. value) are within an interval of 10 % of the declared voltage, measured as a mean value over one minute,± in points of connection in the low voltage network, both during normal operation and during maintenance and temporary supply arrangements that are made to prevent or reduce loss of supply.” This text is the translation of paragraph 3.3 of the Norwegian power quality regulation [65, 68], with a scope extension for the network conditions. This scope extension is made based on the observation that if ”a certain condition causes voltage deviation that may lead to damage for electrical equipment, than for the customer it is better to experience an interruption” [65]. For the MV network the interval is assumed to be 5 % (see further paragraph 2.4). ± 2.3.2 Flicker Incandescent light bulbs and other electrical light sources are sensitive to voltage variations. A change of the voltage amplitude results in a change of the luminance of the light source, which brings annoyance to human beings. The human brain responds in a complex non-linear way to voltage variations of different repetition frequencies and shapes. To quantify the annoyance observed by a human being the parameters short and long-term flicker are used. Flicker is measured using the flickermeter, which is defined in IEC61000-4-15 [69]. With the phasing-out of incandescent light bulbs [70] and the increasing use of alternative light sources such as energy saving lamps and LEDs, this flickermeter needs revision [71]. Energy saving lamps and LEDs are less sensitive to voltage variations. As a worst-case approach, the thresholds and methods developed for the incandescent lamp are used in this thesis. For this, the curve that indicates the annoyance threshold can be used. For rectangular voltage changes, this curve is shown in Figure 2.4.

For non-repetitive events, the following analytical method can be used. For a series of events the short-term flicker level is equal to Σt 1/3.2 P = f (2.1) st T p with the flicker impression time in seconds t = 2.3 (F d )3.2 (2.2) f · max and the interval Tp equal to 600 s [72]. According to these equations the maximum rectangular voltage amplitude change (F = 1), due to non- repetitive events such as network reconfiguration or load connection, 24 Chapter 2

101

0 (%) 10 d

10−1 10−1 100 101 102 103 number of voltage changes per minute

Figure 2.4: Curve for Pst = 1 for rectangular equidistant voltage changes with amplitude d [72].

can have a maximum amplitude dmax of 5.7 %. The maximum relative steady-state voltage change is however limited to 3.3 % according to [72], while the Dutch grid code uses 3 % as the maximum amplitude of fast voltage variations. On the other hand, the mentioned power quality limits for flicker are not applicable in all network conditions. For example, the Dutch grid code excludes situations where there is loss of production, large loads or lines. In this thesis it is assumed that for deliberate network reconfiguration actions the limits for normal operations apply. For the IN this means that the rate at which the power flow in the network is changed must be limited, to avoid excessive flicker levels.

2.3.3 Voltage dips A voltage dip is defined as a ”sudden reduction of the voltage at a particular point on an electricity supply system below a specified dip threshold followed by its recovery after a brief interval” [73]. The voltage thresholds are equal to 90 % and 1 % of the declared voltage. Conventionally the duration of a voltage dip is between 10 ms and 1 minute. The depth of a voltage dip is defined as the difference between the minimum r.m.s. voltage during the voltage dip and the declared voltage. The residual voltage is the minimum r.m.s. voltage during the dip. In this thesis both the term ’depth’ and ’residual voltage’ are used. Voltage dips occur due to a number of reasons, such as short-circuits, the energization of transformers or the connection of heavy (inductive) loads, and cannot be prevented entirely. The limits for acceptable voltage dips are not well defined in international standards and only indicative values of occurring residual voltage, duration and frequency of occurrence are given [64]. Table 2.1 can be used to classify measured dips. In the table the commonly used ITIC curve [74] is shown, which is an indication of the voltage dip immunity of computers and other digital devices. Network Distribution systems 25

Table 2.1: Classiﬁcation of voltage dip measurement results.

td (s) 0.01 0.02 0.1 0.5 1.0 2.0 5.0 10 20 ------V (%) 0.02 0.1 0.5 1.0 2.0 5.0 10 20 60 80 - 90 ITIC 70 - 80 60 - 70 50 - 60 40 - 50 30 - 40 20 - 30 10 - 20 1 - 10 reconﬁgurations in a network with an IN can cause voltage dips. To assess the impact of such voltage dips, in this thesis the voltage dips that are measured and calculated are compared with the ITIC curve.

2.3.4 Phase angle jumps Phase angle jumps are not a frequently used power quality indicator and little information exists on equipment immunity. Although extensive research was performed to qualify phase angle jumps as a parameter of voltage dips [75], this has not resulted in agreement on more general phase angle jump immunity levels. In the IN concept, phase angle jumps can occur, and therefore here an attempt is made to determine acceptable values. In order to do so, other events where phase angle jumps occur are examined: the synchronization of a synchronous generator and the connection of two grid areas. Also, circuit breaker specifications are examined. In case of generator synchronization, the voltage angle difference is the most critical factor, since it causes a transient torque on the generator shaft after connection, which has to be limited. The sensitivity of generating units to phase angle jumps can be deduced from the threshold values for synchronization and the functioning of protection systems. Exemplary values of 10° and 20° for synchronization [76, 77] are reported. When connecting two grid areas, it is the voltage angle difference which is the most important parameter: in cable grids with a low X/R ratio, the transformer inductive impedance is the dominant impedance limiting the current after closing the circuit breaker. In [78] for a medium voltage distribution system the maximum allowable voltage angles were calculated to prevent overloading of the concerned feeder. Threshold 26 Chapter 2 values of a few degrees were found. Another factor limiting the phase angle difference on both sides of the circuit breaker is the circuit breaker itself. General purpose circuit breakers are tested with voltages up to 90° phase difference [79], which is a rather high value when compared to other limiting factors. From the above, it is observed that connected loads and generators are already exposed to phase angle jumps of 5°, 10° and even 20°. No compatibility problems are known in relation to these phase angle jumps. From this, it is concluded that phase angle jumps up to at least 10° are acceptable for loads and DG units. This limit is used in the development and implementation of the IN concept, where it is a critical parameter in the connection of different network areas.

2.3.5 Power frequency The frequency of the supply voltage is controlled by generators, which adjust their power output accordingly when the network frequency changes and thus balance out supply and demand deviations [80]. In large interconnected power systems, such as the European ENTSO-E networks, the resulting frequency fluctuations are small. For 99.5 % of the time the power frequency should be within 1 % of the nominal frequency and during 100 % of the time within +4± % and 6 %. In island networks larger frequency variations are allowed: 2 %− during 95 % of the time and 15 % during 100 % of the time [64]. In± the EMC standard IEC61000-2-2± [81] a compatibility level of 1 % is given for temporary frequency variations. For the implementation± of the IN concept, the power frequency characteristics are used as an input for the definition of suitable threshold values to detect switching actions during network reconfiguration.

2.4 Voltage control

In typical cable distribution networks the only available technique to control the voltage at the low voltage supply terminals is the operation of the HV/MV on-load tap changers (OLTC). The OLTC is operated based on local measurements of the loading of the connected feeders. This information is meant to compensate the voltage drop and rise along the feeder by means of a voltage oﬀset [82]. This method is called compounding. The tap changers of the MV/LV transformers are of the oﬀ-load type, and cannot respond to voltage changes. Voltage drops in the low voltage network add to the voltage drops in the MV network, that is why voltage variations in the LV network are larger than in the MV network. Figure 2.5 shows typical voltage variations in a network Distribution systems 27

HV MVb MVe LVb LVe

PSfrag replacemen 1.05 1.00

0.95

0.90 Voltage variation (p.u.)

Figure 2.5: Typical voltage coordination in a radially operated MV/LV network [83].

with a radial structure in the MV and LV parts of the network, assuming only loads. In this thesis these typical values for voltage bands are an input for the application of the IN to control voltage proﬁles in the distribution network, which is elaborated in paragraph 4.2.

2.5 Conclusion

In the current regulatory environments the distribution network operator is faced with some contradictory responsibilities. On one side, the DNO must provide and operate a network of sufficient capacity in an economical way. On the other side, the DNO must provide new connections at very short notice. Anticipating these connections by strengthening the network in advance, brings uncertainty in the return on investments. Furthermore, the expected control of load and DG units for balancing supply and demand, and thus reducing generation costs, stresses the distribution system, but is out of the DNO’s sphere of control. The application of an IN can support the DNO here, as will be explained in Chapter 4. The Intelligent Node can influence some power quality parameters, such as flicker, voltage dips and phase angle jumps. Currently, for many of these parameters no binding compatibility levels exist, but steps are made towards this. For each of the relevant PQ indicators, the limits which are used in this thesis to assess the influence of the IN are defined. The power quality parameters steady state voltage amplitude and power frequency are discussed as they are input for the development and implementation of the IN concept. Finally, the currently implemented voltage control technologies are discussed, as well as the resulting typical voltage ranges for cable networks. The more advanced FACTS and D-FACTS devices to influence power flow and voltages are discussed in the following chapter.

Chapter 3

FACTS in distribution systems

In this chapter the state of the art of power electronics applications in electrical power systems is given. When applied to transmission systems, this kind of devices is often referred to as FACTS devices (Flexible AC Transmission Systems), while for distribution system applications the terms D-FACTS (Distribution FACTS) and Custom Power are used. In this thesis, the term FACTS is used both for transmission and distribution system applications and D-FACTS for applications that are only for distribution systems. FACTS are used to influence the electrical voltages and currents in the grid to add flexibility to the operation of the power system. In a meshed power system power flows in the ’path of least resistance’, or more precisely, power flows according to the ratio of the inverse of the path impedances. This may lead to overloading of the path with the lower impedance, while the other paths are under-utilized. Voltages in a power system depend on the power flow conditions and impedances. If these voltages exceed certain limits, power system stability is at risk or end-users are no longer supplied with a stable voltage. Having the ability to control power flow and influence voltages can help prevent these problems. In this chapter, first the principles of power flow control are discussed, using simplified power systems representation. Second, the basics of FACTS technologies are discussed, followed by an overview of the state of the art of FACTS devices applied in distribution systems and their typical applications.

3.1 Principles of power ﬂow control

To investigate the possibilities to influence the active and reactive power flow in a certain line of a meshed network we first analyze the basic configuration of an overhead line connecting two generators, which can freely exchange active and reactive power. Next, the analysis is repeated for an underground cable.

29 30 Chapter 3

3.1.1 Power ﬂow in overhead line The power ﬂow in an overhead line (OHL) depends on the voltages on both sides and on the impedance of this line. Figure 3.1a shows the basic

P1, Q1 P2, Q2 I X G1 G2

V1 V2

(a) Single-line diagram.

jX S1 = P1 + jQ1

V1 I δV V2

(b) Phasor diagram. Figure 3.1: Single-line and phasor diagram for overhead line.

conﬁguration used to analyze this dependency [84]. For this analysis the overhead line impedance is assumed to be purely inductive, which is a good approximation. The general equation for complex power S, using the complex quantities V for voltage and I for current, is ∗ S = V I = P + jQ (3.1) while the current through the line equals

V1 V2 I = − (3.2) jX

Figure 3.1b shows the phasor diagram for arbitrarily chosen phasors V1 and V2. The angle between the dotted lines is equal to the angle of the impedance jX, i.e. 90 ◦, as follows from (3.2). It can be more intuitive to express complex voltages and currents in polar form using amplitudes and angles. The above mentioned voltages and current then become

V1 = V1 (cos(0) + j sin(0)) = V1 = V1 (3.3)

V2 = V2 (cos(δ) + j sin(δ)) = V2 (cos(δ) + j sin(δ)) (3.4)

FACTS in distribution systems 31

V sin(δ) j (V V cos(δ)) I = − 2 − 1 − 2 (3.5) X where δ is the phase angle between the voltages on both sides. Note that the given amplitudes of current and voltage are r.m.s. values. This results, in this lossless situation, in the active and reactive power injected by generators G1 and G2 to be equal to

V (V sin(δ)) P = P = − 1 2 · (3.6) 1 − 2 X

V (V V cos(δ)) Q = 1 1 − 2 · (3.7) 1 X

V (V V cos(δ)) Q = − 2 2 − 1 · (3.8) 2 X The given formulas can be expressed in the following more intuitive relation between impedance, voltage and power: In a line with a purely inductive impedance the flow of active power causes a phase difference between the sending and receiving end of the line while a reactive power flow causes a voltage amplitude difference. Indeed, overhead lines have a dominantly inductive impedance: the ratio of R/X is typically larger than 10. These relations result in the following basic options to change the active and reactive power flow through an overhead line:

1. By inserting a series voltage source in the line it is possible to:

a) Increase or decrease the active power transfer by changing the phase angle δ between the voltages on both sides of the line. b) Increase or decrease the reactive power ﬂow by changing the amplitude diﬀerence between the voltages on both sides of the line.

2. By inserting a series reactor or capacitor in the line it is possible to increase or decrease the impedance X of the line, and thus decrease or increase both the reactive and the active power ﬂow.

3. By connecting a parallel device active and reactive power can be injected or consumed which inﬂuences the power ﬂow accordingly. 32 Chapter 3

Series voltage source Methods 1a and 1b are illustrated using the network shown in Figure 3.2a where a series voltage source ∆V is connected in series with generator G2 (V3 = V1). Figure 3.2b shows the phasor diagram for method 1a, applying a series voltage in quadrature with the receiving end voltage V2 (i.e. δ = 90°), which results in active power ﬂow. Figure 3.2c shows the phasor diagram for method 1b, by using a series voltage source which is in phase with V2 (i.e. δ = 0°), which results in a reactive power ﬂow.

P1, Q1 P2, Q2 I X G1 G2

V1 V2 ∆V V3

(a) Series voltage source. S

jX jX S,I V1 V2 V1 ∆V ∆V V2

I (b) Quadrature voltage. (c) In phase voltage. Figure 3.2: Power ﬂow control using a series voltage source.

Series impedance Method 2 for controlling power ﬂow by changing the impedance of a line is illustrated using the network as shown in Figure 3.3a. In Figures 3.3c the resulting vector diagram is shown for the situation without and with a series capacitor, for an arbitrarily chosen voltage across the line. From the diagram it can be concluded that a series capacitor decreases the line impedance and increases the power ﬂow.

Parallel injection Besides series voltage sources and impedances, parallel devices can be used to control the power ﬂow in a overhead line network. A parallel device can inject/consume reactive power, if FACTS in distribution systems 33

P1, Q1 P2, Q2 I X G1 G2

V1 ∆Z V2

(a) Series impedance.

jX S S 1 jX + jωC V1 V1 I ∆V ∆V V2 I V2

(b) No series capacitor. (c) Series capacitor. Figure 3.3: Power ﬂow control using a series impedance.

capacitive or inductive components are used, or also inject/consume active power if, for example, a storage device is used. In Figure 3.4a the diagram is given for a device connected in parallel with an overhead line network. The distribution of the injected power P3 and Q3 towards the left and the right circuits depends on the ratio of their impedances. The eﬀect on the voltage V3 depends on the equivalent (mostly inductive) network impedance as seen from the connection point of the parallel device, as shown in Figure 3.4b. The equivalent network impedance is equal to the parallel connection of the short-circuit impedances of the left and the right part of the network. The resulting phasor diagrams for active and reactive power injection are shown in Figures 3.4c and 3.4d. From these diagrams it is concluded that in an overhead line network reactive power injection inﬂuences the voltage amplitude at the point of connection while active power injection changes the phase angle of the voltage. 34 Chapter 3

X1 P1,Q1 P2,Q2 X2 G1 G2

V1 P3,Q3 V3 V2

Parallel Device

(a) Parallel injection.

Sk = ∞ Xeq I

Veq P ,Q V3

Parallel Device

(b) Equivalent network. I

jXeq jXeq S,I V1 V1 V2 ∆V ∆V V2

S (c) Active power injection. (d) Reactive power injection. Figure 3.4: Power ﬂow control using a parallel device. FACTS in distribution systems 35

3.1.2 Power flow in underground cable An underground cable has different impedance characteristics than an overhead line: the series impedance of cables ranges from dominantly resistive for thin low voltage cables (X/R < 0.1) to mixed resistive/inductive (X/R 1) for thick medium voltage cables. To assess how power flow is influenced≈ by this, the analysis from above is repeated for a purely resistive line, as shown in Figure 3.5.

P1, Q1 P2, Q2 I R G1 G2

V1 V2 ∆V V1

(a) Series voltage source.

I R V1 S,I R V2 V1 ∆V ∆V S V2

(b) Quadrature voltage. (c) In phase voltage. Figure 3.5: Power ﬂow in cable with series voltage source.

Looking at these diagrams and formulas and comparing them to the ones for a purely inductive impedance, we can see the exact opposite relation between impedance, voltage and power: In a line with a resistive impedance, reactive power flow causes a phase angle difference and active power flow a voltage amplitude difference. And when a line has a mixed resitive/inductive impedance, the relation between impedance, voltage and power is a combination of the two previously analyzed cases: In a line with a mixed resistive/inductive impedance reactive power flow causes both a phase angle difference and an amplitude difference, and also active power flow causes these differences. More precisely, a voltage amplitude difference across a line causes a power flow with a P/Q ratio equal to the R/X ratio of the line, while 36 Chapter 3 a voltage angle difference causes a P/Q ratio equal to the X/R ratio of the line. Following these relations, shunt reactive power injection in a resistive network influences the phase angle of the voltage, while active power injection influences the voltage amplitude. As described in the previous chapter, the typical distribution system lines in the Netherlands, consist of underground medium and low voltage cables, with mixed resistive/inductive impedances. This implies that the use of reactive power injection for voltage control or the use of series voltage sources that introduce a phase shift for power flow control are not efficient methods. Instead, to control power flow, (also) the voltage amplitude needs to be changed and active power must be injected to influence the voltage. This is an important aspect to consider when analyzing the suitability of existing FACTS technologies for cable distribution networks.

3.2 FACTS technologies

FACTS devices are designed to use the above mentioned principles of power ﬂow control. Before presenting the diﬀerent topologies and applications of existing FACTS devices here an overview is given of the basic FACTS technologies. First, a summary of the available solid- state switching techniques is given, followed by a description of converter topologies, and concluded by addressing the use of mechanical switches.

3.2.1 Solid-state switching devices The enabling technology for FACTS devices is the application of solid- state switching devices. Devices available for high power applications are thyristors, gate turn-off thyristors (GTO), insulated gate bipolar transistors (IGBT) and integrated gate commutated thyristors (IGCT). Thyristors are naturally commutated devices, i.e. they can be turned on and start conducting at any moment, but only be turned off and stop conducting when the current crosses zero. The main distinguishing feature of GTOs, IGBTs and IGCTs when compared to thyristors is that they can interrupt the current at any moment, i.e. they have turn-off capabilities. The typical power and switching frequency ratings are displayed in Figure 3.6. Numerous text books and reports such as [51, 85, 86] are available for details on the technology and developments of solid-state switches. Due to their lower cost per MVA, the most widely used solid-state switch for the highest power levels is the thyristor. For lower power levels, as typical in distribution systems, the use of devices with turn-off capabilities starts to become more feasible. Research and development of solid-state switches is an ongoing process and the voltage and current ratings, as well as switching frequencies are FACTS in distribution systems 37

100M

10M

PCT IGCT 1M HV IGBT

Power Rating (VA) 100k

IGBT 10k

1k SJ MOS

100 MOSFET

10 10 100 1k 10k 100k 1M Frequency (Hz)

Figure 3.6: Ratings of solid-state switching devices [87].

continuously increasing. New solid-state devices and topologies are being developed, such as high power metal oxide semiconductors (MOS) controlled thyristors (MCT) and IGCTs [88]. New wide band-gap semiconductor materials are becoming available, which can replace silicon as the basic raw material for diodes, power MOSFETS (MOS field-effect transistors), thyristors, GTOs (gate turn-off thyristors) etc. [89]. Silicon carbide (SiC) is the most promising technology of this new generation of materials, which is already making it possible to make 10 kV MOSFETS and 13 kV IGBTs [49].

3.2.2 Converter topologies and switching strategies In FACTS applications solid-state switches are used to dynamically control the behaviour of a device by conducting during a controlled percentage of the time, thus deﬁning the eﬀective output or operating characteristic of the device. As a building block for FACTS two basic converter topologies exist:

• The three-phase voltage source converter.

• The three-phase current source converter.

The diagrams for the two basic types of converters are shown in Fig- ure 3.7. As can be seen, the two topologies are identiﬁed by the type of 38 Chapter 3

− (a) Thyristor based current source converter.

− (b) IGBT based voltage source converter. Figure 3.7: Basic diagrams for three-phase current and voltage source converters.

energy storage on the DC side of the converter: a capacitor for the voltage source converter and an inductor for the current source converter. Depending on the solid-state switch, different switching strategies are used. The output of a thyristor based current source converter is controlled by changing the time between voltage zero-crossing and turning on the thyristor, in other words by adjusting the firing angle. In Fig- ure 3.8, the case is shown of square wave switching, where each valve is conducting 50% of the time. Two types of voltage source converters are shown, the 6-pulse version, which has large harmonic and requires big filters, and the 12-pulse version, with reduced harmonic content. De- vices with turn-off capabilities allow different switching strategies such as pulse width modulation (PWM) and harmonic elimination technique. In the PWM technique, the switching moments are determined by comparing a reference signal with a saw tooth signal of the desired switching frequency and creating a logic control signal for the switches, as shown in Figure 3.9. The output of the converter is a square wave voltage, with a harmonic content that depends on the applied saw tooth frequency: higher switching frequencies reduce the amplitude and increase the frequency of the harmonics, allowing smaller, thus cheaper, filters. Practical high power applications use more advanced topologies, such as FACTS in distribution systems 39

1:1

VSC

(a) 6-pulse converter.

1:1

VSC

30° √3:1 (b) 12-pulse converter. Figure 3.8: Topology and output voltage source converters.

(a) Saw tooth signal and reference voltage.

(b) Switching pattern. Figure 3.9: PWM switching technique applied to a single-switch topology.

multi-level converters [90], and various modulation principles exist [50].

3.2.3 Mechanical switches Although not as the enabling technology of FACTS, also mechanical switches play an important role in FACTS applications. Mechanical switches are applied in situations where the high control speed of power electronics is not required or where they are complementary to power electronics, deﬁning the bigger control steps of a FACTS device, leaving 40 Chapter 3 the ﬁner and faster control to power electronics. Mechanical switches in FACTS are used to connect or disconnect a shunt component [91] or bypass series components [92], but also to switch between taps of a transformer, when applied in a on-load tap changer (OLTC) [93], see Figure 3.10.

(a) Series and shunt. (b) Tap changer. Figure 3.10: Mechanical switches applied in FACTS.

3.3 FACTS and D-FACTS applications

Using the technologies and building blocks as described above, diﬀerent FACTS topologies have been developed to perform diﬀerent functions in electric power systems. The basic functions of these topologies follow from the analysis in paragraph 3.1:

1. Inject or consume active and/or reactive power to control the voltage at the point of connection. Devices with these capabilities are connected in shunt with the power system, see Figure 3.11a.

2. Insert a series voltage source or impedance in a line in order to control the power ﬂow in a meshed system. This functionality requires a series connected FACTS device, see Figure 3.11b.

3. The two previous capabilities can be combined in one FACTS device, which then requires both a shunt and a series connection, which is called a mixed form device in this thesis, see Figure 3.11c. All previous capabilities can also be applied to more than one line by one device. The terms shunt and series can then often no longer be used to describe the topology. Also for those forms the term mixed form device will be used in this thesis. FACTS in distribution systems 41

(a) (b) (c) Figure 3.11: Basic (D-)FACTS connection methods: (a) shunt, (b) series and (c) mixed form.

Depending on the technology, (D-)FACTS devices can respond quickly to events in the power system. While in transmission systems FACTS are also used to increase voltage stability or to damp power oscillations, the main applications for distribution systems are to control power flow (including voltage control), to limit short-circuit current and to mitigate voltage unbalance. Other distribution systems applications are the mitigation of harmonic voltages, damping of harmonic resonance, and the mitigation of voltage dips and flicker. An international power electronics applications survey [94] shows that series devices are mostly applied for limiting the short-circuit current, protection against voltage sags, and shunt devices mostly for voltage regulation, phase balancing and flicker reduction. Little experience is reported in using series or mixed form devices for voltage profile and power flow control in distribution systems. In the following an overview is given of existing FACTS topologies and their applications. Extensive literature exists detailing the application, topology, and analysis of FACTS, such as [51, 95, 96]. Here only those topologies are discussed that are relevant to distribution systems. This excludes typical transmission system topologies such as high voltage DC systems (HVDC) and phase shifting transformers or quadrature boosters (PST/QB).

3.3.1 Shunt FACTS and D-FACTS devices Two mature shunt devices are relevant topologies for distribution systems, both used for reactive power injection or consumption. The ﬁrst device that is discussed is the Static Var Compensator (SVC), a device based on reactors and capacitors in combination with thyristors, widely deployed in transmission systems and on higher voltage levels of distribution systems. The second device that is described, is the more recently introduced Static Synchronous Compensator (STATCOM), a converter- based device using solid-state switches with turn-oﬀ capabilities. 42 Chapter 3

3.3.1.1 Static var compensator (SVC) A static var compensator (SVC) is defined by IEEE [97] as ”a device for fast reactive compensation, either inductive or capacitive, brought about by thyristor-based control of an effective shunt-susceptance. It is typically used to regulate voltage at a bus on the high voltage transmission system”. Or, as defined by CIGRE [98]: ”SVCs are shunt connected static generators and/or absorbers of reactive power whose outputs are varied so as to maintain or control specific parameters of the electrical power system”. An SVC consists of a thyristor controlled shunt reactor and/or a thyristor switched capacitor, optionally in combination with mechanically switched capacitors and/or reactors and filters, as illustrated in Figure 3.12a. By adjusting the firing angle of the thris- tor bridge, the equivalent value of the shunt reactor can be controlled. In situations where it is not required to have full control of the entire range of installed reactive power at any moment, mechanically switched susceptances are applied in parallel with the thyristor-controlled susceptance. This way the current rating of the thyristors is limited and so are the associated losses, both reducing costs. The amount of reactive power

TCR TSC ﬁlter MSC MSR TSR Capacitive Current Inductive Current (a) Single-line diagram. (b) Operating characteristic. Figure 3.12: Single-line diagram and operating characteristic SVC.

injection is determined by a controller that has a voltage reference value ∗ VSV C as input and as output the reference value for the injected reactive power. The relation between system voltage and reactive current is a droop function. When the system voltage equals the reference voltage, no reactive current is injected. When the measured system voltage is higher than the set-point, reactive current is consumed, the SVC be- FACTS in distribution systems 43 haves as a shunt reactor. When the system voltage is lower than the set-point the SVC behaves as a shunt capacitor and reactive power is injected. Note that the maximum reactive current decreases linearly with decreasing voltage, and thus reactive power decreases quadratically with decreasing terminal voltage. That is to say,

V V 2 I = Q = (3.9) L jωL L jωL I = jωC V Q = jωC V 2 (3.10) C · C ·

The actual operating point is determined by the reference voltage, the droop setting, the power system voltage when there is no reactive power injection, and the system impedance. The slope of the droop function is a design parameter of the SVC and determines the operating area of the device. Figure 3.12b illustrates the operating characteristics of an SVC. In this diagram, the injected reactive current is on the horizontal axis, while the system voltage is on the vertical axis. The positive x-axis is for inductive current, the negative x-axis is for capacitive current. For reference, the operating characteristics of a fixed reactor and a fixed capacitor are also depicted, as hashed straight lines through the origin. The nominal operating characteristic is indicated by the solid line, characterized by zero current injection at nominal reference voltage (1.0 p.u.), the nominal droop and the capacitor and reactor sizes. The enclosed area indicates the possible operating region of an SVC, for different voltage reference values and droop settings. The operating region is bounded by the capacitor and reactor size, the solid-state switches voltage rating and by the minimum and maximum droop settings, in combination with the minimum and maximum reference voltage. The solid lines outside the filled area indicate the characteristics during short over- and under-voltages: the solid-state switches do not switch anymore, but are continuously conducting, making the SVC impedance equal to the full capacitor or reactor size. The reactive power injection capability of the SVC is mainly applied to control voltage and damp oscillations, but also to improve the steady state power flow and transient stability [99]. Volt- age control can be used to compensate under-voltage and unbalance but also faster phenomena such as dips and flicker. Note that this technology is mostly applicable in overhead line networks, since the injection of reactive power only leads to a voltage increase in lines with an inductive impedance. The device is therefore classified mainly as FACTS device although it can also be applied as a D-FACTS device in overhead line distribution systems. 44 Chapter 3

3.3.1.2 Static synchronous compensator (STATCOM) This device was originally called static condensor (STATCON), while later the term static synchronous compensator (STATCOM) was ac- cepted. ABB also uses the proprietary name SVC Light® [100]. A STATCOM is a solid-state DC to AC switching power converter, consisting of a three-phase voltage source converter. On the AC side, the converter is shunt connected to the grid, normally through a transformer. The DC side of the converter is connected to a capacitor. The function of the capacitor is to provide a constant DC voltage, not to supply reactive power, which is why its size is only small when compared to that of a classic static shunt capacitor or SVC. Figure 3.13a shows the single- line schematic of a STATCOM. The IGBT converter can in principle inject any current into the grid, with any desired amplitude and phase angle. However, a basic STATCOM conﬁguration includes only a small DC capacitor, which prohibits any exchange of active power, since that would quickly charge or discharge the capacitor, leading to an under- or over-voltage on the DC bus. As a result, the injected current generally has a phase angle of either 90° leading or 90° lagging the grid voltage, thus injecting or consuming reactive power as desired. To also allow active power exchange, storage or generation must be connected to the DC bus. The operating characteristic of a STATCOM (without stor-

STORAGE/ GENERATION

Capacitive Current Inductive Current (a) Single-line dia- (b) Operating characteristic. gram. Figure 3.13: Single-line diagram and operating characteristic STAT- COM with optional storage and generation. age) is shown in Figure 3.13b. The injected reactive current is on the x-axis and the terminal voltage on the y-axes, identical to the axes of FACTS in distribution systems 45

Figure 3.12b, which shows the operating characteristics of an SVC. The nominal operating characteristic is indicated by the solid line, characterized by zero current injection at nominal reference voltage (1.0 p.u.), nominal droop and the maximum converter current rating. The enclosed area indicates the operating region of a STATCOM, which is bounded by the current and voltage ratings of the power electronics valves and by the minimum and maximum droop settings, in combination with the minimum and maximum reference voltage. Unlike it is the case with an SVC, with a STATCOM the maximum reactive current does not linearly decrease with the terminal voltage, but is constant, which results in a linear decrease of reactive power with decreasing terminal voltage, instead of a quadratic decrease. The reactive power injection capability of the STATCOM is used for similar applications as the SVC, and additionally for harmonics mitigation. When applied to distribution systems, the STATCOM is also called D-STATCOM and besides the normal applications, also used to increase the loadability of distribution feeders [101] by improving the power factor. Due to the fast response of the STATCOM controller, its contribution to the short-circuit current during a fault in the power system, is limited to around 1 p.u. The most important distinguishing features of a STATCOM when compared to an SVC device are its smaller footprint and higher reactive power output at reduced system voltage. Technological developments continuously increase the voltage and current ratings of IGBT valves, as well as the allowable switching frequency. For lower voltage levels this allows the omission of the grid coupling transformer and a direct grid connection through series reactors, which reduces the costs of the STATCOM installation. Higher switching frequencies allow smaller filters, also reducing costs and footprint. Relocatable units exist. Also here, as with the SVC, it must be noted that the application of a STATCOM is most efficient in overhead line networks and less efficient in cable networks. Like the SVC, this device is more used as a FACTS than as a D-FACTS device.

3.3.2 Series FACTS and D-FACTS devices Series FACTS devices are based on the principle of inserting a series voltage source or changing the impedance of a circuit in order to change the power ﬂow in the meshed grid that the line is part of, as explained in paragraph 3.1. The basic impedance changing types are the Thryristor Switched Series Capacitor (TSSC) and the Thryristor Controlled Series Capacitor (TCSC). An implementation of a series voltage source is the so called Distributed Static Synchronous Series Compensator (D-SSSC), consisting of a series transformer in combination with a voltage source converter. The Phase Shifting Transformer (PST) is another FACTS device that operates as a series voltage source, but it is more typical 46 Chapter 3 for transmission systems than for distribution systems and therefore not considered here.

3.3.2.1 Thyristor switched series capacitor (TSSC) and thyristor controlled series capacitor (TCSC) The TSSC is a FACTS device based on a number of series connected capacitors, each with a parallel thyristor branch, which gives the option of short-circuiting the capacitor. By using a sufficient amount of capacitors in series, the impedance can be controlled in small discrete steps. The thryistors are not angle-controlled, but are either conducting or blocking the entire sine wave. The more advanced TCSC consists of a few series connected capacitors, each with a parallel reactor that is phase angle controlled by thyristors. This configuration prevents discrete impedance control steps and also allows inductive behaviour instead of only capacitive. Both devices and their operating ranges are shown in Figure 3.14. When applied in overhead line networks, the TSSC and TCSC can lower or increase a circuit’s inductive impedance, thus controlling power flow. In cable networks, the TSSC and TCSC are less efficient in doing this, since the resistive component of the network impedance, which is significant if not dominant, is not affected.

3.3.2.2 Static series synchronous compensator (SSSC) The SSSC is a voltage source converter connected to the grid through a series transformer, as shown in Figure 3.15a. Its function is to inject reactive power into the grid, which in a meshed network with its inductive impedances results in circulating active power. No storage or generation is connected to the DC bus which prevents substantial active power exchange (some active power is consumed to compensate losses and maintain the DC bus voltage). A D-FACTS variation to this topology, the Distributed Static Series Synchronous Compensator (D- SSSC) [102], has a direct connection of the converter to the grid without a transformer, as shown in Figure 3.15b and is applied on the lower voltage levels. The (D-)SSSC can only inject reactive power and is therefore more suitable for overhead line networks than for cable networks.

3.3.3 Mixed form FACTS and D-FACTS devices 3.3.3.1 Unified power flow controller (UPFC) and interline power flow controller (IPFC) The UPFC can be seen as a combination of a STATCOM and an SSSC connected by their DC bus, as shown in Figure 3.16a. Due to its ability to allow active power flow from the shunt to the series connection, FACTS in distribution systems 47

(a) TSSC topology.

(b) TCSC topology.

V V

both capacitive capacitive capacitive 1 section inductive, 1 section capacitive 0 n sections 0 1 section both inductive inductive inductive

I I (c) TSSC characteristic. (d) TCSC characteristic. Figure 3.14: Single-line diagrams and operating characteristics TSSC and TCSC.

(a) SSSC. (b) D-SSSC. Figure 3.15: Single-line diagrams SSSC and D-SSSC. 48 Chapter 3 and independent of that reactive power injection in both interfaces, the UPFC has a very wide range of capabilities and possible applications. The UPFC can support voltage by reactive power injection, but also change the series impedance of a circuit and insert a series voltage in phase in or in quadrature with the line voltage, thus controlling the power ﬂow. Despite its diverse capabilities, UPFCs are not frequently applied in power systems, due to unfavorable economics. The IPFC

(a) UPFC. (b) IPFC. Figure 3.16: Single-line diagrams UPFC and IPFC.

can be seen as a combination of two or more SSSCs connected by their DC link, see Figure 3.16b. The application allows bi-directional active power flow control between the different circuits. Each SSSC can independently inject a quadrature voltage, which in overhead line systems results in an active power flow. In cable systems, a combination of in- phase and quadrature voltage must be used, also requiring active power injection. Both the UPFC and the IPFC can be effectively used both in overhead line and cable systems due to their ability to inject both in-phase and quadrature series voltages, and can therefore be classified both as FACTS and as D-FACTS devices.

3.3.3.2 Solid-state load tap changer (SSLTC) Besides in converters, power electronics are also being applied in the control of transformer tap changers. The resulting D-FACTS application is known under the name of solid-state load tap changer (SSLTC). The power electronics can simply replace mechanical switches and apply full-wave switching, maintaining the discrete character of the tap changer, or it can apply a PWM switching pattern, which transforms the discrete tap changer steps in a continuous characteristic. Imtech, a Dutch supplier, provides a transformer equipped with a power electronics controlled tap changer under the name of Smarttrafo®. The application consists of a 10/0.4 kV transformer with IGBTs connected to the high voltage tap changer [103], which allows it to maintain a con- FACTS in distribution systems 49 stant voltage on the secondary side. This is in large contrast with the normal situation where it decreases linearly with increasing load, since typically the tap changers of these distribution transformers are oﬀ-load tap changers, only being adjusted when the distribution system or the demand changes structurally. A similar idea was proposed in [104], where a thyristor switched tap changer is applied for the larger, discrete voltage steps and an additional PWM controlled series controller using GTOs or IGBTs ensures an overall continuous voltage control.

3.3.3.3 Transfer switches A transfer switch is a device which is applied to select the supply path for the connected loads, see Figure 3.17. Typically, one of the supply

Supply Path 1 Supply Path 2

Switches

Control

Loads

Figure 3.17: Transfer switch. paths is the public electricity network, while the other is formed by a local generator. Mechanical switches cannot be operated fast enough to perform this transition without a (short) power interruption, so instead, power electronic valves are used. The resulting device is the static transfer switch (STS), which has the capability to transfer loads within half a power frequency cycle from one source to the other, which is suﬃ- ciently fast for uninterrupted power supply [105]. This topology has been applied successfully on medium voltage levels to achieve uninterrupted power supply for the connected loads [94]. The topology of the STS has been extended with mechanical switches to take advantage of both the speed of power electronics and the low losses of mechanical switches, while still achieving uninterrupted supply, resulting in the hybrid transfer switch (HTS) [106]. The STS and HTS are, besides for uninterrupted supply, also applied to increase power quality, by switching from one source to the other if one source has a higher power quality level than the other. 50 Chapter 3

The application of the STS and HTS is targeted to the increase of the reliability and power quality of power supply to critical loads, and not to the optimization of the power ﬂow in a power system.

3.3.3.4 Back-to-back converter The back-to-back converter is the topology most related to the topic of this thesis, and is applied both as FACTS and as D-FACTS device. The device consists of two voltage source converters (i.e. STATCOMs) that are connected on the DC side. Besides the STATCOM capabilites (fast reactive power injection) the device has the additional capability of active power exchange between the two AC connections, thanks to its DC bus interconnection. The single-line diagram of a back-to-back converter is shown in Figure 3.18a. The solid-state switching devices that are used

Q Vac = 0.9 p.u. Vac = 1.0 p.u. Vac = 1.1 p.u.

Imax

Maximum DC link power (a) Single-line diagram. (b) Operating characteristic. Figure 3.18: Back-to-back device single-line diagram and operating characteristic. in the converters need to have turn-oﬀ capabilities, typically IGBTs are used. The operating range of the device is, as described in [107], determined by: • The current rating of the converter, which can be seen as a circle around the origin in the P Q plane. • The maximum DC voltage level, which limits the output voltage of the converter, which in its turn limits the reactive power injection capabilities in overhead line networks. In cable networks, this also limits the active power transfer capability. FACTS in distribution systems 51

• The maximum DC connection current, which limits the active power transfer capability.

Depending on the presence of external passive reactive components, as part of the grid-side filter or otherwise, the entire operating area, which is shown in Figure 3.18b, can be shifted along the Q-axis. The shown limitation in reactive power in case of higher AC system voltage is true for inductive networks. For a resistive network, the limitation would be on the positive P -side of the graph. As illustrated, this FACTS device can both control active and reactive power, making it suitable to be applied in both overhead line and cable networks. Different suppliers use different names for their back-to-back converter products. ABB calls their application BtB Light® [92] when the two converters are on the same location and HVDC Light® [108] when the two converters are geographically separated and connected by a DC cable. SIEMENS uses the name SIPLINK® [109] for their medium voltage back-to-back application, used for connecting different grid areas, and the name Siharbor® for a similar installation used for ship-to-shore connections. In Japan the Central Research Institute of Electric Power Industry performs research on an application which is called a Loop Balance Controller or a Loop Power Flow Controller. In the following, some existing installations are discussed to illustrate the state of the art of this technology.

Eagle Pass/Piedas Negras back-to-back interconnection. This HVDC Light® installation, commissioned in 2000, interconnects the asynchronous grids of Mexico and Texas, USA, with the purpose of providing voltage support for both grids using reactive power injection, and allowing controlled power exchange from the stronger Mexican grid to the weaker Texas grid to prevent voltage instability in the Eagle Pass network [92]. The 36 MVA 138 kV installation with a footprint of 35 m x 45 m also facilitates supplying Eagle Pass when the connection with the rest of the Texas grid is out of service [110], although the reconnection to the rest of the Texas grid requires a black-out for the Eagle Pass grid.

Talega dual STATCOM. This back-to-back installation, commissioned in 2003, consists of two STATCOM installations, capable of providing voltage support using reactive power injection. The ratings of the installation are 100 MVA, 138 kV. The DC link required for active power transfer is physically± in place, but the actual use of this functionality is not possible, since it is not yet implemented in the control system. The control implementation would result in a maximum active power transfer capability of 50 MW [111]. 52 Chapter 3

Ulm and Karlsruhe back-to-back installations. The two SIP- LINK® applications in Ulm and Karlsruhe, Germany, both consist of a back-to-back installation connecting two grids. In Ulm, the main functionality of the installation is to transport energy between the networks of the districts Ulm and Neu-Ulm during times of peak demand to share expensive regulating power. The installation has a capacity of 2 MVA and was commissioned in 2003 [112]. The pilot installation in Karlruhe has similar ratings, and was installed to avoid overloading of a certain cable [113]. Additional functions are the optimization of voltage and the supply of an aﬀected sub-network as an insular system in the event of failure of one of its infeed points.

Japanese distribution system back-to-back research setup. In Japan, the Central Research Institute of Electric Power Industry has developed a medium voltage back-to-back application [114]. The basic functions are the control of voltage and power ﬂow in a distribution system without increasing the short-circuit power level. The installation consists of a transformerless 1 MVA 6.6 kV back-to-back converter with a footprint of 4 m x 1.7 m an 3 m height. The transition from meshed to radial operation is supported, with relatively large voltage changes. The opposite process is not reported [115].

Back-to-back installation for research, development and testing at KEMA. Although not installed as a FACTS device in a distribution system, this installation is worth to be mentioned, as an illustration of the state of the art in power electronic converters. In the Flexible Power Grid Lab (FPGL) in Arnhem, the Netherlands, a back-to-back installation was erected for research and development and testing purposes. The setup consists of a 4-quadrant 1 MVA, 3.3 kV converter, with custom-made controls that allow a freely adjustable voltage output, including adjustable individual harmonic levels, unbalance, frequency variations, dips etc. The footprint of the back-to-back installation is around 5 m x 2 m.

3.4 Conclusion

A FACTS or D-FACTS device is used to influence voltage and loading levels in the electrical network that it is part of. The effect that such a device can achieve is significantly different for cable networks when compared to overhead line networks. In an overhead line network reactive power injection leads to an increased voltage, while in a cable network it (also) causes a phase angle difference. Similarly, a series voltage source that inserts a quadrature voltage causes a circulating active power flow FACTS in distribution systems 53 in a meshed overhead line network, while in a meshed cable network, it (also) causes a circulating reactive power flow. The application of solid-state switches is the enabling technology for FACTS and D-FACTS. An overview is given of existing devices and their capabilities, followed by an overview of the basic topologies in which these devices are applied. FACTS and D-FACTS devices exist in a wide variety of topologies and functions, and may be categorized by their connection type. The shunt devices called SVC and STATCOM both have a relatively simple topology and fulfill one specific function, namely the injection of reactive power for voltage control. The series devices TSSC and TCSC are both based on controlling the series impedance of the line, while the SSSC and D-SSSC operate as series voltage sources, controlling the power flow by injecting a series voltage. None of the devices SVC, STATCOM, TSSC, TCSC, SSSC and D-SSSC can inject active power and are therefore less suitable for application in cable distribution networks where only reactive power injection has a limited effect for voltage and power flow. The mixed form devices UPFC and IPFC are more complex devices, combining several functions in one device, such as series voltage injection, series impedance control, voltage control, active power exchange between circuits and are more suitable for cable distribution networks. Power electronics are also applied in distribution system transformer tap changers, switching from one tap to the other, thus avoiding mov- ing parts and changing the discrete characteristic of a mechanical tap changer into a continuous one. These devices are only applied for voltage control, not for power flow control in meshed networks. Finally the back-to-back converter is discussed, which allows both active and reactive power injection, which makes it suitable for application in cable distribution systems. This configuration is the most relevant D-FACTS topology for this thesis. For illustration purposes, several installed devices are described for this device. In the next chapter, an extension of the back-to-back converter application is proposed.

Chapter 4

Functional concept of the Intelligent Node

The introduction of distributed generation in electrical power systems is changing the operation of that part of the network which tradition- ally has been called the ’distribution system’. The primary task of the operator of these networks, the DNO, is to operate them, to ensure the safety and reliability of these networks and to facilitate the transmission of electricity in an economical way. The DNO also has to do network construction, repair, refurbishment and expansions while considering measures regarding renewable energy, energy saving, demand control and DG, as was discussed in Chapter 2. Given the increasing uncertainty of load and generation (size, location and coming on-line) in distribution systems and the DNO’s obligation to maintain enough spare capacity, investments in grid reinforcements may turn out to be uneconomic. This stimulates the DNO to maximize the use of the existing network and to consider installing ﬂex- ibility and intelligence in the grid, instead of traditional grid reinforcements. It is in this context that the application of the Intelligent Node (IN) is proposed. The IN may be seen as a black box with on the outside a number of AC ports, and, for now, an undeﬁned internal topology. The preliminary functional requirements of this black box are:

• Inject or consume an adjustable amount of active and/or reactive power through each of its AC ports.

• Supply a radial network part from any of its AC ports.

• Improve the power quality of the connected network parts.

• Optionally: store energy.

55 56 Chapter 4

In this chapter, several applications are proposed and it is analyzed how and to what degree they can provide the required ﬂexibility. As the IN applications are discussed, the functional requirements are elaborated in more detail. The ﬁrst application of the IN to be discussed concerns the increase of loading capacity, followed by the IN application to facilitate DG integration. The third application which is presented is the use of the IN to mitigate voltage dips. The chapter ends with a summary of the functional requirements and possible internal topologies, followed by a conclusion.

4.1 Facilitating increased loading

One of the oldest trends in electrical power systems is the gradual growth of electricity demand. As an alternative to traditional grid reinforcements to accommodate this growth, the application of an IN is proposed. Power system lines are generally not loaded to their nominal rating because of redundancy considerations. With this in mind, two grid configurations are identified. The first concerns a radially operated pocket network in which redundancy can be made available by reconfig- uring the network, which can be done in multiple ways through the IN. The second concerns a network where the redundancy is not enough to accommodate a load growth and where an alterative supply point in a certain geographical distance has enough capacity to accommodate the increasing load, but the power flow of the interconnection between the regions has to be controlled to prevent overloading.

4.1.1 Controlled sharing of redundancy This IN application concerns a network that requires reconfiguration in order to resume supply after a certain part of the network was de- energized due to a short-circuit. Such a network is shown in Figure 4.1a. In case of a fault on feeder d the concerned section of the line is de- energized. Subsequently, to limit or avoid customer interruption time, the network is reconfigured and the affected loads are supplied by feeder c. This reconfiguration is only possible if enough spare capacity is available on the concerned feeders, as explained in Chapter 2. To make the spare capacity fully available to accommodate the extra load, the application of an IN is proposed. The concept of sharing redundancy is based on the use of the total spare capacity of several feeders, instead of using only the spare capacity of one feeder. Figure 4.1a shows a part of a distribution grid, with four feeders ending in one geographical location. In that location a coupling busbar is used with 4 load-break switches (LBs) to connect and disconnect feeders. As illustrated, the pre-reconfiguration loading of any of the feeders cannot exceed 50% of Functional concept of the Intelligent Node 57

50% a

b 50%

c 50%+50%=100% Coupling busbar d Sum=200% (a) One feeder supplies the load of an other feeder.

75%+25%=100% a

b 75%+25%=100%

c 75%+25%=100%

d Intelligent Node Sum=300% (b) All feeders share the load of the aﬀected feeder. Figure 4.1: Sharing of redundancy.

their nominal rating, in order to allow the shown reconfiguration. In Figure 4.1b the switches and the coupling busbar are replaced by an IN which actively balances the power demand from feeder d equally across feeders a, b and c. As illustrated, the pre-reconfiguration loading level of each feeder can now be allowed to grow up to 75% of the nominal rating. Note that in case of n feeders connected in the described way, the maximum loading level becomes 100 (n 1)/n %. This also means that in case of only 2 feeders, there is no· benefit− from this application, at least in terms of redundancy, since no further redundancy sharing can be done. The control objective while the network is reconfigured, is to balance the power consumption of feeder d among feeders a, b and c. In this basic example it is, for simplicity reasons, assumed that only the network feeders are approaching their loading limits, and that the supplying transformers have enough capacity to accommodate the load. With all feeders fully in service, the IN can be used to optimize the power flow in the meshed network. Part of this optimization can be the control of voltage profiles, which is discussed in paragraph 4.2 and the minimization of losses. The minimization of losses is not further elaborated in this thesis. In order to establish the requirements for the IN for this application, the sequence of events during such a grid reconfiguration is analyzed. 58 Chapter 4

a a) The power system is in b IN normal operation, the IN controls power ﬂow, for ex- c ample, to minimize losses.

a b b) A permanent short- IN circuit occurs on feeder c c.

a c) The circuit breaker of b IN feeder c opens and the IN disconnects from feeder c. c The loads on feeder c are unsupplied.

a d) The fault is isolated b IN by opening the load-break switches of the aﬀected sec- c tion.

a e) The circuit breaker of b IN feeder c is reclosed and the IN starts to supply the IN c side of feeder c. All loads are supplied again.

a f) The line section is re- b IN paired and the IN ensures voltage synchronism c between the two energized parts of feeder c.

a g) By closing the two load- b IN break switches of feeder c meshed operation of situa- c tion a) is restored. Figure 4.2: Sequence of events during short-circuit in a network with a 3-port IN.

Figure 4.2a through g show and describe the diﬀerent events and actions following a short-circuit in one of the feeder sections. To compare with the sequence of events in situations without an IN, see Chapter 2. Functional concept of the Intelligent Node 59

The sequence of events for planned maintenance are the same, with the exception of steps b) until e), which are skipped, and the supply of the loads on feeder c is never interrupted. The operation of circuit breakers is in many networks performed simultaneously on all three phases. In other medium voltage networks, for example in the Netherlands, however, the phase-by-phase operation of load-break switches is common, given the wide-spread application of manually operated, compact epoxy resin insulated, single-phase switchgear [116], of which an example is shown in Figure 4.3.

Figure 4.3: Manual phase-by-phase opening and closing of medium voltage switchgear [117].

To enable the described events and phases, the IN must be able to:

• Detect a permanent short-circuit and de-energize the appropriate feeder (feeder c in the example).

• Ride through a voltage dip, which occurs due to a permanent fault on an adjacent feeder (feeders a and b should not be aﬀected in the example).

• Detect the opening of the load-break switches which isolate a network section and change from controlling power ﬂow in the meshed network to supplying the radial network part (in case of section de-energization for planned maintenance).

• Synchronize the voltages on both sides of the isolated feeder section.

• Detect the closing of the load-break switches which restore the meshed network operation and change from supplying the radial network part to controlling power ﬂow in the meshed network.

• Support the above also for the phase-by-phase operation of the load-break switches. 60 Chapter 4

4.1.2 Controlled power exchange between grid areas Figure 4.4 illustrates another situation where an IN can facilitate load growth, also by sharing redundancy, but in a diﬀerent manner. For this

T-B1 T-B2 1.0 p.u. 1.0 p.u.

B-B T-A1 Option 1, 0.2 p.u. 1.0 p.u. Ld-B L-BC L-AC1, 1.0 p.u. 0.7 p.u. L-AC2, 1.0 p.u.

L-AC3 B-C T-A2 PST Ld-C 1.0 p.u. 0.9 1.2 p.u. B-A → T-A3 Option 2, 0.2 p.u. Figure 4.4: Traditional grid reinforcements to supply increasing load on busbar B-C (Ld-C).

network, it is expected that, in the future, the load Ld-C will exceed the capacity of the redundant lines L-AC1 and L-AC2 and transformers T-A1 and T-A2. Busbar B-B however, still has enough spare capacity to feed the increasing load, but connecting busbars B-B and B-C, would result in an overloaded line L-BC. This overloading is the result of equalizing currents that occur due to different circuit impedances of the two grid areas, different voltage amplitudes and/or different phase angles. These factors depend on the power system configuration and the network loading conditions and therefore vary during time. Counteract- ing the equalizing currents by classical means, involves, for example, the application of a phase-shifting transformer PST in line L-BC, which is shown as Option 1. This option is only feasible in networks with a mainly inductive impedance, such as overhead line networks. The other traditional method to increase the network capacity is the shown Option 2, which consists of increasing the capacity of line L-AC and adding a third transformer T-A3. A side effect of both options is that the maximum short-circuit current on busbars B-A, B-B and B-C increases. As an alternative, the application of an IN in line L-BC is proposed, Functional concept of the Intelligent Node 61

T-B1 T-B2

B-B 0.2 p.u. T-A1 Ld-B L-BC L-AC1 IN

L-AC2 B-C B-A T-A2 Ld-C Figure 4.5: IN application to supply increasing load Ld-C.

as shown in Figure 4.5. The IN and line L-BC both have a power rating of the excess load, i.e. 0.2 p.u. in the shown numerical example. The IN actively controls the power flow and thus solves the expected power deficit, while preventing overloading. The IN can control the power flow independently from power system impedance and voltage differences, and is thus feasible in both cable and overhead line networks. Depend- ing on the IN technology, the short-circuit rating of busbars B-A, B-B and B-C can be maintained the same. Besides the already mentioned functional requirements, this application does not require any additional functionality from the IN. Whether the IN’s ability to supply a radial network area is useful depends on local load conditions, since the IN is only able to supply a load Ld-C up to the power ratings of the IN and the line L-BC. This situation could, for example, occur in a N-2 contingency with both lines L-AC1 and L-AC2 out of service. An existing back-to-back installation at Eagle-Pass, Texas, USA, which was described in paragraph 3.3.3.4, has the ability to supply a radial network. The ability to resume meshed operation without supply interruption is not reported.

4.2 Controlling voltage proﬁles

A second IN application involves the control of voltage profiles to facilitate the integration of distributed generation. The connection of DG units in a distribution network can result in reverse power flow and volt- 62 Chapter 4 age rise instead of voltage drop on the concerned feeder system. If this occurs in only some of the feeders that are fed from one busbar, the voltage profiles of the connected feeders can possibly no longer be kept within a certain band by compounding (the operation of the transformers tap changers, see also paragraph 2.4). This is illustrated in Figure 4.6. To

(a) One feeder has predominantly generation, the other mainly load.

I Loads

Generators distance → (b) Resulting feeder current, assuming homogenously distributed loads and DG. Generators

Vmax

Loads Vmin distance → (c) Compounding is no longer a solution.

Figure 4.6: Voltage proﬁles with increasing DG.

keep voltages within the required limits it is possible to create dedicated networks to connect generation, for example, by splitting the busbar in generation and load sections and supplying each from a different transformer, each with its own compounding. However, in case of an outage of one of the transformers the busbar cannot be operated split and the voltage profiles can still exceed the allowed limits. The application of an IN can be proposed as a method to control voltage profiles by the control of both active and reactive power flow. As explained in Chap- ter 3, the voltage amplitude can be influenced by the injection of active and/or reactive power, depending on whether the line concerns an underground cable or an overhead line. In Figure 4.7, the location of the IN for this application is shown, as well as an illustration of the effect which can be achieved on the voltage profiles. In this network, only active power is exchanged between the two feeders, resulting in the shown Functional concept of the Intelligent Node 63

IN P

(a) IN controls power ﬂow between feeders.

I Loads 0 Generators distance → (b) Resulting feeder current has partly changed direction.

V max Generators

Loads V min distance → (c) Voltage proﬁles are within limits.

Figure 4.7: Voltage proﬁle ’bending’ with the application of a 2-port IN.

current profiles. Power flow calculations show that the voltage profiles are then ’bent’ towards each other, which makes it possible to bring all voltages within the network within the required voltage band. Besides active power, also reactive power can be injected or consumed, resulting in a similar effect on the voltage profile. The optimal active and reactive power set-points of the IN depend on the loading situation in the network and on the ratings and impedances of the various components. The shown example concerns a two-port IN, but the concept can be expanded to an n-port device. In the following paragraph, a four-port example application is described.

4.2.1 Example application The intention of this example application is to assess the maximum achievable voltage profile improvement in a certain network by controlling the power flow with the IN. It is not the intention to find the optimal IN settings for this network, considering all effects. This means that, for example, the minimization of losses is not taken into account. To illustrate the possible benefits of this IN application, we consider the network 64 Chapter 4

6900 feeder length (m) 240Al cross section (mm2) and material 0.08 total load per feeder (MVA) normally open

150/10.5 kV 2x66 MVA 20 %

6900 1900 240Al 150Al 0.08 0.32 4900 400 4600 150Al 150Al 35Cu 0.32 0.04 0.64 1 MVA 2500 500 400 150Al 150Al 150Al 0.16 0.04 0.08 1 3 6600 900 1100 150Al 150Al 150Al P1,Q1 IN P3,Q3 0.08 0.04 0.04 1300 3400 2 4 P2,Q2 P4,Q4 240Al 25Cu 600 0.16 0.04 800 35Cu 1600 35Cu 0.08 35Cu 0.16 0.32 700 3200 3200 1200 95Al 95Al 35Cu 35Cu 0.00 0.08 0.16 0.16

Figure 4.8: Medium voltage cable network with a 4-port IN.

shown in Figure 4.8. The network consists of several cable feeders, each supplying some loads (cos(φ) = 0.9 inductive), equidistantly connected to their feeder (not shown). The total load per feeder is shown in the figure. One 1 MVA generator (cos(φ) = 1.0) is connected to a busbar in the grid. The rectangle shows the location of the IN, which does not have storage connected. Three configurations are considered. In configuration I, an IN is connected in the indicated location. In configuration II, the IN is replaced by a busbar, which interconnects the four feeders, resulting in meshed operation of the network. In configuration III, reactive power compensation is connected to the busbar of configuration II.

4.2.1.1 Voltage optimization method To obtain the optimal voltage profiles in the network, we must first define what is considered ’optimal’ and, second, find the IN settings to achieve these optimal conditions. In order to define the optimal voltage profiles two terms are introduced: ’voltage band’ and ’associated nodes’. The ’voltage band’ is defined as the difference between the lowest and the highest voltage occurring on a certain set of grid nodes at a certain moment. The ’associated nodes’ are defined as all nodes for which a Functional concept of the Intelligent Node 65 galvanic path exists to the IN and for which this galvanic path does not pass through the MV busbar of the HV/MV substation. The node voltages are calculated by solving the power flow equations, a technique which is elaborated in many text books, for example in [118]. In this example the power system analysis tool PowerFactory from DIgSILENT is used for this task. In the shown example, this excludes all nodes of the radial network section connected to the center of the MV busbar of the HV/MV substation. The voltages on the k associated nodes are Van,1 until Van,k. The optimal voltage profiles in the network occur when the voltage band of the associated nodes has its minimum value. The objective function is expressed mathematically as

minimize max(V , ..., V ) min(V , ..., V ) (4.1) an,1 an,k − an,1 an,k which is subject to the following boundary conditions: • The sum of the injected/consumed active power through all IN N ports is zero: i=1 Pi(t) = 0 (in the example no storage is connected) P • The apparent power of each IN port is not exceeded: Si

• The feeder ratings are not exceeded: Ii(x) < Imax for i = 1..N and 0

• The node voltages are limited: 0.97 p.u. < Van,i < 1.03 p.u. for i = 1..k As an example of a possible approach towards optimal settings for the IN, Cauchy’s gradient method [119] is used. The gradient method determines the effect of each control parameter on the objective function in order to find the steepest gradient towards the maximum or minimum of the objective function. Figure 4.9 shows a graphical impression of the application of the gradient method for maximization of a 2 dimensional problem. By applying small disturbances in the two dimensions (x and y), the gradient of the objective function (z) is determined. Iteratively applying a stimulus in the direction of and proportional to the gradient will bring us to the desired operating point: the global maximum. A flowchart of the individual steps to be performed to implement the gradient method for voltage profile optimization is shown in Figure 4.10 and described in more detail hereafter. In the example network, eight variables exist, which can be adjusted to find the optimal voltage band: the active and reactive power injected from each of the four IN ports. Thus, the optimum IN setting is com- posed of a linear combination of the eight variables Q1 to Q4 and P1 66 Chapter 4

Figure 4.9: Graphical impression of gradient method applied to 2 dimensional problem.

to P4, see Figure 4.8. In the example, we assume that the IN does not contain storage, which reduces the degree of freedom with one, since the sum of P1, P2, P3 and P4 must be equal to zero at all times. The first three of the seven vectors, S1 until S3, concern the active power flow control and are chosen such that this condition is met for all possible combinations of Si, see Equation (4.2). The remaining four vectors S4 until S7 concern the reactive power flow control and are equal to Q1 until Q4, which are independent variables. The resulting seven independent vectors S1 until S7 are equal to

P1 S1 1 −1/3 −1/3 −1/3 0 0 0 0 P2 S2 −1/3 1 −1/3 −1/3 0 0 0 0       P3 S3 −1/3 −1/3 1 −1/3 0 0 0 0  P4   S4  =  0 0 0 0 1 0 0 0 ×   (4.2)      Q1   S5   0 0 0 0 0 1 0 0        Q2   S6   0 0 0 0 0 0 1 0        Q3   S7   0 0 0 0 0 0 0 1        Q4          Starting from the initial condition, with zero active and reactive power injection by the IN, successively each IN control vector Si is triggered with a small amplitude dSi and a power ﬂow calculation is performed. The resulting change dVi of the voltage band is calculated. Performing this calculation for each vector Si gives the gradient of the objective function for each vector. Subsequently, a linear combination of S1 until S7 is applied, where each vector Si has an amplitude that is proportional to dVi. In this way, per iteration step, the maximum reduction of the voltage band is obtained by applying a linear combination of the vectors. Repeating these steps n times results in the optimal settings for the IN to minimize the voltage band. Functional concept of the Intelligent Node 67

apply actual loads in model

perform loadﬂow

calculate global voltage range

apply dSi

perform loadﬂow =1..7

calculate change global voltage range dVi i for ﬁnd local gradient

apply combination of S1..S7

perform loadﬂow

calculate global voltage range iterations n ﬁnd optimal controller settings seconds

apply controller settings m every respond to changes of load and DG

Figure 4.10: Flowchart for voltage proﬁle optimization using Cauchy’s gradient method. 68 Chapter 4

Subsequently, the settings are used as set-points by the IN. This optimization process can be repeated regularly, in order to respond to changing power ﬂow and network conditions.

4.2.1.2 Optimization results The described method is applied to the network of Figure 4.8. The boundary conditions for the optimization method are the power rating of the IN, 1 MVA for each of the IN ports, and the maximum total reactive power injection by the IN, which is set to 1.5 Mvar. With these limitations in place, the described optimization process is implemented in the software tool DIgSILENT PowerFactory. The resulting iteration path of active and reactive power, and the resulting voltage band and the total reactive power injection are shown in Figure 4.11. The calculated voltages are given in per unit values, where 1 p.u. equals 10.5 kV. In the optimization process the constraints due to network component ratings are not implemented yet. However, inspection of the results shows that no components are overloaded. From the results, it can be seen that with the IN starting set-point equal to zero, the voltage band is 4.1 % and that with optimal set-point the IN reduces the voltage band to around 0.57 %. In Table 4.1 these simulation results are given numerically, and also the results for configurations II (IN replaced by busbar) and III (same as II but with reactive power injection) are given. For configuration II and III the given Pi and Qi values represent the power flow through the corresponding connections between the IN-replacing busbar and the network. Configurations II and III result in a voltage band of 1.26 % and 0.78 % respectively. From these results it can be concluded that in the example network, a large voltage band reduction is obtained by changing from radial to meshed operation. A further reduction is obtained by injecting reactive power. Finally, optimizing the active and reactive power flow reduces the voltage band even more. Thus, for the example network it can be concluded that the voltage performance that is achieved with the application of an IN is comparable to the results in meshed operation with reactive power injection. As stated earlier, an actual implementation of an optimization method as described above would also need to take into account other factors, such as, for example, network losses. The method described above requires measurement data from several network locations. It is likely that this information is available in the central control room that controls the network that the IN is part of, rather than at the IN location. In such a central location the optimal IN settings can then be determined. Research is performed on determining the optimal IN settings only from locally available information [83, 120]. This is an interesting development, which would make the IN operation independent of Functional concept of the Intelligent Node 69

1 1

0.5 0.5

0 0 (Mvar) (Mvar) 2 1 Q Q -0.5 -0.5

-1 -1 -1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

P1 (MW) P2 (MW)

1 1

0.5 0.5

0 0 (Mvar) (Mvar) 3 4 Q Q -0.5 -0.5

-1 -1 -1 -0.5 0 0.5 1 -1 -0.5 0 0.5 1

P3 (MW) P4 (MW)

0.04 1.5 1 0.03 0.5

0.02 (Mvar) 0 i Q

Σ -0.5 0.01 Voltage band (p.u.) -1

0 -1.5 10 20 30 40 50 60 10 20 30 40 50 60 iteration n iteration n Figure 4.11: Conﬁguration I: Iteration path towards optimal IN settings: active and reactive power set-points, voltage band and total reactive power injection. 70 Chapter 4

Table 4.1: Simulation results voltage proﬁles.

I: IN II: meshed III: meshed + Q Voltage band (p.u.) 0.00567 0.01262 0.00777 V1 (p.u.) 0.996 V (p.u.) 0.996 2  0.991  0.996 V (p.u.) 1.000 3   V4 (p.u.) 0.999   ◦ φ1 ( ) -0.717 ◦   φ2 ( ) -0.079   ◦  0.002  -0.507 φ3 ( ) -1.344 ◦   φ4 ( ) -1.098   P (MW) -0.03 0.15 0.09 1   P2 (MW) -0.91 -1.03 -0.98 P3 (MW) 0.36 0.37 0.37 P4 (MW) 0.58 0.52 0.52 Q1 (Mvar) 0.36 -0.10 0.70 Q2 (Mvar) 0.13 -0.19 0.39 Q3 (Mvar) 0.47 0.13 0.17 Q4 (Mvar) 0.52 0.16 0.23 communication and thus more robust. In this thesis, the optimization process is not further developed and not included in the list of requirements for the IN. The already listed IN requirements are suﬃcient for controlling voltage proﬁles.

4.3 Voltage dip mitigation

A third application of the IN involves the mitigation of voltage dips, with as potential benefit the reduced disturbance of connected customer’s processes. Dips are caused by current surges due to e.g. a short-circuit, motor starting or the energization of a transformer. The duration of a dip due to a short-circuit is determined by the time needed for the protection equipment to detect the short-circuit, the duration of the opening process of the circuit breaker and the protection system’s delay time. The delay time is chosen such that selectivity amongst different protection systems is ensured. The duration of a dip due to motor starting or transformer inrush is a characteristic of the type of equipment. The observed depth of a dip is determined by the short-circuit power of the power system at the location of its cause, the location of the observer, grid impedances and the magnitude of the current surge. A situation in which dips cause problems can be solved my mitigating the voltage dips Functional concept of the Intelligent Node 71 in the grid and/or by increasing the dip immunity of equipment. The proposed IN application concerns the mitigation of the dips themselves by reducing their depth by controlling the power flow. The effect of power flow control on the voltage amplitude depends on the impedance of the network. In a network with inductive impedance, a reactive power flow changes the voltage amplitude, while in a resistive network, the voltage amplitude is changed by active power flow, as explained in Chapter 3. To obtain the maximum change of voltage amplitude in a network with mixed resistive/inductive impedance, active and reactive power must be injected with a ratio of P/Q equal to the R/X ratio of the network impedance. To assess the feasibility of this IN application, we consider the cable network as shown in Figure 4.12a. We assume that in this network the R/X ratios of the network impedances as seen from the IN are equal to 2 and that a voltage dip occurs due to a fault on feeder d. The resulting voltage dip on the MV busbar is observed on all feeders. To mitigate this dip, active and reactive power can be injected from the IN. However, if no storage is connected to the IN, the sum of active power delivered by the IN must be zero, rendering it impossible for the IN to inject P and Q in the optimal ratio of 2:1 on each of its ports. Therefore, if no storage is connected, only reactive power can be used for voltage dip mitigation. This results in only limited ability of the IN to mitigate voltage dips in the example network. However, in some cable grids each cable is equipped with a series reactor connected between the MV busbar and the cable, as shown in Figure 4.12b. The purpose of these reactors is to limit the short-circuit current in the connected cables. The series inductive impedance also makes it possible to use reactive power to increase the voltage amplitude on the healthy feeders. This allows the IN to operate as an independent STATCOM on each of its ports and to support the voltage during disturbances.

4.3.1 Example application To illustrate the potential of an IN to mitigate voltage dips in a cable network, the medium voltage network of Figure 4.8 is considered. This is the network that was used to illustrate the IN application of controlling voltage proﬁles. In Figure 4.13, this network is expanded with the reactors L1,L2 and L3, and again no storage is connected to the IN. Also the location is indicated where a short-circuit is applied in order to create a voltage dip. The voltage dips in four situations are calculated. In the base case no reactors are applied and no reactive power is injected. In the other three situations combinations of reactors and/or reactive power injection from the IN are applied. In the situations with reactive power injection, the amount of reactive power which is used is equal on all ports and is limited by the maximum output current (1 MVA at 72 Chapter 4

Q a

b Q

c Q

Q Intelligent Node d

(a) Limited mitigation in mixed resistive/inductive network.

Q a

b Q

c Q

Q Intelligent Node d

(b) Improved mitigation with series reactors. Figure 4.12: Voltage dip mitigation in cable network, IN without storage.

10.5 kV 55 A) of the IN. The reactors’ rating is 6 MVA with U = 4%. ≈ k The calculation results are given in Table 4.2, where VMBB is the main busbar voltage, VT Rn is the voltage on the secondary side of reactor Ln and VINn is the voltage on port n of the IN.

Table 4.2: Simulation results voltage dip mitigation example network.

Base Base Case Reactors Reactors Case and Q and Q

ISC (kA) 4.68 4.74 3.29 3.33 VMBB (p.u.) 0.854 0.864 0.845 0.855 VTR1 (p.u.) - - 0.845 0.862 VTR2 (p.u.) - - 0.600 0.608 VTR3 (p.u.) - - 0.841 0.870 VIN1 (p.u.) 0.838 0.863 0.825 0.870 VIN2 (p.u.) 0.850 0.867 0.841 0.864 VIN3 (p.u.) 0.827 0.859 0.815 0.865 VIN4 (p.u.) 0.829 0.859 0.816 0.865

From the calculation results it can be seen that the short-circuit current is limited signiﬁcantly by the application of the reactors. In the situation with reactors the short-circuit has a large reactive component, Functional concept of the Intelligent Node 73

6900 feeder length (m) 240Al cross section (mm2) and material 0.08 total load per feeder (MVA) normally open

150/10.5 kV 2x66 MVA 20 %

L1 L3 6900 1900 240Al 150Al 0.08 0.32 4900 400 4600 150Al 150Al 35Cu 0.32 0.04 0.64 2500 500 400 150Al 150Al 150Al 0.16 0.04 0.08 1 3 6600 900 1100 Q Q 150Al 150Al 150Al IN 0.08 0.04 0.04 1300 3400 Q2 4 Q 240Al 25Cu 0.16 0.04 800 600 1600 35Cu 35Cu 35Cu 0.16 0.08 0.32 700 3200 3200 1200 95Al 95Al 35Cu 35Cu 0.00 0.08 0.16 0.16

Figure 4.13: Medium voltage cable network with IN and reactors.

and due to this the voltage on the main 10.5 kV busbar is not significantly different from the base case, where the short-circuit current has a large active component. The use of reactive power to mitigate the voltage dip results in an increase of the remaining voltage at the medium voltage busbar of around 0.01 p.u. on the non-faulted feeders when no reactors are applied. With reactor the voltage dip on the secondary side of the reactors is mitigated with around 0.017 to 0.029 p.u. Further down the feeders, at the terminals of the IN, higher mitigation is achieved. The achieved voltage dip mitigation depends on the maximum amount of reactive power injection by the IN and on the grid impedance. The impedance of the reactor is typically not more than several percents, which means that the maximum induced voltage across the reactor cannot be more that a few percents either, assuming the reactive power current stays within the nominal feeder ratings. This limits the application of the IN to mitigate voltage dips in cable networks and therefore it is is likely only a secondary benefit from an IN application for a different use. In overhead line networks, the network impedance is mainly inductive and reactive power injection by the IN is more useful. 74 Chapter 4

4.4 Possible Intelligent Node topologies

The optimal internal topology and technology of an IN depend on the requirements. Not all mentioned functional requirements are necessary in each IN application. In some applications, the power exchange will, for example, always be in one direction, or only concern reactive power injection or consumption. Other applications may only require meshed operation, and do not need the ability to supply a radial network section. In this paragraph, possible IN topologies are given, along with their typical capabilities. As available building blocks of the IN we consider power electronics controlled auto transformers and impedances, and power electronic converters.

4.4.1 Power electronics controlled auto transformers An auto transformer consists of a single coil for each phase, which has a neutral/ground connection and two different winding connections, one for the input, and one for the output. The auto transformer can be seen as a series voltage source, with an inserted voltage that is in phase with the network voltage. The amplitude of the inserted voltage is controlled by adjusting the tap changer. This is a technique applied in radially operated medium voltage systems to compensate the voltage drop along a line. The tap changer position in such an application is normally fixed. Applying power electronics to operate the tap changer prevents mechanical wear and tear and the tap changer can be operated continuously. Multiple outputs can be created by connecting several arrays of power electronics valves to a number of taps of the autotransformer. Each array selects its own tap position and thus controls the voltage on that feeder. By coordinating the various tap positions, the voltage amplitude differences between several parts of a meshed network can be controlled. Such a configuration is given in Figure 4.14a, where only one of the three phases is shown. By controlling the voltage amplitude differences between the different ports, reactive or mixed reactive/active power flow is controlled, depending on the network impedances, as was shown in Chapter 3. In a network with inductive impedances, such as an overhead line network, only reactive power is influenced and there this configuration can thus only be used to control voltage profiles, not for active power flow control. Note that this configuration only circu- lates reactive power, and does not generate any (i.e. the sum of reactive power injected through all ports is zero). In a network with mixed resistive/inductive impedances, such as a cable network, active and reactive power flow can be controlled, with a ratio equal to the ratio of the resistive and inductive components of the network impedance. No active or reactive power is generated, so both the active and reactive power Functional concept of the Intelligent Node 75

(a) Single line diagram. Q Q Inductive Inductive

Q0 P /Q R/X 0 0 ∝ P P0 P Capacitive Capacitive

(b) Operating characteristic inductive (c) Operating characteristic mixed in- grid impedance. ductive/resistive grid impedance. Figure 4.14: Intelligent Node consisting of a multi-output power electronic controlled auto transformer.

add up to zero for all ports. The operating characteristics shown in Figure 4.14b and c illustrate the limited degree of freedom in controlling the power ﬂow. Coupling networks using a power electronics controlled auto transformer contributes to the short-circuit level in the connected networks.

4.4.2 Power electronics controlled series impedances The application of a TCSC in a network allows the control of power ﬂow in a network, as discussed in Chapter 3. Connecting multiple controllable series impedances allows the exchange of power between diﬀerent circuits, such as shown in Figure 4.15. By adjusting the reactive se- 76 Chapter 4

Figure 4.15: Intelligent Node consisting of multiple controlled impedances.

ries impedances, the reactive component of the network impedance is increased or decreased and thus the power flow is influenced. In a cable network, where the resistive component of the network impedance is dominant, this method is less efficient. To effectively increase the network impedance in that situation involves the application of series resistors, which is undesirable because of the resulting energy losses. Decreasing the impedance implies inserting a negative resistor, i.e. a power source, which is contradictory to the concept of adjusting impedances.

4.4.3 Power electronics converters A versatile configuration for an IN consists of a number of power electronics converters, connected by their DC side, as shown in Figure 4.16a. This configuration allows the independent control of reactive power on each of its ports and can exchange active power freely among its connections. In principle any number of converters can be used, although practical applications will not likely involve more than four converters, since in practice the number of feeders ending in one geographical location is limited. The number of converters affects the rating of the DC bus, which depends on the expected voltage and load unbalance (see also Appendix A), the rate of expected power flow changes, the PWM frequency of the converters and the control speed of the DC bus voltage controller. The optional energy storage can be connected to the DC bus. The voltage amplitude, phase angle, and frequency of the different ports do not need to be equal. The operating region of this configuration per port is equal to the operating region of the back-to-back device, as shown in Chapter 3 and repeated here in Figure 4.16b. Each converter can operate in any P, Q point within the circular contours, with the boundary condition that the sum of the active power of all ports is zero. Functional concept of the Intelligent Node 77

V = 0.9 p.u.

Q ac Vac = 1.0 p.u. Vac = 1.1 p.u.

Imax

Maximum DC link power (a) Single line diagram (b) Operating characteristic Figure 4.16: Intelligent Node consisting of multiple converters.

This results in a versatile and ﬂexible device to control the power ﬂow. Power electronic converters typically do not contribute to the short- circuit current in the network.

4.5 Conclusion

In this chapter three IN applications were discussed: • Facilitating increased network loading by controlled sharing of redundancy or by controlled power exchange between grid areas. • Controlling voltage profiles to facilitate integration of distributed generation. • Voltage dip mitigation. It is concluded that facilitating increased loading is the most important IN application and that the control of voltage profiles and voltage dip mitigation only offer limited benefits when compared with alterative solutions. From the mentioned applications the following functional requirements were formulated: • Inject or consume an adjustable amount of active and/or reactive power through each of its AC ports. • Supply a radial network from any of its AC ports. • Improve the power quality of the connected networks. • Optionally: store energy. 78 Chapter 4

• Detect a permanent short-circuit and de-energize the appropriate feeder.

• Ride through a voltage dip, which occurs due to a permanent fault on an adjacent feeder.

• Detect the opening of the load-break switches, which isolate a network section and change from controlling power ﬂow in a meshed network to supplying a radial network part.

• Synchronize the voltages on both sides of a remote opened load- break switch.

• Detect the closing of the load-break switch which restores meshed network operation and change from supplying a radial network part to controlling power ﬂow in a meshed network.

• Support the above also for the phase-by-phase operation of the load-break switches.

The following three possible topologies for the IN were described:

• Multiple power electronics controlled auto transformers. • Multiple power electronics controlled series impedances.

• Multiple back-to-back connected power electronics converters.

In the next chapter, the controls for the most versatile of these topologies is elaborated, i.e. the multiple back-to-back connected power electronics converters. Chapter 5

Intelligent Node control and protection

In order to allow the Intelligent Node (IN) to perform the tasks as described in the previous chapter, the IN converters need to be able to respond quickly to planned and unplanned events in the power system, such as load changes, short-circuits and the opening and closing of load- break switches. The ability of the converters to do so, depends, besides on their ratings, mainly on the controls that drive them. Besides, the protection system of the IN needs to prevent the IN components and the power system from over-currents and over-voltages. In this chapter the implemented IN control and protection algorithms are described, starting at the converter level in paragraph 5.1, where the AC current and voltage controllers, the DC voltage controller and the power ﬂow control algorithm are described. In paragraph 5.2, it is explained how the IN responds to uncontrolled power system events, such as the occurrence of a permanent or temporary short-circuit or under- voltage. Also the protection concept is described here. In paragraph 5.3, the detailed description is given of the IN behavior during controlled power system events, such as the transition from supplying a radial network to controlling the power ﬂow in meshed network operation (from P Q to V control) upon closing of a sectionalizer, and also for the reverse process after opening of a sectionalizer.

5.1 Basic converter controls

At the converter level two basic operating modes exist: power flow control and voltage control. In the first operating mode, which we call the P Q control mode, the converter follows the voltage of the grid that it is connected to and controls the power flow. In the second operating mode, which we call V control mode, the converter defines the amplitude, fre-

79 80 Chapter 5 quency and phase angle of the voltage on its AC port, and supplies or consumes the active and reactive power as required to and from the connected loads and generators. A converter must be operating in P Q control mode, if it has a synchronous connection to the network that has voltage amplitude, frequency and phase control, such as the public power system. A converter must operate in V control mode, when connected to a network that has no voltage amplitude, frequency and phase control, such as a feeder with only loads, and/or DG that does not have such controls and follows the grid voltage. In the proposed IN concept, at least one of the converters of the IN is galvanically connected to the ’central grid’, and operates in P Q control mode, in order to supply the connected sections and to control the DC bus voltage. In Figure 5.1 an example of an IN application is given, with the operating mode indicated per converter. In this chapter, it is assumed that a human oper-

Intelligent Node PQ PQ

V V

Figure 5.1: Basic converter operating modes.

ator or an automatic process provides the IN with active and reactive power set-points for each converter that operates in P Q control mode, for example based on the optimization process described in the previous chapter, when maximizing the penetration of DG, paragraph 4.2. The basic blocks of the converter control structure (software) are shown in Figure 5.2, along with the converter components (hardware). The DC side of each of the converters is connected directly to one common DC bus capacitor, while each AC side is connected to a diﬀerent part of the power system through an LC ﬁlter. The heart of the control topology for each converter is a current control loop, which is always active. This is a fundamental choice, common in modern converters, which allows current waveform control, peak current protection, overload rejection and good dynamics [121]. Depending on the operating mode of the converter, the current reference is provided by an active and reactive power controller (P Q Controller) or by the AC voltage controller (V - controller). A feed-forward signal is added to the output of the current controller for improved dynamic response. The simultaneous switching Intelligent Node control and protection 81

Software Hardware

S2 0.5 . C Feed Forward dc

S1 I Vpwm Vac∗ PR ∗ PR PWM

AC AC Voltage Current Controller Controller Iac Lf Q Vac ∗ f( ) PQ Controller P ∗

2 V PI DC Voltage Controller dc∗ V dc Cf x2

Figure 5.2: Converter with control system.

of the software switches S1 and S2 changes between P Q and V control mode. The shown switch positions correspond to V control mode.

5.1.1 Controller discretization In the practical set-up, which is described in Chapter 6.1, the IN and its converters are controlled using a digital signal processor (DSP). This implies that discrete control methods are used, which are speciﬁed in the z domain. Here, the various control and converter transfer functions are given in the s-domain, using the Laplace transform. To implement the controller transfer functions, a relationship is needed between the s and the z domain. To ﬁnd this, we look at a time-domain delay of Ts, which can be expressed in the s and z domain as

−sTs/2 1 − e = e sTs = (5.1) z esTs/2 Expanding the numerator and denominator into their equivalent Taylor series results in

n ∞ (−sTs/2) 1 n=0 n! = m z ∞ (sTs/2) P m=0 m! (5.2) 1 1 2 1 3 1 4 1 5 1 sTs + (sTs) (sTs) + (sTs) (sTs) + ... = P− 2 4 − 12 48 − 240 1 1 2 1 3 1 4 1 5 1 + 2 sTs + 4 (sTs) + 12 (sTs) + 48 (sTs) + 240 (sTs) + ... 82 Chapter 5

By truncating the numerator and denominator expressions to the ﬁrst power of s, we obtain the (1, 1)-Pad´eapproximation of 1/z [122]. Rewrit- ing this, results in the approximation of z by

1 1 + sTs z 2 (5.3) ≈ 1 1 sT − 2 s This expression is commonly referred to as the Tustin transformation [123], and can be written as 2 z 1 s − (5.4) ≈ Ts z + 1 which is used to translate the s domain controller transfer functions into the z domain. The error that is introduced by this approximation is illustrated for each controller.

5.1.2 AC current control In three-phase converter applications a commonly used current control method is a proportional integral (PI) controller applied in the synchronous dq reference frame [124]. The main advantage of this controller type is a zero error at the synchronous frequency. Disadvantages are the relatively high computational effort requirement and the need for additional synchronous reference frames to control negative-sequence and harmonic currents [125]. An alternative method to achieve zero error at the fundamental frequency is the application of a proportional resonant (PR) controller in the stationary domain [126], which is a modifica- tion of the normal resonant controller [127], allowing the width of the resonance to be adjusted and having a large, but finite, gain at the resonance frequency. This type of controller has a very high open loop gain at the fundamental frequency, resulting in negligible error at that frequency. The main advantage of this controller over the synchronous PI controller is the intrinsic ability to control both positive- and negative- sequence currents using only one controller and reduced computational requirements [125]. Furthermore, PR controllers can be extended easily to also control higher harmonic currents with zero error, by adding resonances in the transfer function for the desired frequencies. In the practical set-up used in Chapter 6.1 a PR controller is used to control the converter current. The transfer function of the resonant controller is given in Table 5.1, Equation (5.12), with Kpcc and Kicc the proportional and resonant controller constants, ω1 the resonance frequency and

ωccc the bandwidth of the resonance. The resonance frequency is chosen equal to the nominal grid frequency of 100π rad/s (=50 Hz). Figure 5.3 displays the amplitude and phase angle of the current controller transfer Intelligent Node control and protection 83

-10 HP R(s) Tustin transformation

-20

-30 Gain (dB)

-40

101 102 103 f (Hz)

HP R(s) Tustin transformation 50

0 Angle (degrees) -50

101 102 103 f (Hz)

Figure 5.3: Transfer function proportional resonant current controller.

function, as applied in the practical set-up. The solid line is the transfer function in the s domain, while the dots indicate the s domain equivalent of the z domain implementation. The control error introduced by the discretization is negligible.

5.1.3 AC voltage control Also for the control of the AC voltage a proportional resonant controller is chosen, for the same reasons as mentioned in the discussion of the

AC current controller. The transfer function of the controller HPRvac is shown in Table 5.1, Equation (5.13), with Kpvac and Kivac being the proportional and resonant controller constants, ω1 the resonance frequency and ωcvac the bandwidth of the resonance. To compensate for voltage drops in the connected power system, a line-drop compensation (LDC) algorithm is implemented, which adjusts the amplitude of the voltage reference signal according to the measured output power of the converter and the power system characteristic impedances. This droop characteristic is similar to the LDC algorithm commonly implemented 84 Chapter 5 in the tap changer controller of high to medium voltage transformers, as described in Chapter 2. The optimal droop characteristics depend on the impedances and conﬁguration of the power system, as well as on the location of loads and generators. Figure 5.4 displays the amplitude

HP R(s) 0 Tustin transformation

-10

-20 Gain (dB) -30

-40

101 102 103 f (Hz)

HP R(s) Tustin transformation 50

0 Angle (degrees) -50

101 102 103 f (Hz)

Figure 5.4: Transfer function proportional resonant voltage controller.

and phase angle of the voltage controller transfer function, as applied in the practical set-up. The solid line is the transfer function in the s domain, while the dots indicate the s domain equivalent of the z domain implementation. The high frequency components of the measured voltage signal are removed using low-pass ﬁltering, so the control error introduced by the discretization, which increases above several kHz, is negligible. In paragraph 5.3.2.2, a modiﬁcation of the given voltage controller is presented, which controls only the voltage of phase C, in the situation that phases A and B are already connected to the ’central grid’. Intelligent Node control and protection 85

5.1.4 Active and reactive power control To control the active and reactive power of the converter, the reference current signal is calculated from the measured AC voltage and the reference values for active power P ∗ and reactive power Q∗. This calculation is performed in the αβ reference frame defined by the Clarke transformation, which transforms a three-phase system in a system consisting of the orthogonal α and β components and a zero sequence component. This transformation is mostly used in electrical machines and power electronics control, and described in, for example, Chapter 3 of [128]. Expressed in αβ components, the relationship between balanced currents, voltages and active and reactive power is ∗ ∗ ∗ P = vαiα + vβiβ ∗ ∗ ∗ (5.5) Q = v i + v i − β α α β Rewriting these equations results in reference signals for the current controller equal to ∗ ∗ ∗ vαP vβQ iα = 2 − 2 vα + vβ ∗ ∗ (5.6) ∗ vβP + vαQ iβ = 2 2 vα + vβ This method only works perfectly for balanced and sinusoidal AC voltages. In case of voltage unbalance, higher harmonic components are introduced in the reference currents. These harmonics appear due to the fact that voltage unbalance causes the denominator of Equation (5.6) to be not constant in time. To inject active and reactive power that is associated only to the positive-sequence voltage, the harmonic voltages and the negative-sequence component are removed from the measured voltage, by using a low-pass filter and by substracting the negative-sequence from the measured voltage signal, using the output of the negative- sequence stationary αβ reference frame filter shown in Fig. 5.5. The parameter ω1 is the grid frequency and ωb is the bandwidth of the filter.

5.1.5 DC bus voltage control The DC bus is a central energy buﬀer connected to all converters and its main function is to provide a constant DC voltage. The voltage amplitude of the DC bus depends on the electrical energy that is stored inside and can therefore be controlled by injecting or taking power from it. The relationship between the capacitor voltage Vdc and the active power exchange P is as shown in Table 5.1, Equation (5.11). To control the 86 Chapter 5

v ω 1 α b s vα−

ω1

v ω 1 β b s vβ−

Figure 5.5: Negative-sequence ﬁlter αβ reference frame [129].

DC bus voltage a proportional integral (PI) controller with anti-windup is applied with as input the diﬀerence between the squared values of the reference and measured DC bus voltages. The output signal of the controller is added to the active power set-point P ∗ of the P Q controller. The transfer function of the PI controller is shown in Table 5.1,

Equation (5.14), with Kpvdc and Kivdc the proportional and integral constants. Figure 5.6 displays the amplitude and phase angle of the DC voltage controller transfer function, as applied in the practical set-up. The solid line is the transfer function in the s domain, while the dots indicate the s domain equivalent of the z domain implementation. The control error introduced by the discretization is negligible. The exchange of balanced three-phase reactive power does not inﬂu- ence the DC bus voltage, since no energy is drawn from the DC bus. However, in case of unbalanced reactive power exchange, such as, for example, during the mitigation of an unbalanced voltage dip, a ripple appears in the current drawn from the DC bus, with a frequency that is twice the grid frequency, as explained in Appendix A.

5.2 IN response to unplanned power system events

The converter controls described above deﬁne the dynamic behavior of the converters for each control mode. To fully utilize the capabilities of the interconnected converters, the IN control concept also includes speciﬁc detection schemes and additional control and protections, which, based on power system events, change the operating mode and set-points of the converters or shut down the IN. In the following this is described. The described events are the occurrence of power system faults, power system over-voltages and the unintentional opening of a circuit breaker leading to the creation of a radial network area only supplied from an IN converter. Also the protection against faults within the IN is treated. Intelligent Node control and protection 87

Table 5.1: Converter and controller transfer functions.

Description Transfer function

PWM signal to 1 Hinv(s) = Vdc (5.7) AC current sLf

PWM ((1, 1) Pad´e Ts 1 4 s approximation of a Hpwm(s) = − (5.8) 1 + Ts s Ts/2 delay) 4

1 AC current to Hload(s) = (5.9) AC voltage (incl. sCf + Yload resistive load P , with inductive load QL 1 Q 100π Q and capacitive Y = (P + L · + s C ) (5.10) load V 2 s 100π load QC ) ac

2 Active power to Vdc 1 Hdc(s) = = (5.11) DC bus voltage P sCdc

PR AC current 2K ω s H (s) = K + icc ccc (5.12) controller PRcc pcc 2 2 s + 2ωccc s + ω1

PR AC voltage 2Kivac ωcvac s HPRvac (s) = Kpvac + 2 2 controller s + 2ωcvac s + ω1 (5.13)

PI DC voltage Kivdc HP Ivdc (s) = Kpvdc + (5.14) controller s 88 Chapter 5

40 HP I (s) 30 Tustin transformation

Gain (dB) 0

-10

-20 10−2 100 102 f (Hz)

0 HP I (s) Tustin transformation -20

-40

Angle (degrees) -60

-80

10−2 100 102 f (Hz)

Figure 5.6: Transfer function DC voltage controller.

5.2.1 Voltage dip mitigation by injecting reactive power When the power system is in normal operation conditions and the applicable converters are in P Q control mode, the IN controls the power ﬂow according to centrally determined P and Q set-points. During a short-circuit, and the resulting voltage dip, the IN must no longer follow these set-points, but inject reactive power to mitigate the voltage dip, as described in paragraph 4.3. In order to change the P Q controller set-points, the residual voltage must be determined. In the following paragraph the speed of diﬀerent methods to do so is compared. Subse- quently, the reactive power control method is described.

5.2.1.1 Speed comparison of methods to determine residual voltage An important performance criterion for the determination of the residual voltage is its speed. Below, three methods are described and compared. The speed of each of the methods is characterized by determining the Intelligent Node control and protection 89 time delay between the start of the voltage dip and the moment that the calculated residual voltage amplitude falls below 0.9 p.u. This delay is calculated for synthetic dips with different amplitudes, starting at different phase angles and with different phase angle jumps.

Single-phase half-cycle r.m.s. calculation This calculation method calculates a single-phase r.m.s. voltage in a sliding time window with a length of a half-cycle for a sampled signal [130]:

1 N V [n] = (v[n N + i])2 (5.15) rms vN − u i=1 u X t fs with N = 2f0 , fs the sampling frequency of the digitization, f0 the grid frequency of 50 Hz and n the moment for which the r.m.s. value is calculated. Calculations were performed for diﬀerent voltage dips and Figure 5.7 shows the resulting delay. The delay for shallow dips

0.9

0.8

0.7

0.6 -90◦ -70◦ 0.5 ◦ -50 -30◦ 0.4 -10◦ 10◦ 30◦

Residual voltage (p.u.) 0.3 50◦ 70◦ 0.2 90◦ PAJ 0.1 0 2 4 6 8 10 Detection delay (ms) Figure 5.7: Calculation delay for single-phase r.m.s. detection method for voltage dips of diﬀerent residual voltage, starting at diﬀerent phase angles. The thick line indicates the extreme values for phase angle jumps (PAJ) between 90° and 90°. − lasts up to 10 ms, while for deeper dips this value decreases to 5 ms. For voltage dips that are only 0.01 p.u. deeper than the threshold, the delay can increase up to 18 ms, depending on the phase angle jump 90 Chapter 5

during the dip. Depending on the angle at which the dip starts and the magnitude of the phase angle jump, the detection delay can be smaller. This method can be applied to each phase, resulting in three separate amplitude signals, one per phase. Although this method is not fast enough to allow the IN to control the voltage fast enough to eliminate PSfrag replacemen any voltage dip completely (after 10 ms a voltage lower than 0.9 p.u. is called a voltage dip), it is a method that provides a stable voltage amplitude signal.

Three-phase voltage rectification In this method, the three-phase voltages are mathematically rectified, and subsequently filtered to remove the rectification ripple. The resulting signal is an indicator for the amplitude of the voltage dip residual voltage. Figure 5.8 shows the

0.9

0.8

0.7

0.6

0.5

0.4 0◦ 10◦ 20◦ ◦ Residual voltage (p.u.) 0.3 30 40◦ 0.2 50◦ PAJ 0.1 0 2 4 6 8 10 Detection delay (ms) Figure 5.8: Calculation delay for three-phase rectification method for voltage dips of different residual voltage, starting at different phase angles. The thick line indicates the extreme values for phase angle jumps (PAJ) between 90° and 90°. −

calculated delay, for similar conditions as above, except here the dip is simultaneously applied to three phases. The calculations were performed for starting angles in a smaller range, because of the three-phase symmetry. The calculated delay is less dependent on the angle at which the dip starts and is also smaller, when compared to the single-phase r.m.s. method. The method results in one amplitude signal for all three phases. During unbalanced voltage sags, the resulting amplitude signal Intelligent Node control and protection 91 contains stronger harmonics and has an amplitude intermittently exceeding 0.9 p.u. or even 1 p.u. during dips with a residual single-phase r.m.s. voltage a little lower than 0.9 p.u. The increased harmonic content can be ﬁltered out, but makes the method too slow for adequate calculation of the residual voltage. This method is rejected given the unbalanced nature of around 75 % of the voltage dips that originate from medium voltage networks [131].

Three-phase αβ domain calculation This method uses a transformation of the three-phase voltages into an orthogonal system deﬁned by the α and β vectors (and a zero sequence component in case of ground or neutral referenced voltage systems). This transformation, the Clarke transformation, is often used in electrical machines and power electronics control, and described in, for example, Chapter 3 of [128]. A characteristic of this transformation is that a balanced three-phase voltage system results in α and β signals that describe a circle in the αβ plane with an 2 2 amplitude equal to Vα + Vβ . Evaluating this amplitude during a voltage dip and comparingq it with a value of 0.9 p.u. results in the calculation delay times as shown in Figure 5.9. The resulting delay is negligible for

0.9

0.8

0.7

0.6 -90◦ 0.5 ◦ -70 -50◦ 0.4 -30◦ -10◦ 10◦

Residual voltage (p.u.) 0.3 30◦ 50◦ 0.2 70◦ 90◦ 0.1 0 2 4 6 8 10 Detection delay (ms) Figure 5.9: Calculation delay for αβ domain detection method for balanced three-phase voltages dips of different residual voltage, starting at different phase angles. voltage dips deeper than 0.85 p.u. residual voltage and very small for 92 Chapter 5 voltage dips between 0.85 p.u. and 0.9 p.u. In other words, this method allows an almost instantaneous determination of the residual voltage of a balanced voltage dip. The method is, for balanced dips, insensitive to phase angle jumps. Similar to the three-phase rectification method, an unbalanced voltage dip causes harmonics in the calculated signal, also here intermittently exceeding 0.9 p.u. or even 1.0 p.u. during dips with a residual single-phase r.m.s. voltage a little lower than 0.9 p.u. Strong filtering would be required to remove these harmonics, rendering the method too slow for adequate detection and mitigation of unbalanced voltage dips. Also this method is rejected given the unbalanced nature of many voltage dips. To allow the IN to control reactive power injection during a voltage dip it needs an adequate residual voltage magnitude signal. Since many voltage dips are unbalanced, the αβ domain method and the three-phase rectification method are unsuitable. In order to be able to mitigate both unbalanced and unbalanced dips, the per-phase r.m.s. method is most suited to determine the voltage amplitude and detect voltage dips.

5.2.1.2 Reactive power support If the calculated residual voltage is smaller than the defined threshold of 0.9 p.u., the applicable converter must inject reactive power to mitigate the voltage dip. The provision of reactive power can be controlled in different ways. The method that is normally implemented in SVC and STATCOM applications, see also paragraphs 3.3.1.1 and 3.3.1.2, consists of a proportional (P) voltage feedback controller, and a constant reference voltage. The characteristic of the resulting Q V droop function is based on the error that exists between the reference− voltage V ∗ and the actual grid voltage. Also, the full rating of the converter can only be used when the voltage deviates largely from the reference voltage, as can be seen in Figure 5.10a. An alternative method uses a proportional- integral (PI) voltage feedback controller. This controller eliminates, in steady state conditions, the error between the reference voltage and the grid voltage, as far as converter power ratings allow. The proposed PI controller, which adds an offset to the reactive power set-point Q∗, has a deadband and is only active when the voltage is lower than the minimum voltage Vmin (during a voltage dip) or higher than Vmax (during a voltage swell). When the converter voltage is between Vmin and Vmax, the centrally determined reference value Q∗ is used. When the grid voltage is lower or equal to Vmin, the PI controller can only add a negative offset ∗ to Q , and when the voltage is higher than or equal to Vmax it can only add a positive offset. The proposed reactive power control topology is shown in Figure 5.11 and the resulting Q V characteristic is shown in − Intelligent Node control and protection 93

V Q∗ V

∗ max ∗ V V ∗ Vmin

Freely Adjustable

Q (capacitive) Q (inductive) Q (capacitive) Q (inductive)

(a) P feedback (b) PI feedback with deadband Figure 5.10: Comparison of proportional and proportional-integral voltage control methods.

Vmax∗

Vac,error Vac PI

Vmin∗ to P Q Controller Q∗ Figure 5.11: Proposed PI controller with deadband, for voltage dip and swell mitigation.

Figure 5.10b. When comparing the two control types, the PI controller with a deadband has the advantage that it allows the IN to freely control P and Q when there is no need for local voltage control, while it oﬀers the most accurate voltage reference tracking. For this reason, the PI controller with deadband is implemented in the practical set-up of this ∗ ∗ thesis. The voltage reference values Vmin and Vmax are set to 0.9 p.u. and 1.1 p.u. respectively, which ensures maximum voltage dip and swell mitigation. The control speed of the feedback loop, which is deﬁned by the PI controller constants, must be not too high because of the limited speed at which the voltage amplitude is determined. When using the 0.5 cycle r.m.s. method, the typical delay time is 10 ms. The proposed reactive power support method is implemented in the practical set-up of this thesis. In the practical tests only balanced dips are used. To mitigate unbalanced voltage dips, the reactive power injection must be controlled per phase, which is not further examined in this research. For unbalanced reactive power support the reader is referred to research results from simultaneously performed research at Eindhoven 94 Chapter 5

University of Technology, for example reported in [129, 132].

5.2.2 IN protection concept To protect the IN primary components, i.e. the converters and the DC bus capacitors, and at the same time allow the IN to perform its functions, the design of the protection must be very dedicated. Especially during temporary grid faults and during intentional opening and closing of load-break switches, the IN must continue to operate and must provide the required support to the grid in the form of reactive power injection or making a change between voltage and power ﬂow control and should not disconnect from the grid. On the other hand, in case of a permanent fault or during unintended islanding, the IN must disconnect the corresponding converters from the power system. In the following paragraphs, the proposed IN protection concept is described.

5.2.2.1 DC bus protection The DC bus is connected to each of the converters through thermal DC fuses, disconnecting the capacitors in case of over-current. Further, the DC bus voltage is measured and, after initial charging, compared to over- and under-voltage threshold values, resulting in an instantaneous trip of all converters if any of the thresholds is exceeded. When a converter is tripped, all valves are instantaneously switched to the non-conducting state. Subsequently, the breaker, connecting the converter to the grid, is opened.

5.2.2.2 AC over-voltage protection Power electronic valves are sensitive to over-voltages and can easily be damaged due to this. To mitigate transient over-voltages, an adequate insulation coordination philosophy needs to be implemented, which includes surge-mitigation devices on all AC connections of the converters. To protect against voltage swells, the grid voltages are expressed in r.m.s. values and compared to a time independent threshold value. When the threshold value is exceeded, the applicable converters are tripped instantaneously as described in the paragraph on DC bus protection.

5.2.2.3 AC under-voltage protection An AC under-voltage can occur due to a short-circuit in the grid or the connection of a large inductive load. The IN converters are in principle not damaged or negatively inﬂuenced directly by a low AC voltage, although it may cause a lowered or increased DC bus voltage, due to the resulting reduced power exchange capability of the aﬀected converter. Intelligent Node control and protection 95

This can, in its turn, trip all converters, and thus the entire IN. During a temporary AC under-voltage, i.e. a voltage dip, the function of the IN, when in P Q control mode, is to mitigate the under-voltage by injecting reactive power into the grid. If an under-voltage occurs due to a fault that is located on the network that the IN converter is connected to, the inverter is tripped. This situation can be recognized because the under-voltage lasts longer than the normal fault clearing time of the power system and because of exceeding the fault ride-through curve that the IN must respect. If the converter is in V control mode, supplying a radial grid area, an under-voltage indicates the presence of a short- circuit. Since the IN converters are current controlled and the output currents are limited, the IN only feeds a limited amount of short-circuit current into the fault. The detection of a fault is therefore not based on over-current, but on under-voltage lasting longer than a certain time. After this time, the applicable converter is tripped and a power outage occurs on the radial network. As an alterative method, which allows the continuation of the currently implemented philosophy of short-circuit detection by over-current, an additional device can be used, which has as its sole purpose to supply the fault current, and thus trip the relevant protection devices. A prototype of such a device was presented in [133]. A more elaborate discussion of networks that can become an island and on the diﬀerent adaptive protection system settings for islanded and grid-connected operation can be found in [134].

5.2.2.4 AC over-current protection The IN converters are current controlled, with the controllers limiting the current amplitude. Due to this, over-current is not likely to happen due to events in the power system and the IN output current is, therefore, a poor indicator for faults in the power system. Should the mentioned current controllers fail to limit the current within certain limits, all valves are instantaneously switched to the non-conducting state by a separate protection system. An internal short-circuit in the IN does however result in large short-circuit currents fed from the power system. To disconnect the converter in these conditions, fuses are installed. When connected to a radial network and in V control mode, an internal fault results in an AC under-voltage and a DC over-current, both causing the applicable converter to be tripped.

5.2.2.5 AC over- and under-frequency protection Unintentional islanding occurs when the opening of a circuit breaker causes a certain part of the network to be without a galvanic, synchronous connection to the ’central grid’, in other words, to become an 96 Chapter 5 island. This can, for example, occur after a permanent fault in one of the IN-connected feeders. If the technical means to control voltage amplitude and frequency are available within the resulting island, and the role of system operator, who is responsible for the balance of load and generation, is defined within the resulting island, the island could remain energized. Here we assume that, in case of unplanned islanding, de-energization of the island is required. In paragraph 5.3.2 the situation is treated where the network must stay energized after planned islanding. In Chapter 4, these two possibilities were discussed in more detail. For the protection against unintentional islanding the network frequency is measured. An over- or under-frequency indicates a situation of islanding, since the power system frequency becomes arbitrary after it is no longer defined by the power system. This is a characteristic from the P Q controller, which calculates the current wave shape from the voltage wave shape. If the voltage wave shape is no longer defined, the IN output current has no longer the normal grid frequency, causing the converter output voltage to further deviate from 50 Hz. When islanding is unintentional, the IN is tripped and disconnected from the power system. In case of intentional islanding, the converter needs to change its operating mode and must start controlling the voltage, instead of the power flow.

5.3 IN role in planned power system events

Several planned mode transitions occur during the operation of an IN. In the following, it is ﬁrst described how the IN is energized and de- energized. Then a description is given of the role of the IN in the intentional and controlled transition process from controlling power ﬂow in a meshed power system to supplying a radial network area, after the opening of a load-break switch. Finally, the role of the IN during the opposite process is described: the (re)connection of a radial network area to the rest of the grid. During both events the load-break switch is assumed to operate either simultaneously for all three phases or on a phase-by-phase basis, the latter being normal practice in the Dutch distribution systems.

5.3.1 Energization and de-energization Firstly, the not-energized IN must be connected to the grid. Connect- ing a converter to the grid with a de-charged DC bus would damage the power electronics valves, so first the DC bus is charged to a suffi- ciently high voltage using a separate, controlled rectifier. After initial charging, the P Q controller receives set-points of 0 MW and 0 MVA respectively, resulting in reference currents equal to zero. The voltage Intelligent Node control and protection 97 feed forward of the P Q controller, as shown in Figure 5.2, ensures that the converter output voltage is equal to the grid voltage, which allows the safe connection of the converter to the grid. After connecting the first converter, the DC bus voltage is regulated by that converter by importing or exporting active power, superimposed on the active power set-point, which still equals zero. The charging rectifier is disconnected now. The connected converter can immediately inject or consume reactive power. Now also the other converters, having the same control topology with voltage feed forward, can be connected to their respective feeders. Then, active power can be exchanged amongst the connected feeders. The converter(s) in V control mode are now also connected to the grid. Before de-energizing the IN, all of the converters must stop injecting or consuming power. To this end, any converter in V control mode must be disconnected from the grid using load-break switches, and receive a zero-voltage set-point. The P and Q set-points of all converters in P Q mode are ramped to zero. Now, the switching pulses to the converters are stopped, and mechanical switches or separators are opened. The DC bus can now be decharged using decharging resistors.

5.3.2 Disconnecting grid areas In normal system operating conditions, the IN controls the power ﬂow in a meshed network. In order to perform maintenance or repair work, it can be necessary to isolate part of the network by opening the load- break switches on each of its sides, as shown in Figure 5.12. Then, the IN must control the voltage on the resulting radial network. The switchgear can be of the type that closes or opens all phases simultaneously or of the type that is operated manually per phase. The Magneﬁx switchgear is an example of the latter category and is widely deployed in medium voltage systems in the Netherlands. To maintain supply to load L2 after

Intelligent Node Area 1 LB1 F1 LB2 F2 Area 2

L1 L2 Figure 5.12: Grid conﬁguration during transition from meshed to radial operation. opening of the load-break switch LB2, it is desirable that the applicable 98 Chapter 5 converter can change from P Q control mode to V control mode without disconnecting from the grid and without excessive voltage disturbances. In the following it is described how the IN is prepared for this transition and how the transition from P Q to V control takes place.

5.3.2.1 Three-phase load-break switch opening Detection of three-phase load-break switch opening The load- break switch opening is a controlled action, which is announced in advance to the IN, and to which the IN reacts by enabling the here described detection algorithm. By only activating this detection mechanism during a limited amount of time preceding a planned network configuration manoeuvre, the risk of an unintentional detection and the resulting mode transition of the converter, is minimized. For simplicity reasons, the detection process is chosen to not have communication with remote load-break switches, since there can be many along a feeder, but to only use local measurements and the mentioned communication with a central control room. It depends on the P Q control scheme of the converter and on the loading situation how voltages and current change after load-break switch opening. The implemented P Q controller, which is described in paragraph 5.1.4, calculates the current reference signal from the measured grid voltage amplitude and wave shape. When the voltage is no longer defined by the grid, it becomes, if not zero, arbitrary due to coincidental resonances between capacitive and inductive grid elements and loads. The calculated current reference thus becomes arbitrary too, creating a situation that should only exist for a short time. The arbitrary voltage has a large deviation from nominal values, making it relatively easily to detect the opening of the load-break switch. Several parameters of the grid voltage are candidates to be used for detection, such as the r.m.s. values or frequency of the voltage, or their time derivatives. The amplitude of the grid voltage is a parameter that has a rather broad band of values in normal operation conditions. The standard EN50160 [64] that describes the voltage characteristics in public medium and low voltage distribution networks states that 95 % of the 10 minute average voltage amplitude values are between plus and minus 10 % of the nominal value, and that all 10 minute values are between 15 % and +10 %. To prevent false detection, this would imply relatively− wide detection threshold values, if the voltage amplitude was used as a triggering parameter, making the detection slow. On top of that, false detections would not be prevented, since occasionally, the voltages can be out of the mentioned bands even in normal operation. This also remains true if the expected future restriction of the limits, as discussed in Chapter 2, becomes effective. To determine whether the grid frequency in normal conditions is Intelligent Node control and protection 99 stable enough to be used as a parameter for load-break switch opening detection, measurements of the grid frequency have been performed over several periods. The resulting histograms are shown in Figure 5.13, together with plots of the probability that the frequency deviates from 50 Hz more than ∆f Hz. The measurements show that in normal operating conditions the grid frequency stays firmly within 100 mHz around the nominal value of 50 Hz. These measurements confirm the tight range in which frequency variations normally occur [64], i.e. 99.5 % of all 10 second values must be within 1 %, and 100 % of them between 6 % and +4 %. The large change± of the frequency after load-break switch− opening, and the small deviations in normal operation make the grid frequency a suitable parameter to detect load-break switch opening, with a small chance of false detection. An over- and under-frequency detection scheme is chosen for the detection of load-break switch opening. The grid frequency is determined using a robust phase locked loop (PLL) which is insensitive to voltage distortions and unbalance [135].

Change from P Q control to V control After detection of the load- break switch opening, the converter switches from P Q control to V control, defining the voltage on the radial network area. To prevent loads from experiencing excessive voltage amplitude, frequency or phase angle jumps after the opening of the load-break switch, the voltage reference for the V controller must be a continuation of the grid voltage after opening of the load-break switch. The control topology as shown in Figure 5.14 is proposed to ensure this. The amplitude, frequency and phase angle of the locally measured grid voltage are calculated using the same PLL that is mentioned before. From the obtained frequency, together with the amplitude, a three-phase voltage reference oscillator is constructed, which is running in parallel with the rest of the controller functions. To synchronize this oscillator with the grid, the phase angle of the oscillator is determined using a PLL, and compared with the phase angle of the grid-PLL. The resulting angle difference ∆φ serves as an input to a PI controller that modifies the phase angle of the oscillator, thus synchronizing the oscillator with the grid voltage. The PI controller is relatively slow, and its response within the time between opening of the load-break switch and the moment of detection is negligible so that no phase shifting of the oscillator occurs. After detection of the opening of the load-break switch, the operating mode of the converter is instantaneously switched from P Q to V control by operating the software switches S1 and S2, which are shown in Figure 5.2. The ∗ output of the oscillator is used as the reference signal Vac for the AC voltage controller. The grid voltage amplitude and frequency, and the PI controller output are sampled at the moment of detection and ramped 100 Chapter 5

1 Duration:

) 21:10h f

40 ∆ Period: >

| Friday 0

f 0.5 20090207

20 − 16h02 f Density (%) | until ( Saturday P 0 0 20090208 49.9 49.95 50 50.05 0 0.05 0.1 13h13 f (Hz) ∆f (Hz)

1 Duration:

) 1:36h f

40 ∆ Period: >

| Monday 0

f 0.5 20090202

20 − 12h09 f Density (%) | until ( 13h45 P 0 0 49.9 49.95 50 50.05 0 0.05 0.1 f (Hz) ∆f (Hz)

1 Duration:

) 1:58h f

40 ∆ Period: >

| Monday 0

f 0.5 20090202

20 − 16h22 f Density (%) | until ( 18h20 P 0 0 49.9 49.95 50 50.05 0 0.05 0.1 f (Hz) ∆f (Hz)

1 Duration:

) 8:24h f

40 ∆ Period: >

| Wednesd. 0

f 0.5 20090204

20 − 9h56 f Density (%) | until ( 17h30 P 0 0 49.9 49.95 50 50.05 0 0.05 0.1 f (Hz) ∆f (Hz)

1 Duration:

) 0:10h f

40 ∆ Period: >

| Monday 0

f 0.5 20090209

20 − 7h26 f Density (%) | until ( 7h36 P 0 0 49.9 49.95 50 50.05 0 0.05 0.1 f (Hz) ∆f (Hz) Figure 5.13: Frequency density and probability plots of ﬁve frequency logs. The bin width of the frequency density plots is 10 mHz. Intelligent Node control and protection 101

ωt ∆φ Vac ω PI V detect | | LB PLL opening

to S1 and S2 ttr ω nom V sin(ωt + φ) Vac∗ Vnom | | | φnom|

Figure 5.14: Control topology for detection of load-break switch opening and transition from P Q control to V control. towards nominal amplitude, frequency and phase angle values in a time interval of ttr seconds. The time interval must be chosen long enough to prevent ﬂicker. Based on the power quality limits as elaborated in paragraph 2.3, a value of 5 s is proposed.

5.3.2.2 Phase by phase load-break switch opening Detection of opening of load-break switch in phase C In the proposed concept it is assumed that, if the load-break switch is opened in a controlled manner on a phase-by-phase basis, the order in which the individual phases are opened is always the same: phase C is opened first, followed by phase B and finally phase A. Prior to the announced opening of phase C of load-break switch LB2 (LB2-C), the converter is in P Q control mode. After opening of LB2-C, the voltage between phases A and B, vab, is still defined by the grid, but the voltages vbc and vca are no longer defined, which results in an unbalance of the voltage. To detect the opening of LB2-C, the negative-sequence component of the voltage is calculated using the filter given in Figure 5.5. As a threshold level, a value of 0.05 p.u. is used. Typical voltage unbalance in public low and medium voltage networks does not exceed 3 % or 0.03 p.u.

Change from P Q control to voltage control of phase C After detection of the opening of load-break switch LB2-C, the converter is instantaneously switched from P Q to V control. The voltage between phases A and B is deﬁned by the grid and from this (measured) voltage, ∗ an orthogonal reference voltage vαβ system is calculated using the orthogonal system generator (OSG) shown in Figure 5.15. This reference voltage system is used as a feed forward signal to the PWM block and ∗ forms the basis for the calculation of the phase C reference voltage vc , as shown in Figure 5.16. The measured voltage of phase C is compared with this reference signal and the error signal is used as the input to a PR voltage controller with the same controller characteristics as the 102 Chapter 5

ǫ K ω0 1 v s v′

LPF

1 ω s 0

vq′

Figure 5.15: Orthogonal system generator (OSG) [136]. controller used to control three-phase voltages. Both the measured and reference phase C voltages are calculated with an artiﬁcial neutral voltage as reference. The PR-controller output is the reference value for the

c Voltage Feed Forward c a b a b vα∗ vab∗ vα vab vαβ∗ 0.5 2/3 vαβ OSG to v∗ vβ v c to bc vab∗ ,vbc∗ vbc∗ vab,vbc vβ∗ v 0.5 2/3 c

i iαβ∗ c∗ f() PR ic,q∗ OSG

Figure 5.16: Voltage control two-wire grid connection.

∗ phase C converter output current ic . Again, an OSG is used to derive from this signal a balanced current reference system, thus obtaining a balanced power supply from the converter. In case of balanced loads on the converter side of LB2, a minimum current through phase A and B of LB2 is obtained. In case of unbalanced loads, the current through phases A and B of LB2 increases, proportional to the load unbalance.

Detection of opening of load-break switches in phases A and B After the opening of LB2-B the voltage vab is no longer defined by the grid. A characteristic of the proposed single-phase voltage controller, which is initially active at that moment, is that the calculated reference current only has a positive-sequence component. Furthermore, Intelligent Node control and protection 103 the OSGs, which are part of the controller, work as a band-pass filter around 50 Hz, which ensures that the frequency of the reference current is around that value. These controller characteristics make the positive- sequence voltage amplitude the best parameter to detect the opening of LB2-B. This signal is obtained from the measured voltages using the filter shown in Figure 5.17 and compared with threshold values of 0.9 p.u.

v ω 1 + α b s vα

ω1

v ω 1 + β b s vβ

Figure 5.17: Positive-sequence ﬁlter αβ reference frame [137]. and 1.1 p.u. Since the power system is without a reference to earth and without a neutral conductor, the opening of phase B also interrupts the path for current through phase A (eﬀects associated to parasitic impedances to ground are neglected). Therefore, the subsequent opening of phase A does not change the operating conditions for the converter and no further converter mode change is required after opening of phase A.

Change from voltage control of phase C to voltage control of phase A, B and C After detection of the opening of LB2-B, the voltage control is switched from controlling vc to controlling the voltage on all three phases. The reference voltage is a continuation of the measured three-phase voltage and after opening of LB2-B the amplitude, frequency and phase angle are ramped towards nominal values.

5.3.3 Connecting grid areas After the maintenance or repair work, as described in the previous section, is ﬁnished, the radial part of the grid is to be reconnected to the rest of grid by (re)closing load-break switches LB1 and LB2. In the following it is assumed that LB1 has already been closed. Before closing LB2, the loads connected to the radial part of the grid are supplied through the IN, with a supply voltage that not necessarily has the same frequency, phase and amplitude as the grid voltage on the other side of 104 Chapter 5

Intelligent Node Area 1 LB1 F1 LB2 F2 Area 2

L1 L2 Figure 5.18: Grid conﬁguration during transition from radial to meshed operation.

LB2. This situation is illustrated in Figure 5.18. During the reconnection excessive voltage amplitude or phase angle changes for the load L2, or overloading of any of the grid elements or the IN is to be prevented. The geographical distance between the load-break switch LB2 and the IN can be many kilometers and communication across this distance is not fast enough to synchronize the converter in real-time. To develop a generic synchronization strategy, it is assumed that the network of Area 2, as shown in Figure 5.18, is not synchronous with Area 1, i.e. the frequency of Area 1 cannot be measured locally by the IN. Further, it is assumed that information from a certain location in Area 1 is available. The information (voltage amplitude, frequency, phase angle, power flow situation) is assumed to have a random, but limited time delay. When comparing this situation with a similar situation of closing a ring in a meshed power system, this situation is different in that sense that there can be a difference between the angles and amplitudes of the voltages on both sides of the load-break switch, but also a frequency difference. Different frequencies on both sides eliminates the usefulness of manual synchronism verification before manual load-break switch closing: the frequency difference would cause a phase angle difference within already a short time after checking synchronism. To overcome this limitation, an automatic synchronizer could be used, which verifies equal voltages on both sides of the load-break switch before automatically closing it. How- ever, in the proposed concept, it is considered undesirable to equip each load-break switch along a feeder with such a device, and in combination with manually operated switches, it is even impossible. Compared to another similar situation, namely the closing of a circuit breaker during synchronization of a single synchronous generator or an island grid [138], the situation is different in that respect that the converter can react very fast, faster than the mechanical inertia of one or more synchronous machines allows. This aspect will be used in the proposed mode transition concept. Also, such situations use real-time measurements and a feedback loop to control the generator and to operate the switch [139]. Due Intelligent Node control and protection 105 to distance, and only indirect measurements, this feedback loop is in the IN concept non-real-time and the command to the switch cannot be given since they are assumed to be manually operated. Based on the above boundary conditions, the following control strategy is proposed, which consists of the following three distinct steps:

• The grid operator announces the planned closing of the load-break switch LB2 to the IN, upon which the IN synchronizes its output voltage with the remote grid voltage, using periodically received, remote measurements.

• The load-break switch is closed which is detected by the IN and it changes instantaneously from V to P Q control, using the latest measured P and Q values of the load on the radial network part, as reference values P ∗ and Q∗.

• The IN ramps P ∗ and Q∗ to system optimal values.

In the next paragraphs, ﬁrst the synchronization of the IN with the remote grid is discussed. Second, the proposed method for detection of load-break switch closing is discussed, including the mode transition of the converter and the change of active and reactive power set-points.

5.3.3.1 Synchronization: communication and control In order to allow a smooth reconnection of the radial part to the rest of the grid, the voltages on both sides of load-break switch LB2 must be equal, or in other words, synchronized, before load-break switch closing. The maximum allowed voltage changes during the reconnection process are deﬁned by power quality limits, which are described in paragraph 2.3. Figure 5.19 shows the proposed control scheme for synchronizing the voltages on both sides of the load-break switch. In a location in Area 1,

P ∆φ remote comm. inv measurements delay Q V | |inv Vac V | |LB P φLB ωtLB ∆φ Q tmeas. ω PI GPS PLL GPS

sync. command

ttr Vnom V sin(ωt + φ) Vac∗ | ωnom| | | φnom

Figure 5.19: Control scheme to synchronize load-break switch voltages. 106 Chapter 5 measurements are performed of the power ﬂow situation and voltage amplitude, frequency and phase, and provided with a time-stamp obtained from a GPS (global positioning system) receiver. From these measurements and using the network topology, the voltage amplitude and phase angle at the location of LB2 are calculated. This information, which is the target voltage on the IN side of LB2, is transmitted to the IN location, where it is received with an arbitrary, but limited communication delay. From the locally measured active and reactive power injection, and the target voltage amplitude at the LB2 location, the required converter voltage amplitude is calculated, as well as the phase shift across the impedance between the converter and LB2. From the received frequency and phase angle information, the measurement time-stamp and a locally received GPS signal, a sawtooth signal is reconstructed that represents the phase angle of the voltage on the grid side of LB2. After receiving the synchronization command, which announces the planned closing of LB2, the amplitude and frequency of the voltage reference signal are changed from the ﬁxed nominal values, to the calculated values that are described above, in ttr seconds. Subsequently, the phase angle of the reference voltage is controlled with a PI controller, resulting in a voltage at the IN side of LB2 that is equal to the voltage on the grid side of LB2.

Synchronous frequency and angle The quality of the synchronism depends on several factors:

1. The accuracy of the frequency and phase angle measurement on the grid side.

2. The diﬀerence in clocks on grid side and on IN side.

3. The accuracy of the frequency and phase angle reproduction by the IN.

4. Frequency variations in the grid and the interval between ’synchronism updates’.

The interval between ’synchronism updates’ is deﬁned as the sum of the time between two remote measurements, the communication delay and the time needed by the PI controller to synchronize the voltage. Each of the factors inﬂuencing the synchronism quality will be discussed here under.

Ad 1: Accuracy of frequency and phase angle measurement The frequency and phase angle measurements are performed using the mentioned robust PLLs, which are insensitive to waveform distortion or Intelligent Node control and protection 107 unbalance. The accuracy of the PLL is considered to be very good, and the associated error negligible in the synchronization process.

Ad 2: The difference in clocks on grid side and on converter side To achieve an accurate voltage synchronization, it is essential that the remote and local clocks are accurately synchronized. Phasor measurement units (PMUs) are increasingly applied in power systems. The development of PMUs started in the second half of the 1980s and further matured in the late 1990s, while new applications of the technology are being developed continuously. The availability of GPS technology has made the time-synchronization of PMUs economically and practi- cally feasible, making it possible to time-stamp measurements with an accuracy of 0.5 µs, which corresponds to a maximum angle error due to clock differences of 0.009° in a 50 Hz power system [34]. This error is negligible, and error-free clock-synchronization is assumed in the presented synchronization strategy. Indeed, literature reports a successful application of GPS time-stamped phase angle measurements in a similar situation where the synchronism of two areas needed to be verified [140].

Ad 3: The accuracy of the frequency and phase angle reproduction by the converter Due to the high gain of both the current and voltage controller at 50 Hz, the amplitude and phase angle diﬀer- ence between the controller reference signal and the actual voltage is considered negligible.

Ad 4: Frequency variations in the grid and the interval between ’synchronism updates’ In the electrical power system, frequency variations occur due to changes in the loading and generation situation. The consequence for the synchronization process of the converter with the remote grid is that some time after starting synchronizing, the synchronism may be lost and then, to restore the synchronism, new measurement values of frequency and phase angle are needed. The interval between two of these frequency and phase angle measurements needs to be suﬃciently small and depends on the characteristics of frequency variations in the grid. In order to obtain practical values for frequency variations in the network, measurements have been performed. The power system frequency was measured every 100 ms during in total 33 hours, in 6 measurement sessions, of which the longest lasted longer than 21 hours and the shortest 10 minutes. Of these measurements, the following analysis is made: • Sliding time windows of diﬀerent durations are used on the measured frequency data. Based on the extrapolation of the measured frequency value at the beginning of the nth window, it is 108 Chapter 5

determined which, during the entire time window, is the maximum phase angle deviation ∆φn between this extrapolation and the actual angle. The actual angle is calculated from the series of frequency measurements within the time window. The angle deviation, in radians, after n samples is equal to

n−1 ∆φ = 2πT f nf (5.16) n s i − 0 i=0 ! X th with the sample time Ts =100 ms and fi the i measured frequency value. The maximum absolute angle deviation is determined for the entire time window. This time window is repeatedly shifted one measurement sample, while the calculation is repeated, until the end of the frequency measurement has been reached. The 100 ms sample time of the frequency measurements was chosen to have negligible eﬀect on the error on the calculation of the actual angle.

• The analysis above is performed for diﬀerent lengths of the sliding time windows. From these results, for each window length, it is calculated how big the chance is that a phase angle error of a certain magnitude is exceeded. This analysis is repeated for several phase angles and the results are plotted in Figure 5.20. One curve represents the probability (on the vertical axis) that the corresponding phase angle threshold is exceeded if the synchronization input is renewed after a certain interval (on the horizontal axis). By select- ing an acceptable chance of exceeding the angle threshold, on the horizontal axis the maximum interval can be read, which should be used to update the synchronization. This interval includes the communication delay and the time needed for the PI controller to control the phase angle.

Interestingly, the curves of diﬀerent measurements in Figure 5.20 are very similar, despite the diﬀerent measurement session durations. To explain this, a closer look is taken at the frequency control in the UCTE (Union for the Co-ordination of Transmission of Electricity) grid in which the measurements were performed. In this grid, the majority of the electrical power is produced with the use of synchronous electrical machines. The initial response of such synchronous generators to a load variation or loss of a generation unit, is a change of their rota- tional speed, and thus system voltage frequency. If the load increases or a generation unit is disconnected, more electrical power is taken from the generators, while the mechanical power from the turbines is initially still the same. This results in a decrease of the generator speed and Intelligent Node control and protection 109

1 ◦ 2 5 ◦ ◦ 0.8 10 20 ◦ 45 ◦ 0.6 90 ◦

0.4

0.2

Probability of exceeding angle difference 0 0 5 10 15 20 Update interval (s) Figure 5.20: Measured probability of exceeding phase angle difference, as function of sampling interval. Different traces represent different frequency measurements.

grid frequency. In reaction to this speed reduction, the speed governor of the turbine will increase the power output of the turbine according to a negative droop function of the frequency (if such droop control is installed): the active power injection is increased when the frequency decreases. This results in a new equilibrium, at a lower grid frequency. In ”Policy 1 Load-frequency Control and Performance” [80] of the Oper- ation Handbook of UCTE, this is called primary frequency control. The primary frequency control responds within seconds as a joint action of all involved generators. The control that regulates the frequency back to 50 Hz (and restores cross-border power exchanges to their programmed set-point values) is called secondary control and starts 15 to 30 s after a frequency disturbance. So, for the synchronization of the converter, with the remote grid, the only relevant control algorithm is the primary frequency control. It is plausible that the parameters that define this response, namely the amplitude of load and generation changes, the inertia of the generators and the joint control action of all turbine governors, are on the UCTE scale more or less constant parameters, which explains why the curves of the different measurements are so similar. The other forms of frequency control, tertiary control and time-control, are on a longer time scale and no longer play a role for the synchronization of the converter with a remote grid, and are therefore not discussed. From the measurements it is concluded that, in order to not exceed a 10° phase angle jump after closing LB2, as defined as a threshold in paragraph 2.3, 110 Chapter 5 the maximum interval between ’synchronism updates’ must be smaller than 4 s.

Amplitude and angle of voltage at load-break switch location Besides phase angle and frequency, also the voltage amplitudes on both sides of LB2 must be made equal before closing. As shown in Figure 5.19 and described in the accompanying text, the voltage and power ﬂow in the remote grid is measured and from this, the voltage on the grid side of LB2 is calculated. Similarly, the converter voltage amplitude and phase angle is calculated to match the voltage amplitude and phase angle at the location of LB2. If the grid only consists of radials, the mentioned calculations can be done analytically, otherwise power ﬂow calculations need to be performed. Here, the analytical expressions are given for the relationships between power and voltage amplitude and angle along a feeder, as shown in Figure 5.21. The general relationship between

PG,QG XG RG VL

V G load

Figure 5.21: Generator supplying load through an impedance.

the complex voltages VG and VL on the two sides of the feeder with impedance XG and RG and load with active and reactive power PG and QG is

RGPG + XGQG XGPG + RGQG VL = VG j (5.17) − VG − VG

For small amplitude differences, the imaginary part of Equation (5.17) is a measure for the voltage angle difference between sending and receiving end, while the second term for small angles gives the amplitude difference. The exact expressions for voltage amplitude and angle on the grid side of the load-break switch are

R P + X Q 2 X P R Q 2 V = V = V 1 G 1 G + 1 G − 1 G | GLB| | L| G − V V s G G (5.18) and X P R Q φ = φ = arctan 1 G − 1 G (5.19) GLB L V 2 (R P + X Q ) G − 1 G 1 G Intelligent Node control and protection 111

where R1, X1, VG, PG and QG are the impedance and the voltage and power values of the remote measurements. These values serve as target values for the voltage VILB, the voltage on the converter side of LB2. To calculate the required converter voltage, again (5.18) and (5.19) are used, now substituting PG and QG with the locally measured active and reactive power P and Q and substituting V with the target − I − I G voltage amplitude VILB, and further usin the impedances R2 and X2 of the network elements between converter and LB2. The results are VI and φ . The converter reference values for amplitude and angle are |V | X | I | and φX + φGLB.

5.3.3.2 Three-phase load-break switch closing Detection of three-phase load-break switch closing After receiving the announcement of the planned reconnection of the radial network section to the rest of the grid, the described algorithm synchronizes the voltages on both sides of LB2. The action of closing the load-break switch parallels two voltage sources: the grid and the IN converter. Perfect synchronization does not occur in practise and small amplitude, angle or frequency differences will cause equalizing currents to be exchanged between the two voltage sources, only limited by the grid impedance, which is in general small. The occurrence of a resulting large current is proposed for the detection of the closing of LB2. A detail that plays a role here is the output current limitation of the converter. The voltage controller, seeing the grid voltage that differs from its reference signal, will try to increase the current reference signal amplitude, in order to make the converter voltage equal to the reference value. How- ever, the output of the AC voltage controller is limited, so the current reference signal saturates and approaches a square wave function. The r.m.s. value of a square wave current is equal to its peak value I (i.e. 1.41 p.u.), while in normal operating conditions, up to the nominal current, this value does not exceed I/√2 0.707I (i.e. 1 p.u.), whichb is the r.m.s. value of a sinusoidal signal with≈ the same peak value. This leaves some margin for an adequate detectionb thresholdb setting that prevents false detection. An r.m.s. current threshold level of 1.05 p.u. is proposed and used in the practical set-up. In the theoretical situation that the voltages on both sides are and stay perfectly equal, no detection takes place. The drawback of this is that the IN cannot perform its task of power flow control, but connected loads keep on being supplied. In the unlikely event that no automatic detection has taken place, it is proposed to change the converter operating mode manually after verification that the load-break switch was closed. A short-circuit in the power system in the time interval between announced LB2 closing and the actual closing also triggers the proposed 112 Chapter 5

over-current detection. This leads to the converter changing its operating mode from V control to P Q control. Since the network frequency is no longer deﬁned in this situation, after a short time, the under- and over-frequency islanding detection protection will disconnect the converter. In cable systems, this is the desired IN response, since short- circuits are almost always permanent, and re-energization after a fault is in general undesired. In overhead line systems, most faults are temporary and re-energization after a fault is more common. In the proposed mode transition, this functionality is not supported.

Change from V control to P Q control After detection of the closing of LB2, the converter operating mode is instantaneously changed from V control to P Q control by the operation of switches S1 and S2 of Figure 5.2. The latest measured values of the local load P and Q are used as initial set-points for the P Q controller, and are ramped towards the system optimal reference values, as shown in Figure 5.22. The pro-

detect PSfrag replacemen I LB Iac rms P closing Vac Q

ttr to PQ controller P ∗ Q∗

Figure 5.22: Power control scheme after closing of load-break switch.

posed transition time is 5 seconds to avoid sudden changes in power ﬂow and the resulting sudden voltage changes in Areas 1, 2 and 3. The mode transition block samples-and-holds the locally measured active and reactive power P and Q at the moment of the detection and ramps these ∗ ∗ values within a time ttr to the system optimal values P and Q , which are, for example, calculated in a process as described in paragraph 4.2.

5.3.3.3 Phase-by-phase load-break switch closing

In distribution systems, not all load-break switches connect or disconnect all three phases simultaneously. Some medium voltage load-break switches are operated on a phase-by-phase basis. In the following paragraphs, the role of the IN during the phase-by-phase closing of LB2 is described. Intelligent Node control and protection 113

Detection of closing of load-break switches in phases A and B Due to the floating network without reference to earth and without neutral connection, the closing of LB2-A does not cause any change in the power flow situation. By the subsequent closing of the load-break switch of phase B, two three-phase voltage sources are interconnected by two wires. Through these two wires, an equalizing current can flow. This causes an unbalanced current delivered by the converter, which is detected by extracting the negative-sequence component from the measured converter currents by the application of the filter shown in Figure 5.5. A threshold level of 0.4 p.u. is used, which is considered a value high enough to not cause false detection by unbalance due to load or generation in medium voltage systems. In case the converter is fully loaded before load-break switch closing, the peak current limitation prevents a large negative-sequence current. Therefore, additionally, the r.m.s. converter currents are used as an additional signal to detect load-break switch closing, with a detection level of 1.05 p.u.

Change from three-phase voltage control to control of phase C voltage Before connection of any of the phases, the converter is operating in V control mode to supply Area 2 and is synchronized with the voltage of Area 1. After detection of the connection of phases A and B, the converter stays in V control mode, but now only controls the voltage of phase C, thus reconstructing a balanced three-phase voltage from the measured voltage between phases A and B. The output of the voltage controller controls the current in phase C. From this current, the reference currents for phase A and B are determined in order to create a balanced current system. In case of balanced loads, this minimizes the current exchange through the two wires connecting Area 1 and Area 2.

Detection of closing of load-break switch in phase C Before the closing of LB2-C, the converter controls the voltage of phase C, based on the measured voltage between phases A and B. Closing LB2-C connects the two voltage sources, resulting in an equalizing current due to small voltage diﬀerences. The converter voltage controller structure is such that the negative-sequence current injection is minimized, and only a positive-sequence current occurs. The detection of closing of LB2-C is based on the positive-sequence current exceeding the threshold level of 1.05 p.u. To ensure detection, the phase C voltage is controlled such that a deliberate voltage diﬀerence with the network voltage exists.

Change from control of phase C voltage to P Q control After automatic detection of closing of LB2-C or after receiving a manual command, the converter changes instantaneously from phase C voltage 114 Chapter 5 control to P Q control. The active and reactive power reference settings are initially made equal to the power that the converter was delivering before closing of LB2-C and ramped towards the system optimal values. The used transition time is 5 s.

5.4 Conclusion

In this chapter the basic IN converter controls are given, followed by a description of the control and protection structures that allow the converter to perform its task during unplanned and planned power system events, without being damaged or damaging power system components. The ﬁrst category, unplanned events, includes the mitigation of voltage dips. The second category, planned power system events, consists of the energization and de-energization of the IN and the disconnection and (re)connection of grid areas using the IN. To support the disconnection and (re)connection of grid areas the IN detects the opening or closing of a remote load-break switch by observing locally measured signals. After detection of the load-break switch operation the applicable converter control structure is changed. The proposed control algorithms support both the phase-by-phase operation of load-break switches as well as the simultaneous opening and closing of all three poles. For the connection of grid areas an important characteristic is the synchronization before load-break switch closing. Since remote voltage amplitude, frequency and phase angle information is assumed to be only periodically available, the level of frequency variations in the ’central power grid’ is important. By evaluating frequency measurements in the public network, the maximum synchronization interval was determined. Chapter 6

Laboratory-scale demonstration

To demonstrate the concept and controls as described in the previous chapters, a three-phase laboratory-scale set-up is realized. In this set-up experiments are carried out which focus on the innovative parts of the IN concept, such as the synchronization between remote grid areas, detection of load-break switch operation and the transition between different converter operating modes. The process to determine the optimum for active and reactive power IN reference values is not experimentally verified, since this is a system optimization aspect, which is not related to the functioning of the IN itself. In this chapter, first the hardware and the control features of the set-up are described, followed by a brief presentation of the measured converter response to load and reference value variations. Next, the transition from radial to meshed operation is discussed, as well as the reverse process, both for 3-phase and for phase-by-phase load-break switch operation. Then, an implementation is presented of the IN’s capability to mitigate voltage dips and swells. Subsequently, it is discussed how the detection mechanisms change when translating the concept to higher voltage and power ratings. Finally, the conclusions from the experiments are presented. Throughout the chapter, the most important experimental results that characterize the IN behavior are presented, as well as the most important simulation results of parameter variations, which were obtained using a software model of the practical set-up. In Appendix B all measurement and simulation results are given.

6.1 Experimental set-up

The set-up consists of two 3-phase 400 V IGBT Semikron SKiiP converters, connected on their DC sides to a common DC bus and to a rotating DC generator. In most of the experiments, only one converter was used. In one of the experiments, also the second converter was used. The AC

115 116 Chapter 6 connection of each converter is made through a star connected ﬁlter and a 3-phase isolation transformer. Pictures and a single-line diagram of the set-up are given in Figures 6.1 and 6.2. The three poles of load-break

Figure 6.1: Pictures of laboratory-scale set-up: converter cabinets, DSP connections, user interface, isolation transformers and load.

switch LB2 can be operated simultaneously or on a phase-by-phase basis. Area 1 is supplied either from the public low voltage network or from a programmable voltage source, a Spitzenberger & Spies DM15000/PAS. The nominal converter power is limited by the ﬁlter rating and is equal to 5.2 kVA (=1 p.u.) at a phase-to-phase voltage of 400 V (= 1 p.u.), which corresponds to a nominal current of 7.5 A. On the AC side of the converter, a resistive load can be connected, which has a power rating of 0.9 p.u. In Tables 6.1 and 6.2, an overview is given of the electrical components of the set-up and converter controllers constants used in the experiments. Laboratory-scale demonstration 117

”Area 1” LB2 ”Area 2”

ILB VGLB VCLB Public LV Reactor Network or Programmable Isolation Voltage Source Transformer 0.9 p.u. Load Control Measurements

Filter

Converter 1 Converter 2

Rotating DC bus DC Source

Figure 6.2: Single-line diagram of laboratory-scale set-up.

Table 6.1: Electrical components practical set-up

Component Parameter [unit] Value

Converter Nominal power Sn [VA] 5200 Nominal phase-to-phase voltage Vn [V] 400 Filter Capacitor [µF] 10 Reactor [mH] 2 Isolation Transformer Nominal power Sn [VA] 5200 Nominal phase-to-phase voltage Vn [V] 400 Short-circuit voltage [%] 5 Reactor Nominal power Sn [VA] 5200 Short-circuit voltage [%] 8.2 Public LV grid Short-circuit impedance [mH] 0.1

6.1.1 Converter control implementation Converters 1 and 2 are controlled by a single digital signal processor (DSP), of the type dSPACE DS1103. The controller functions, de- 118 Chapter 6

Table 6.2: Converter controller parameters practical set-up

Controller Parameter [unit] Value AC current Proportional gain 0.01 Resonant gain 0.2 Center frequency [Hz] 50 Bandwidth [Hz] 0.8 AC voltage Proportional gain 0.01 Resonant gain 1.3 Center frequency [Hz] 50 Bandwidth [Hz] 1.1 DC voltage Proportional gain 0.25 Integral gain 2 scribed in the previous chapter, are discretized, implemented in MAT- LAB Simulink and, after compilation, run on the DSP. In the software tool ControlDesk a graphical user interface is implemented which allows the adjustment of reference values and the converter operating mode, as well as the monitoring of selected control and measurement signals in real-time during the experiments. An example of this interface is shown in Figure 6.3. Due to computational limitations of the DSP, the controller functions only calculate a new PWM reference value every second PWM cycle, which results in a discretization of the controller functions with a sampling rate of half the PWM frequency, i.e. 5.7 kSa/s.

6.1.2 Modeling of experimental set-up In order to develop and verify the controls, the experimental set-up was modeled in MATLAB® R2007b with Simulink and the SimPowerSys- tems Toolbox. This model is also used to assess the effect of those parameter variations which were not experimentally determined. The analog electrical components are modeled with a discrete sampling rate of 60 times the PWM frequency, i.e. 684 kSa/s. The controller and detection blocks in the DSP are simulated with a discrete sampling rate equal to the PWM frequency, i.e. 11.4 kSa/s. This is different from the experimental set-up in which the DSP controller and detection functions operate at half the PWM frequency. This mainly impacts the effect of delay times, which occur due to the discretization of analog signals. By comparing the experimental results with the simulation results, which are both presented, it is concluded that the basic control functions and the hardware of the set-up are represented adequately in the computer model. Laboratory-scale demonstration 119

Figure 6.3: Example of interface implementation in ControlDesk software.

6.2 Basic converter step responses

The response speed of the IN as a whole is limited by the dynamic behavior of each of its converters, which in its turn is determined by the controller functions and detection methods. To characterize the individual converters in their diﬀerent operating modes, the following basic converter responses are experimentally veriﬁed:

• Step change of current amplitude while in P Q control mode

• Step change of voltage amplitude while in V control mode

• Connection of load while in V control mode

• Connection of load while one converter is in V control mode and the other converter is in P Q control mode 120 Chapter 6

As described in the previous chapter, the heart of each converter controller consists of the current controller which is always active. De- pending on the operating mode of the converter, the current reference signal is either provided by the P Q controller or by one of the voltage controllers.

6.2.1 Changing power reference values

When LB2 is closed, the converter is in P Q control mode, and the current reference signals are algebraically calculated from the measured voltage and the active and reactive power reference values P ∗ and Q∗. Therefore, determining the converter response to a step function of active and reactive power reference values, is equal to determining the converter response to a change of current reference. In Figures 6.4 and 6.5 the

∗ 1 I I 0 (p.u.) I -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (a) Simulation.

∗ 1 I I 0 (p.u.) I -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (b) Measurement. Figure 6.4: Converter response to step function in active power reference value from 0 to 1 p.u.

measured and simulated converter responses to step changes of P ∗ and Q∗ from 0 to 1 p.u. are shown. The shown signal is the current through the ﬁlter reactor, with the PWM ripple removed from the signal using a low pass ﬁlter. The converter reaches its steady state behavior within around 30 ms after the change of active or reactive power reference value. Laboratory-scale demonstration 121

∗ 1 I I 0 (p.u.) I -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (a) Simulation.

∗ 1 I I 0 (p.u.) I -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (b) Measurement. Figure 6.5: Converter response to step function in reactive power reference value from 0 to 1 p.u.

6.2.2 Changing voltage reference value When Converter 2 supplies a radial network section (LB2 is open), it is in V control mode. During mode transitions, small changes in the reference value for the voltage amplitude can occur. To assess the converter response to such changes, the voltage reference signal is changed from 0.9 to 1.1 p.u. while supplying a 0.9 p.u. load. Figure 6.6 shows the simulated and measured converter response. The converter response to a stepwise change of the voltage reference signal reaches steady state within a few milliseconds.

6.2.3 Changing load A converter in V control mode supplies or consumes the power to and from connected loads and generators as needed. In the experimental set-up, load-break switch LB2 is open. To assess the converter response to load changes, a load is connected at t = 0. This test is performed to verify the response of two controllers: the AC voltage controller and the DC voltage controller.

AC voltage controller response After connection of the 0.9 p.u. load, the AC voltage controller of Converter 2 changes the reference current signal to maintain the voltage on the AC port of Converter 2. In 122 Chapter 6

∗ 1 V V

0 (p.u.) V -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (a) Simulation.

∗ 1 V V 0 (p.u.) V -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (b) Measurement. Figure 6.6: Converter response to step function in voltage reference value from 0.9 to 1.1 p.u. while supplying 0.9 p.u. load.

this test, the rotating DC generator supplies the DC bus. The simulated and measured AC voltages responses are shown in Figure 6.7. The load

∗ 1 V V

0 (p.u.) V -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (a) Simulation.

∗ 1 V V 0 (p.u.) V -1

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 t (s) (b) Measurement. Figure 6.7: Converter response to load step from 0 to 0.9 p.u.

connection leads to a voltage dip of around 0.15 p.u. and the voltage is Laboratory-scale demonstration 123 restored to the reference value within around 60 ms.

DC voltage controller response When a change occurs in the loading situation in the network supplied by one of the converters in V control mode, the corresponding power change is taken from the DC bus. To maintain the DC bus voltage, this power change must be compensated by the other converter(s) connected to the DC bus. In order to do so, the DC bus voltage controller changes the active power reference values of the converter(s) which are operating in P Q mode. In the experimental set-up, this situation was reproduced by using both Converter 1 and Converter 2. Converter 2 supplies a radial network area (island operation) with a switchable load and Converter 1 is connected to the public low voltage grid, operates in P Q control mode and controls the DC bus voltage. During this experiment, the rotating DC generator is disconnected from the DC bus. At t = 0 a 0.67 p.u. resistive load is connected to the converter operating in V control mode. The DC controller active power and voltage response is shown in Figure 6.8. Note that the active

0.5

(p.u.) 0 dc

P -0.5

-1 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 t (s) (a) Active power (measurement).

1.04

1.02 (p.u.)

dc 1 V 0.98 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 t (s) (b) DC voltage (measurement). Figure 6.8: DC controller response to load step from 0 to 0.67 p.u. power necessary to maintain the DC bus voltage is not equal to zero before t = 0 due to the energy losses of both converters. A smaller load of 0.67 p.u. is used, to prevent overloading Converter 1. The DC voltage controller maintains the DC bus voltage deviation within 0.015 p.u. and within 0.3 s the voltage is restored to the reference value. 124 Chapter 6

6.3 Transition from radial to meshed operation

One of the functions of the IN is to facilitate the controlled transition from radial to meshed network operation. During this transition, first the two grid areas are synchronized and then connected. This connection is detected, the converter operating mode is changed and finally the power exchange is controlled to the desired values, as described in the previous chapter. In the following, first the performance of the synchronization process is experimentally determined. Next, the experimental results of the detection of load-break switch closing are presented, both for 3-phase and for phase-by-phase load-break switch operation. Finally, a discussion is given of scaling effects that have to be taken into account when applying the concept in medium voltage networks.

6.3.1 Synchronization The synchronization process consists of ensuring equal frequency, phase angle and amplitude of the voltages on the grid side and on the converter side of the load-break switch LB2, before closing it. Due to the varying frequency of the public electricity network, a critical parameter of the synchronization is the (constantness of the) phase angle difference between the two voltages. To minimize this parameter, a synchronization concept was presented in the previous chapter. To verify the adequate- ness of the proposed 4 s synchronization update interval and method, the ∗ phase angle between the grid voltage and the reference voltage Vac of Figure 5.19 was recorded every 10 ms during one hour while the synchronization mechanism is activated. During the measurement, the network frequency changes due to supply and demand balancing. The angle difference between the voltage reference signal and the actual voltage as generated by the converter is considered to be negligible. During the measurements, a remote frequency and phase angle measurement interval of 2.5 s was used, and a communication delay of 1 s. Together with the 0.5 s that it takes for the PI controller to reach steady state, this results in the proposed ’synchronism update’ interval of 4 s. A histogram of the measured phase angle difference is shown in Figure 6.9, which confirms that the use of a ’synchronism update’ interval of 4 s results in a phase angle error smaller than 10°.

6.3.2 Three-phase load-break switch closing After closing of LB2, the converter has to change from V to P Q control mode. Until this change, two voltage sources (the network and the converter) are interconnected, which results in equalizing currents between them. The time between load-break switch closing and the control Laboratory-scale demonstration 125

4 Percentage (%) 2

0 -15 -10 -5 0 5 10 15 ◦ ∆φ ( ) Figure 6.9: Angle deviation while synchronization algorithm is active (measurement). The bin width of the histogram is 0.5°.

mode change must therefore be as short as possible. The adequacy of the detection algorithms presented in the previous chapter is crucial for this. Here, the detection speed is studied for the situation of 3-phase load-break switch operation. The equalizing current between network and converter is determined by the impedance between them. In the practical set-up, the impedance consists of the isolation transformer impedance, the reactor, plus the grid impedance. The r.m.s. values of the converter currents are used to detect load-break switch closing, and a 1.05 p.u. threshold level is used. To exceed this threshold current through the mentioned impedance, which is dominated by the reactor, an angle shift of 5° is required between the voltages on both sides of the load-break switch. As can be seen in Fig- ure 6.9, such an angle shift (either positive or negative) only occurs in a very limited percentage of the time when a ’synchronism update interval’ of 4 s is used. This means that it can take some time before the required 5° angle occurs, but also that the angle, and thus the current, can be in the wrong direction, or that no detection takes place at all. To ensure adequate detection, a deliberate phase angle difference of 5° between the two sides of the load-break switch is used during the measurements, effectively shifting the histogram of Figure 6.9 by 5°. In paragraph 6.3.4 this deliberate phase angle difference is analyzed in more detail. The resulting converter currents and the voltage and current wave shapes at the location of the load-break switch are shown in Figure 6.10 for an unloaded Area 2. The horizontal and vertical dashed lines in this and other figures represent the detection threshold levels, the moments of load-break switch closing and detection. The time until detection is not constant among different experiments due to the varying angle error and 126 Chapter 6

the different point-on-wave of load-break switch operation. In this figure and in many other figure that follow, the differences between individual signal traces cannot be distinguished, since they are very similar, which is the desired situation. The scale of this figure and the other figures that follow is chosen such that only relevant differences between signals are visible.

1 (p.u.) 0.5

PSfrag replacemen abc,rms I

0 0 0.05 0.1 0.15 0.2 t (s) (a) R.m.s. converter currents (measurement).

1 Grid-side 0.5 Converter-side

(p.u.) 0 ab

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Phase-to-phase voltages Vab on both sides of load-break switch LB2 (measurement).

(A) 0 I

-1

0 0.05 0.1 0.15 0.2 t (s) (c) Current phase A load-break switch LB2 (measurement). Figure 6.10: Three-phase closing of LB2, Area 2 without load.

In Figure 6.11a the converter current is shown for the situation that a 0.9 p.u. load is connected in Area 2. In Figure 6.11b the results are shown for the same network conditions, but on a larger time scale, illus- trating the gradual change of the power reference value after load-break switch closing. The experimentally observed detection times for the loaded Area 2 are shorter or similar to those found for an unloaded network. Detection takes place within a few cycles, without visible voltage disturbances. Laboratory-scale demonstration 127

1 (p.u.) 0.5 abc,rms I

0 0 0.05 0.1 0.15 0.2 t (s) (a) R.m.s. converter currents (measurement).

1 (p.u.) 0.5 abc,rms I

0 0 0.5 1 1.5 2 t (s) (b) Transition to zero power exchange (measurement). Figure 6.11: Three-phase closing of LB2, Area 2 has 0.9 p.u. load.

6.3.3 Phase-by-phase load-break switch closing In the following paragraphs the controls that are used during the connection of two grid areas using phase-by-phase operated load-break switches are veriﬁed. Before closing any of the phases, the synchronization mechanism is activated, and a deliberate phase angle of 5° is used.

6.3.3.1 Closing of phases A and B The closing of LB2-A connects the Area 2 network by one phase only to the public low voltage network, which has no eﬀect, since there is no reference to ground. The subsequent closing of LB2-B introduces a circulating current in phases A and B, causing current unbalance. This parameter, the negative-sequence current, is used for detection with a threshold level of 0.4 p.u., as was discussed in paragraph 5.3.3.3. After detection, the converter control is changed from 3-phase voltage control to 1-phase voltage control, only controlling the voltage of phase C. Fig- ure 6.12 shows the measured signals for the connection of an unloaded Area 2. If Area 2 has connected loads, detection based on the negative- sequence current only can take a long time, or even not take place at all due to the current limitation of the converter. To also ensure adequate detection in loaded network conditions the converter r.m.s. phase currents are used as additional detection signals, with a threshold level of 1.05 p.u. Simulation results for both the loaded and unloaded situ- 128 Chapter 6

0.4 0.3 (p.u.)

− 0.2 PSfrag replacemen I 0.1 0 0 0.05 0.1 0.15 0.2 t (s) (a) Negative-sequence converter current (measurement).

1 Grid-side 0.5 Converter-side

(p.u.) 0 bc

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Phase-to-phase voltages Vab on both sides of load-break switch LB2 (measurement).

1 0.5 0 (p.u.)

I -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (c) Current phase A load-break switch LB2 (measurement). Figure 6.12: Closing of LB2-B, Area 2 without load.

ation are shown in Figure 6.13. The simulation results illustrate that both detection signals are needed: in the unloaded network the negative- sequence current reaches its theshold first, while in the loaded network the r.m.s. phase currents reach the threshold value first. As a further illustration of this, Figure 6.14 shows the results of an experiment where only the negative-sequence current is used for detection in a loaded network: the resulting detection time exceeds 2 s. The time until detection is influenced by the frequency fluctuations in the low voltage network and by the activated synchronization mechanism that tries to maintain the deliberate 5° phase angle, using periodically received frequency and phase angle information. Laboratory-scale demonstration 129

− 1 I Iabc,rms

(p.u.) 0.5 I

0 0 0.05 0.1 0.15 0.2 t (s) (a) Converter currents unloaded Area 2 (simulation).

− 1 I Iabc,rms

(p.u.) 0.5 I

0 0 0.05 0.1 0.15 0.2 t (s) (b) Converter currents loaded Area 2 (simulation). − Figure 6.13: Detection of LB2-B closing based on I and Iabc,rms.

0.4 0.3 (p.u.)

− 0.2 I 0.1 0 0 0.5 1 1.5 2 t (s) Figure 6.14: Detection of LB2-B closing based on only I−. Area 2 has 0.9 p.u. load (measurement).

6.3.3.2 Closing of phase C

Before closing of LB2-C, the converter controls the voltage of phase C, based on the measured phase-to-phase voltage Vab. This allows an accurate voltage reconstruction by the converter, and only a minimal voltage across LB2-C. Closing LB2-C in this situation would result in only minimal voltage or current changes, hampering detection. There- fore, to ensure adequate detection of LB2-C closing, a deliberate voltage ∗ diﬀerence is introduced by rotating the vector VC (Figure 5.16, paragraph 5.3.2.2) by 5°. Figures 6.15 and 6.16 show the simulation results for Area 2 without and with load. The presence of a load in Area 2 causes faster detection, and detection times vary between 0.5 and 4 periods. In the next paragraph, where scaling aspects are discussed, it is 130 Chapter 6

1 (p.u.) 0.5 rms

PSfrag replacemen I

0 0 0.05 0.1 0.15 0.2 t (s) (a) Converter phase currents (simulation).

1 Grid-side 0.5 Converter-side

(p.u.) 0 bc

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Load-break switch phase-to-phase voltage (simulation). Figure 6.15: Closing LB2-C, island without load.

1 (p.u.) 0.5 rms

PSfrag replacemen I

0 0 0.05 0.1 0.15 0.2 t (s) (a) Converter phase currents (simulation).

1 Grid-side 0.5 Converter-side

(p.u.) 0 bc

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Load-break switch phase-to-phase voltage (simulation). Figure 6.16: Closing LB2-C, island with load.

described how the required angle and amplitude of the deliberate voltage diﬀerence is determined. As an illustration of the importance of this voltage diﬀerence, Figure 6.17 shows the measured converter currents when no deliberate voltage angle is applied. The detection time is more Laboratory-scale demonstration 131

1 (p.u.) 0.5 rms I

0 0 0.5 1 1.5 2 t (s) Figure 6.17: R.m.s. converter currents after closing of LB2-C, island without load (measurement). No deliberate phase angle voltage C.

than 0.7 s, which is more than 35 periods.

6.3.4 Ensuring load-break switch closing detection In the experiments an additional voltage phase angle of 5° is introduced to ensure that the equalizing current after load-break switch closing is large enough to be detected. The amplitude of the resulting equalizing current depends on the magnitude of the impedance between the two voltage sources and on the amplitude of the (deliberate or accidental) voltage across the load-break switch before closing. The direction or angle of the equalizing current depends on the angle of the voltage ’error’ and on the R/X ratio of the network impedance. To achieve the maximum converter current with a certain voltage error, the equalizing current must be in phase with the load current as is illustrated in Fig- ure 6.18. Since the converter current amplitude is limited, over-currents in the network are prevented. To achieve the proper voltage difference, knowledge is required of the P/Q ratio of the load, which is measured locally, and the R/X ratio of the network, which is also a known value. In case of a resistive load and an inductive network impedance, as in the experiments, the required voltage difference has a 90° angle, i.e. only a phase angle shift is required, and no voltage amplitude difference. In the experiments, a 5° angle difference was used. To obtain an indication of the required magnitude of this angle in a medium voltage network, we consider the example network shown in Figure 6.18. We assume that

150/10kV Intelligent Node 66MVA, 18% Equalizing Current

Load Current

Figure 6.18: Required current directions to ensure detection. 132 Chapter 6 in this network, the short-circuit impedance of the transformer is the dominant network impedance, and that the cable impedances are negligible. Both the IN converter and the cables between the transformer and the IN converter have a rating of 5 MVA. On this voltage level, a phase angle deviation of 1° corresponds to 183 V and this would result in an equalizing current of 2.2 times the cable and converter ratings, if no current limitation would be active in the converter. This example illustrates that for higher voltage levels, a small voltage angle deviation results in a large equalizing current after load-break switch closing. It is therefore expected that the detection speed increases for higher voltage levels. Since the network frequency is constantly changing, this 1° voltage angle across the load-break switch is easily reached within one synchronization interval, see Figure 6.9. However, the sign of the voltage angle diﬀerence due to changes of the frequency can be both positive and negative, and thus the direction of the resulting equalizing current can well be the opposite of the load current, which would prevent detection. To ensure an equalizing current which is in phase with the load current, a deliberate voltage angle must be introduced which is larger than the one which can occur due to frequency changes. For low and medium voltage applications a value of 5° is proposed, as was used in the experiments. This results eﬀectively in a 5° shift of the histogram shown in Figure 6.9 and phase angle jumps during closing of LB2 are limited between 0 and 10°. If only smaller phase angle jumps can be allowed, for example because of load sensitivity, the ’synchronism update interval’ must be reduced, i.e. the remote measurements need to be transmitted more frequently and/or with a shorter communication delay, resulting in a histogram that is narrower than the one shown in Figure 6.9. This makes it possible to use of a smaller deliberate phase angle, while still ensuring that the occurring phase angle has the required sign. In networks with resistive impedances, together with the angle, the amplitude needs to be varied also, in order to reach an equalizing current which is in phase with the load current. Note that the current limitation of the converter is essential to prevent over-currents in the network.

6.4 Transition from meshed to radial operation

After the opening of load-break switch LB2, the converter must stop controlling power ﬂow, and start controlling the voltage on the resulting radial network. In order to minimize power ﬂow changes and the resulting voltage changes in all associated networks, the current through the opening load-break switch must be minimized before opening. Or, in other words, the power demand in Area 2 must be supplied by the converter. Laboratory-scale demonstration 133

6.4.1 Three-phase load-break switch opening

The opening of LB2, causes the voltage on the radial network to be no longer deﬁned. The frequency is used to detect the load-break switch opening, with threshold values of 49.5 Hz and 50.5 Hz. Figure 6.19 shows

51 50.5

(Hz) 50 f PSfrag replacemen 49.5 49 0 0.05 0.1 0.15 0.2 t (s) (a) Area 2 without load, P ∗ = Q∗ = 0 (simulation).

1 Grid-side 0.5 Converter-side

(p.u.) 0 ab PSfrag replacemen V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Area 2 without load, P ∗ = Q∗ = 0 (measurement).

1 Grid-side 0.5 Converter-side

(p.u.) 0 ab PSfrag replacemen V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (c) Area 2 with 0.9 p.u. load, P ∗ = Q∗ = 0 (measurement).

1 Grid-side 0.5 Converter-side

(p.u.) 0 ab

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (d) Area 2 with 0.9 p.u. load, P ∗ = 0.9 p.u., Q∗ = 0 (measurement). Figure 6.19: Frequency and phase-to-phase voltages LB2 after three- phase opening LB2. 134 Chapter 6

simulated and measurement results for three situations. In the ﬁrst situation, Area 2 is without load. In the second situation a 0.9 p.u. load is connected to Area 2, but the converter does not inject any power, while in the third situation the 0.9 p.u. power consumption in Area 2 is matched by the converter power injection. In the second situation, a voltage dip is observed with a residual voltage of around 0.5 p.u. This voltage dip exceeds the ITIC curve as presented in Chapter 2, and this situation should therefore be prevented as much as possible. In order to do so, it is essential that a proper estimation of the loading situation in Area 2 is made. In all conditions, detection takes place within a few cycles.

6.4.2 Phase-by-phase load-break switch opening In the following paragraphs the controls that are used during the disconnection of two grid areas using phase-by-phase operated load-break switches are veriﬁed.

6.4.2.1 Opening of phase C After the opening of LB2-C, the voltage of phase C in Area 2 is unde- ﬁned, resulting in voltage unbalance. The negative-sequence component of the converter voltage is therefore used for detection, with a threshold of 0.05 p.u., as was discussed in paragraph 5.3.2.2. Figure 6.20 shows

0.1

(p.u.) 0.05 −

PSfrag replacemen V

0 0 0.5 1 1.5 2 t (s) (a) Negative-sequence voltage (measurement).

1 Grid-side 0.5 Converter-side

(p.u.) 0 bc

V -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Load-break switch phase-to-phase voltages (no load, measurement). Figure 6.20: Opening of LB2-C. Laboratory-scale demonstration 135

the measured negative-sequence converter voltage and the voltages at the location of the load-break switch. Detection takes place within a few periods, independent of the loading situation in Area 2 or the power delivered by the converter before load-break switch opening. The observed voltage dip at the load-break switch location is a characteristic of the voltage controller, see also Figure 6.7.

6.4.2.2 Opening of phases A and B A characteristic of the single-phase voltage controller is that it ensures a balanced output current, around the nominal power frequency. There- fore, after the subsequent opening of LB2-B, the voltage in Area 2 remains at 50 Hz and stays more or less balanced, depending on how bal- PSfrag replacemen anced the connected loads are. The voltage Vab is no longer deﬁned by the grid however, which causes the voltage amplitude to change. The positive-sequence component of the voltage is used to detect the opening of LB2-B. Figure 6.21 shows the measured positive-sequence converter

No Load 1.1 0.94 p.u.

(p.u.) 1 + C V 0.9

0.8 0 0.5 1 1.5 2 t (s) (a) Positive-sequence converter voltage (measurement).

1 0.5

(p.u.) 0 bc

V -0.5 Grid-side -1 Converter-side 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 t (s) (b) Load-break switch phase-to-phase voltages (no load, measurement). Figure 6.21: Disconnecting phases A and B.

voltage and the voltages at the location of the load-break switch. Again, the observed voltage dip is a characteristic of the voltage controller. Detection takes place within around 0.5 s. The subsequent opening of LB2-A has no eﬀect. 136 Chapter 6

6.5 Voltage dip and swell mitigation

To illustrate the capability of the IN to mitigate voltage dips and swells, in the practical set-up the connection to the public low voltage network is replaced by a connection to a programmable voltage source. This voltage source is used to expose the IN to balanced voltage dips and swells of various amplitudes. The phase angle and frequency of the output voltage are kept constant during the experiments. Figures 6.22 and 6.23 show the resulting voltage dip and the amount of reactive

1.2

(p.u.) 0.8 V No Mitigation 0.6 Mitigation 0 0.05 0.1 0.15 0.2 t (s) (a) Voltage amplitude (measurement).

0.5

0 (p.u.)

Q -0.5

-1 0 0.05 0.1 0.15 0.2 t (s) (b) Reactive power injection (measurement). Figure 6.22: Mitigation of 20 % balanced voltage dip.

power injected by the converter for voltage dips of 20 % and 50 %. Due to the maximum current rating of the converter, the deeper the dip, the smaller the maximum reactive power output, as can be seen when comparing the reactive power injection in the two presented cases. This also means, that with a higher voltage, e.g. during a voltage swell, a larger reactive power output can be achieved, larger than the nominal converter power, as illustrated in Figure 6.24. As it was already indicated in the previous chapter, the experiments conﬁrm that the typical impedances of a cable network only allow the IN to achieve limited voltage dip or swell mitigation. In the experiments the achieved voltage dip or swell mitigation is around 0.04 p.u. Laboratory-scale demonstration 137

1.2

(p.u.) 0.8 V No Mitigation 0.6 Mitigation 0 0.05 0.1 0.15 0.2 t (s) (a) Voltage amplitude (measurement).

0.5

0 (p.u.)

Q -0.5

-1 0 0.05 0.1 0.15 0.2 t (s) (b) Reactive power injection (measurement). Figure 6.23: Mitigation of 50 % balanced voltage dip.

1.2

(p.u.) 0.8 V No Mitigation 0.6 Mitigation 0 0.05 0.1 0.15 0.2 t (s) (a) Voltage amplitude (measurement).

1 0.5

(p.u.) 0 Q -0.5 -1 0 0.05 0.1 0.15 0.2 t (s) (b) Reactive power injection (measurement). Figure 6.24: Mitigation of 17 % balanced voltage swell.

6.6 Conclusion

In this chapter, the Intelligent Node concept and controls, which were described in the previous chapters, are demonstrated using experiments in a laboratory-scale set-up. After a brief description of the set-up, the 138 Chapter 6 converter response to changes of reference signals and load changes are assessed. The current controller reaches steady state within 30 ms after step-wise changes in the active or reactive power reference values. A step-wise change of the amplitude of the voltage reference signal is tracked within a couple of milliseconds. The connection of a 0.9 p.u. load causes a voltage disturbance of around 0.15 p.u. and the voltage is restored to the reference value within around 60 ms. This voltage disturbance is too small to be considered a voltage dip. A step-wise change in the loading situation of a converter in V control mode causes a disturbance of the DC bus voltage, which is smaller than 1.5 % and lasts shorter than 0.3 s. Before changing from radial to meshed network operation, the IN converter synchronizes the voltages on both sides of the applicable load-break switch. The quality of the synchronization is assessed by measuring the phase angle ’error’ across the load-break switch while the synchronization process is active. The measured phase angle error fluctuates around zero and is always smaller than 10°. The change from radial to meshed network operation and vice versa± is achieved by closing and opening a load-break switch. The three poles of the load- break switch are either operated simultaneously or on a phase-by-phase basis. The experimental results show that in the case of three-phase load-break switch the detection of opening and closing of the load-break switch takes place within a few cycles. Going from radial to meshed network operation is possible without any significant voltage disturbances. For the opposite process it is beneficial that the load in the created radial network area is, before load-break switch opening, supplied by the IN. This allows the opening of the load-break switch without significant voltage disturbances. The phase-by-phase transition to meshed network operation is detected within 100 ms and no significant voltage disturbances are observed. During the phase-by-phase transition from meshed to radial network operation the same voltage disturbance is observed as during connection of a large load, which is a characteristic of the voltage controller. In order to ensure adequate detection of the load-break switch closing, and to prevent overloading in the network, it is essential that a controlled voltage difference across the load-break switch exists before closing. It is described how this aspect is implemented in the demonstration set-up and how this translates to higher voltage networks. Finally, the IN’s capability to mitigate voltage dips and swells is demonstrated. The measurements confirm that the mitigation concept can be implemented successfully. As expected, the amount of mitigation is limited due to the relatively low network impedance, which is characteristic for cable networks. Chapter 7

Conclusions, thesis contribution and recommendations

7.1 Conclusions

Background. Nowadays society is more than ever dependent on energy, and thus demands a high reliability of its energy supply. At the same time environmental concerns stimulate the reduced and more sustainable use of energy. Together with the concern over the increasing scarcity of fossil energy sources and the wish to not be politically dependent for energy needs, this leads to an increased use of renewable energy resources, which are often dispersed by nature. In electrical power systems such distributed generation units inﬂuence the operation of existing protection and voltage regulation systems, power quality, stability and safety. Up to a level of about 20 % of the maximum load it is generally possible to absorb energy from DG units in the electricity distribution network without major costs. The penetration level in many networks is still below this limit, but this will change. Simultaneously with the changing generation mix, in recent years the tasks of generation, transmission, distribution and delivery of electrical energy were unbundled and the roles of network operator, supplier, producer and trader were deﬁned. This has resulted in increasing complexity and uncertainty in the planning and operation of power systems.

Distribution systems. In the current regulatory environment the distribution system network operator is faced with contradictory responsibilities. On one side, the DNO must provide and operate a network of suf- ﬁcient capacity in an economical way. This is generally achieved by loading the network to only a certain percentage of the components’ ratings. This way, in case of maintenance of, or a fault in one of the

139 140 Chapter 7 network components, the network can be reconfigured and supply can be maintained. Thus a reasonable balance is found between network costs and reliability and availability of supply. On the other side, the DNO must provide new connections at very short notice. Anticipating these connections by strengthening the network in advance, brings uncertainty in the return on investments. Fur- thermore, the expected control of load and DG units for balancing supply and demand, and thus reducing generation costs, stresses the distribution system, but is out of the DNO’s sphere of control. This stimulates the DNO to consider other, more flexible, solutions, as an alternative to traditional network reinforcements. The academic and industrial efforts to provide such solutions have resulted in a wide range of possible methods. These methods can be divided into two categories. The first category is based only on the use of communication and automation technology to control the network and the connected load and generation. The second category uses, besides the technologies from the first category, also electrical power equipment with the ability to influence the power flow. Distribution systems already have some flexibility to respond to changes in the power flow situation. The most widely deployed technique is the use of on- and off-load tap changers. More advanced methods include the use of power electronic devices, which are, when applied to transmission systems, also called FACTS devices (Flexible AC Transmis- sion Systems). When applied to distribution systems, these devices are referred to as D-FACTS devices. The Intelligent Node concept, which is presented in this thesis, is a D-FACTS device. The Intelligent Node can influence some power quality parameters, such as flicker, voltage dips and phase angle jumps. Currently, for many of these parameters no binding compatibility levels exist, but steps are made towards this. For each of the relevant PQ indicators the limits, which are used in this thesis to assess the influence of the IN, are defined. The power quality parameters steady state voltage amplitude and power frequency are discussed as they are input for the development and implementation of the IN concept.

FACTS in distribution systems. FACTS and distribution system FACTS (D-FACTS) devices are used to influence voltage and loading levels in the electrical network that they are part of. The effect that such devices can achieve is significantly different for cable networks when compared to overhead line networks. In an overhead line network reactive power injection leads to an increased voltage, while in a cable network it (also) causes a phase angle difference. Similarly, a series voltage source that inserts a quadrature voltage causes a circulating active power flow Conclusions, thesis contribution and recommendations 141 in a meshed overhead line network, while in a meshed cable network, it (also) causes a circulating reactive power flow. The application of solid-state switches is the enabling technology for FACTS and D-FACTS devices and brings the economic application of such devices in distribution networks within reach. FACTS and D-FACTS devices exist in a wide variety of topologies and functions. Of the existing topologies, the UPFC, IPFC and the back-to-back converter can effectively control active power flow in cable networks. This thesis concerns the extension of the back-to-back topology and of its operational concept.

Functional concept of the IN. The IN concept concerns a device that can couple and decouple distribution network areas and control the power flow between these areas. This way the IN adds flexibility to the network. The IN can be applied to facilitate increased loading of a network by controlled sharing of redundancy or by the controlled power exchange between grid areas. Also the integration of distributed generation can be enabled by controlling voltage profiles. Finally, the IN can improve power quality levels, for example by mitigating voltage dips. It is concluded that facilitating increased loading is the most important IN application and that the control of voltage profiles and voltage dip mitigation only offer limited benefits when compared with alterative solutions. From the mentioned applications the following functional requirements were formulated: • Inject or consume an adjustable amount of active and/or reactive power through each of its AC ports when connected in a meshed network. • Supply a radial network from any of its AC ports. • Detect a permanent short-circuit and de-energize the appropriate feeder. • Ride through a voltage dip, which occurs due to a permanent fault on a nearby feeder. • Detect the opening of the load-break switches which isolate a network part and change from controlling power flow in a meshed network to supplying a radial network part. • Synchronize the voltages on both sides of a remote opened load- break switch. • Detect the closing of the load-break switch which restores meshed network operation and change from supplying a radial network section to controlling power flow in a meshed network. 142 Chapter 7

• Support the above also for the phase-by-phase operation of the load-break switches.

• Optionally: improve the power quality of the connected networks.

• Optionally: store energy.

To provide these abilities, the IN can have several internal topologies. If a certain IN application only uses a subset of the mentioned requirements, these can, for example, be realized by using a number of power electronics controlled auto transformers and PE-controlled series impedances. In this thesis, to provide all abilities, the versatile topology of multiple back-to-back converters is used.

IN control and protection. The IN must be able to perform its tasks during normal and steady state operation and during planned and unplanned power system events, without being damaged or damaging power system components. The control system which is proposed to achieve this has a central current controller, which is always active and inherently provides current limitation capabilities. In normal operation, each of the IN converters is either connected to a radial network section or forms part of a meshed network. To supply a radial network, an outer control loop controls the voltage on this network. To control power flow in a meshed network, the injected current is imposed directly. To handle short-circuits on connected feeders and to detect unintentional network disconnections that result in a radial network section, a protection system was developed. This protection philosophy takes into account the current limiting capabilities of the IN. In radial network operation, this means that short-circuits are recognized by detecting under-voltage instead of over-current. Also for other system events the specific controller responses are used to detect changing conditions, such as, for example, the unintentional creation of a radial network. To mitigate voltage dips and swells a controller was implemented that controls the IN reactive power output depending on the network voltage. This controller only operates outside the steady state voltage band, and does not affect the power flow control in normal power system conditions. The planned power system events consist of the energization and de- energization of the IN and the disconnection and (re)connection of grid areas using the IN. To support the disconnection and (re)connection of grid areas an algorithm was developed to detect the opening or closing of a remote load-break switch by observing locally measured signals. After detection of the load-break switch operation the applicable converter Conclusions, thesis contribution and recommendations 143 control structure is changed. The algorithms support both the phase- by-phase operation of load-break switches and the simultaneous opening and closing of all three phase contacts. For the connection of grid areas an important characteristic is the synchronization before load-break switch closing. Since remote voltage amplitude, frequency and phase angle information is assumed to be only periodically available, the level of frequency variations in the ’central power grid’ is important. By evaluating frequency measurements in the public low voltage network, the maximum synchronization interval was determined as a function of allowable phase angle jump.

Laboratory-scale demonstration. The Intelligent Node controls were implemented in a low voltage laboratory-scale set-up consisting of two back-to-back connected converters of 5 kVA each, with a rotating DC generator connected to the DC bus. Connections to the public low voltage grid and to a programmable voltage source are used. Further resistive loads can be connected to the converters. The performed experiments focus on verification of the proposed IN control and detection methods and not on optimization processes on distribution system level. The experiments have confirmed that the control and detection methods allow the connection and disconnection of network areas without interrupting supply, and with an acceptable power quality level. In order to ensure adequate detection of the load-break switch closing, and to prevent overloading in the network, it is ensured that a controlled voltage difference across the load-break switch exists before closing it. It is described how this aspect is implemented in the demonstration set-up and how this translates to higher voltage networks. Finally, the IN’s capability to mitigate voltage dips and swells was demonstrated. The experiments confirm that the mitigation concept can be implemented successfully. As expected, the amount of mitigation is limited due to the relatively low network impedance, which is characteristic for cable networks.

7.2 Thesis contribution

The scientiﬁc contribution of this thesis to the current state of the art of D-FACTS applications in distribution systems is in the following topics.

Definition of the operational benefits of multi back-to-back converter topology. The multi back-to-back converter topology is an extension of the existing back-to-back topology. This thesis defines the additional power system benefits that can be achieved when more than 144 Chapter 7 two converters are connected on their DC sides. By sharing redundancy, circuit reinforcements can be prevented or deferred. If the IN is prepared as a relocatable installation, it can enable the DNO to fulfill its duties of maintaining enough spare capacity and connecting clients in a timely fashion.

Enabling uninterrupted supply during the transitions between radial and meshed network operation. Existing back-to-back applications allow both the control of power ﬂow in meshed operation and the supply of a radial network. However, the transitions between these two modes require supply interruptions. In this thesis an additional functionality is proposed that prevents supply interruption during the mentioned transition. The proposed method is experimentally veriﬁed.

Determination of maximum measurement interval to ensure adequate network synchronization. Synchronizing a radial network area with a remote network area requires information about the remote power frequency, phase angle and voltage amplitude. The power frequency of the public electricity network varies due to the continuous process of balancing supply and demand. In this thesis, the maximum time is determined that can be allowed between two receptions of remote information, as a function of acceptable phase angle diﬀerence between the networks. This knowledge is an eye-opener and opens the possibility to use non-realtime communication methods for synchronization.

7.3 Recommendations

Scaling effects. Back-to-back applications in medium and high voltage networks already exist, so only those aspects that are exclusive to the IN concept require practical verification. In this thesis this verification was performed using a low voltage laboratory-scale model. A theoretical analysis was made of the scale-effect that occurs for applications at higher voltage levels. Before application in public medium voltage networks, further experimental verification is needed in a controlled medium voltage laboratory environment. The focus of such experiments must be on power quality effects and quality of detection.

Storage. In this thesis, the IN concept was developed without including energy storage. Connecting energy storage to the DC bus of the IN would, as a first benefit, increase the flexibility of the IN to optimize the power flow control, since then the instantaneous power balance among the various AC ports of the IN would no longer be required. As a second benefit, energy storage would allow the IN to supply (part of) the fault Conclusions, thesis contribution and recommendations 145 current in case of a network short-circuit. To supply the fault current in a cable network involves mainly active power, as apposed to overhead line networks where the fault current requires mainly reactive power. For the same reason, the use of active power is expected to improve the voltage dip mitigation capabilities of the IN. It is recommended to further quantify the benefits and required size of storage when applied in the IN concept.

Economical analysis. The scope of this thesis is limited to the technical and operational aspects of the Intelligent Node. To support the development of the business case of the IN concept, also the financial costs and benefits of the IN concept need to be quantified, so that it can be compared with, for example, traditional network reinforcements.

Eﬀect of inability to supply short over-current peaks. With the use of power electronics converters to supply a radial network, the parameter ’short-circuit power’ no longer adequately describes the ’strength’ of the network. A radial network sees a relatively ’hard’ voltage source up to the nominal converter current. Beyond that point, the output current of the converter is limited. Some connected equipment, such as, for example, large motors, may be impacted by this. Induc- tion machines are characterized by a large inrush current when starting. If this inrush current causes the converter to reach its current threshold, the network voltage is reduced. The impact of this eﬀect on such equipment itself and on other connected equipment is a topic that needs further research.

Protection systems for converter based networks. The same current limitation requires a rethinking of power system protection concepts. In this thesis, the basic ideas to develop such a new concept have been discussed, but the actual implementation and interaction with other protection systems require further investigation.

References

[1] The Economist, “Special report on the future of energy,” June 21-27, 2008.

[2] ——, “Special report on climate change and the carbon economy,” December 5-11, 2009.

[3] United Nations, “Kyoto protocol to the United Nations framework convention on climate change,” 1998.

[4] “United Nations framework convention on climate change in Copenhagen, Denmark, COP 15/CMP 5,” 7 to 18 December 2009. [Online]. Available: http://unfccc.int

[5] European Commission, “Energy for the future: Renewable sources of energy; White paper for a community strategy and action plan,” COM(97)599, 1997.

[6] Commission of the European Communities, “20 20 by 2020 Eu- rope’s climate change opportunity,” COM(2008), 2008.

[7] European Parliament and Council, “Directive 2001/77/EC of 27 September 2001 on The promotion of electricity produced from renewable energy sources in the internal electricity market.”

[8] CIGRE WG C6.01, “Development of dispersed generation and consequences for power systems,” Electra, vol. 215, pp. 39–49, 2007.

[9] C. Materazzi-Wagner, “Long term perspectives for DG in Europe,” DISPOWER D7.1, 2005.

[10] Commission of the European Communities, “An energy policy for Europe,” COM(2007), 2007.

[11] “Elektriciteitswet 1998 (Electricity Act),” Dutch law, 2009. [Online]. Available: http://wetten.overheid.nl/BWBR0009755

147 148 References

[12] J. A. Pe¸cas Lopes, N. Hatziargyriou, J. Mutale, P. Djapic, and N. Jenkins, “Integrating distributed generation into electric power systems: A review of drivers, challenges and opportunities,” Elec- tric Power Systems Research, vol. 77, no. 9, pp. 1189–1203, 2007.

[13] P. K. Sen and S. L. Larson, “Fundamental concepts of regulating distribution system voltages,” in Proc. Conf. Rural Electric Power, 1994, pp. C1.1 – C1.10.

[14] G. Celli, F. Pilo, G. Pisano, V. Allegranza, R. Cicoria, and A. Iaria, “Meshed vs. radial MV distribution network in presence of large amount of DG,” in Proc. IEEE Power Systems Conf. and Exp., 2004, pp. 709–714.

[15] R. Cossent, P. G´omez, and P. Fr´ıas, “Towards a future with large penetration of distributed generation: Is the current regulation of electricity distribution ready? Regulatory recommendations under a European perspective,” Energy Policy, vol. 37, pp. 1145–1155, 2009.

[16] E. Coster and D. van Houwelingen, “Integration of DG in MV- grids: challenges encountered by the grid operator,” in Proc. CIRED, 2009.

[17] “FENIX, Flexible electricity network to integrate the expected energy revolution.” [Online]. Available: http://www.fenix-project.org

[18] “Microgrids and MoreMicrogrids.” [Online]. Available: http://www.microgrids.eu

[19] R. H. Lasseter and P. Paigi, “Microgrid: a conceptual solution,” in Proc. IEEE Power Electronics Specialists Conf., vol. 6, 2004, pp. 4285–4290.

[20] “EU-DEEP, The birth of a European distributed energy partnership that will help the large-scale implementation of distributed energy resources in Europe.” [Online]. Available: http://www.eu-deep.com

[21] Y. Sun, M. H. J. Bollen, and G. Ault, “Improving distribution system reliability by means of distributed generation,” in Proc. CIRED, 2007.

[22] T. Oyama, “Demand side prospects of multiple menu service in power system with quality control center,” in Proc. IEEE PES Winter Meeting, 1999. References 149

[23] M. Ako, R. Hara, H. Kita, and J. Hasegawa, “A method for balancing supply and demand in an isolated system consisting of voltage control type QCCS in FRIENDS,” in Proc. Power Systems Com- putation Conf., 2008.

[24] “DISPOWER, Distributed generation with high penetration of renewable energy sources.” [Online]. Available: http://www.dispower.org

[25] “DGFACTS, Improvement of the quality of supply in distributed generation networks through the integrated application of power electronic techniques.” [Online]. Available: http://dgfacts.labein.es/dgfacts

[26] “IRED, Integration of renewable energy sources and distributed generation into the European electricity grid.” [Online]. Available: http://www.ired-cluster.org

[27] “SmartGrids, European Technology Platform for the Elec- tricity Networks of the Future.” [Online]. Available: http://www.smartgrids.eu

[28] “GRIDWISE.” [Online]. Available: http://gridwise.pnl.gov

[29] “IntelliGrid.” [Online]. Available: http://intelligrid.epri.com

[30] “Gridstat.” [Online]. Available: http://www.gridstat.net

[31] “U.S. Department of Energy, Oﬃce of Electric- ity Delivery & Energy Reliability.” [Online]. Available: http://www.electricdistribution.ctc.com

[32] F. van Overbeeke, “Active networks: Distribution networks facilitating integration of distributed generation,” in Proc. 2nd Int. Symp. Distributed Generation: Power System and Market Aspects, 2002.

[33] The Economist, “Wiser wires,” October 10-16, 2009.

[34] K. E. Martin and J. R. Carroll, “Phasing in the technology,” IEEE Power & Energy Magazine, vol. 6, no. 5, pp. 24–33, 2008.

[35] “IEC61850:2009 Communication networks and systems in substa- tions.”

[36] F. Caleno, “The Enel Smart Info - A ﬁrst Smart Grids step to addressing in-home energy eﬃciency,” in Proc. CIRED, 2009. 150 References

[37] “ADDRESS Active distribution networks with full integration of demand and distributed energy resources.” [Online]. Available: http://www.addressfp7.org

[38] J. A. Pe¸cas Lopes, P. M. Rocha Almeida, and F. J. Soares, “Using vehicle-to-grid to maximize the integration of intermittent renewable energy resources in islanded electric grids,” in Proc. Int. Conf. Clean Electrical Power, 2009.

[39] G. Celli, E. Ghiani, S. Mocci, F. Pilo, and E. Pazzola, “Demand side management as a support to distributed generation in active networks,” in Proc. CIRED, 2005.

[40] V. Stanojevi´c,V. Silva, D. Pudjianto, G. Strbac, P. Lang, and D. MacLeman, “Application of storage and demand side management to optimise existing network capacity,” in Proc. CIRED, 2009.

[41] M. Gillie, A. Beddoes, M. Gosden, and M. Orme, “An evaluation of alternative voltage control methods enabling increased distributed generation connection capacity,” in Proc. CIRED, 2007.

[42] R. O’Gorman and M. A. Redfern, “Taking advantage of distributed generation for voltage control,” in Proc. CIRED, 2007.

[43] Y. Xu, D. T. Rizy, F. Li, and J. D. Kueck, “Dynamic voltage regulation using distributed energy resources,” in Proc. CIRED, 2007.

[44] T. Tran-Quoc, E. Monnot, G. Rami, A. Almeida, C. Kieny, and N. Hadjsaid, “Intelligent voltage control in distribution network with distributed generation,” in Proc. CIRED, 2007.

[45] M. J. Dolan, E. M. Davidson, G. W. Ault, and J. R. McDon- ald, “Techniques for managing power ﬂows in active distribution networks within thermal constraints,” in Proc. CIRED, 2009.

[46] F. Katiraei and M. R. Iravani, “Power management strategies for a microgrid with multiple distributed generation units,” IEEE Tr. Power Systems, vol. 21, no. 4, pp. 1821–1831, 2006.

[47] A. Nourai and C. Schafer, “Changing the electricity game,” IEEE Power & Energy Magazine, vol. 7, no. 4, pp. 42–47, 2009.

[48] B. Roberts, “Capturing grid power,” IEEE Power & Energy Mag- azine, vol. 7, no. 4, pp. 32–41, 2009. References 151

[49] J. Wang, A. Q. Huang, W. Sung, Y. Liu, and B. J. Baliga, “Smart grid technologies - development of 15kV SiC IGBTs and their impact on utility applications,” IEEE Industrial Electronics Maga- zine, vol. 3, no. 2, pp. 16–23, 2009.

[50] J. Rodriguez, S. Bernet, B. Wu, J. O. Pontt, and S. Kouro, “Mul- tilevel voltage-source-converter topologies for industrial medium- voltage drives,” IEEE Tr. Industrial Electronics, vol. 54, no. 6, pp. 2930–2945, 2007.

[51] L. M. Tolbert, T. J. King, B. Ozpineci, J. B. Campbell, G. Mu- ralidharan, D. T. Rizy, A. S. Sabau, H. Zhang, W. Zhang, Y. Xu, H. F. Huq, and H. Liu, “Power electronics for distributed energy systems and transmission and distribution applications,” Oak Ridge National Laboratory, ORNL/TM-2005/230, 2005.

[52] K. Eriksson, M. Piekutowski, and V. Sood, “HVDC and FACTS for distribution systems,” Electra, vol. 222, pp. 43–55, 2005.

[53] D. O. Neacsu, Power switching converters - medium and high power. CRC Press, 2006.

[54] European Technology Platform SmartGrids, “Strategic deployment document for Europe‘s electricity networks of the future,” September 2008.

[55] C. Sschwaegerl, L. Tao, P. Mancarella, and G. Strbac, “Can microgrids provide a new paradigm for network operation? An evaluation of their technical, commercial and environmental beneﬁts,” in Proc. CIRED, 2009.

[56] J. A. Pe¸cas Lopes et al., “MICROGRIDS - Large scale integration of microgeneration to low voltage grids - Emergency strategies and algorithms,” Deliverable DD1, 2004.

[57] D. Menniti, A. Pinnarelli, and N. Sorrentino, “Operation of decentralized electricity markets in microgrids,” in Proc. CIRED, 2009.

[58] M. Reza, P. H. Schavemaker, W. L. Kling, and L. van der Sluis, “A research program on intelligent power systems: self controlling and self-adapting power systems equipped to deal with the structural changes in the generation and the way of consumption,” in Proc. CIRED, 2003.

[59] P. Kundur, Power system stability and control. McGraw-Hill, 1994. 152 References

[60] European Parliament and Council, “Directive 2009/72/EC of the European Parliament and of the Council of 13 July 2009 concerning common rules for the internal market in electricity and re- pealing Directive 2003/54/EC,” Oﬃcial Journal of the European Union, vol. L211, pp. 55–93, 2009.

[61] T. Short, Electric power distribution handbook. CRC Press, 2004.

[62] H. L. Willis, Power distribution planning reference book, Second Edition. Marcel Dekker, 2004.

[63] EnergieNed, “Energie in Nederland 2009 - Energy in the Nether- lands 2009,” 2009.

[64] “EN50160:2007 Voltage characteristics of electricity supplied by public distribution networks.”

[65] ERGEG, “Towards voltage quality regulation in Europe - An ERGEG public consultation paper,” E06-EQS-09-03, 2006.

[66] ——, “Towards voltage quality regulation in Europe - An ERGEG conclusions paper,” E07-EQS-15-03, 2007.

[67] ——, “ERGEG Public consultation towards voltage quality regulation in Europe - Evaluation of the comments received,” E07- EQS-15-04, 2007.

[68] Norges vassdrags- og energidirektorat, (Olje- og energide- partementet), “Forskrift om leveringskvalitet i kraftsystemet, FOR-2004-11-30 nr. 1557 (in Norwegian),” November 2004. [Online]. Available: http://www.lovdata.no

[69] “IEC61000-4-15:2003 Electromagnetic compatibility (EMC) - Testing and measurement techniques - Flickermeter - Functional and design speciﬁcations.”

[70] European Commision, “Commission regulation (EC) no 244/2009 of 18 march 2009 implementing Directive 2005/32/EC of the Eu- ropean Parliament and of the Council with regard to ecodesign requirements for non-directional household lamps,” Oﬃcial Jour- nal of the European Union, vol. L76, pp. 3–16, 2009.

[71] C. Rong, “Flicker interaction studies and ﬂickermeter improvement,” Ph.D. dissertation, Faculty of Electrical Engineering, Eind- hoven University of Technology, the Netherlands, 2009. References 153

[72] “IEC61000-3-3:2008 Electromagnetic compatibility (EMC) - Lim- its - Limitation of voltage changes, voltage ﬂuctuations and ﬂicker in public low-voltage supply systems, for equipment with rated current 16 A per phase and not subject to conditional connection.” ≤

[73] “IEC/TR61000-2-8:2002 Electromagnetic compatibility (EMC) - Environment - Voltage dips and short interruptions on public electric power supply systems with statistical measurement results.”

[74] Information Technology Industry Council, “ITI CBEMA curve,” 2000. [Online]. Available: http://www.itic.org/resources/iti-cbema-curve

[75] M. Bollen and I. Gu, Signal processing of power quality disturbances. John Wiley & Sons, 2006.

[76] R. Fuchs, “Protection schemes for decentralized power generation,” in Proc. Int. Conf. Developments in Power System Pro- tection, 2004.

[77] Direçcão-Geral de Energia, Ministério da Indústria e Energia, Lis- bon, Portugal, “Guia técnico das instala¸cões de produ¸cão indepen- dente de energia eléctrica (in Portuguese),” 1994.

[78] R. Luijten, “Coupling MV networks (in Dutch),” Master’s thesis, Faculty of Electrical Engineering, Eindhoven University of Technology, 2008. [Online]. Available: http://www.narcis.info

[79] A. L. J. Janssen, R. P. P. Smeets, W. A. van der Linden, and M. J. M. van Riet, “Distributed generation in relation to phase opposition and short-circuits,” in Proc. Int. Conf. Short-Circuit Currents in Power Systems, 2002, pp. 55–62.

[80] ENTSO-E, “Policy 1, Load frequency control and performance, v3.0c,” in Operating Handbook, 2009. [Online]. Available: http://www.entsoe.eu

[81] “IEC61000-2-2:2002 Electromagnetic compatibility (EMC) - Envi- ronment - Compatibility levels for low-frequency conducted disturbances and signalling in public low-voltage power supply systems.”

[82] K. Anna, M. Kari, R. Sami, and J. Pertti, “Active voltage level management of distribution networks with distributed generation using on load tap changing transformers,” in Proc. PowerTech, 2007. 154 References

[83] F. Provoost, “Intelligent distribution network design,” Ph.D. dissertation, Faculty of Electrical Engineering, Eindhoven University of Technology, the Netherlands, 2009. [Online]. Available: http://www.narcis.info [84] E. Acha, V. Agelidis, O. Anaya-Lara, and T. Miller, Power electronic control in electrical systems. Newnes, 2002. [85] M. H. Rashid, Power electronics handbook. San Diego, CA, USA: Academic Press, 2001. [86] N. Mohan, T. M. Undeland, and W. P. Robbins, Power electronics - Converters, applications and design. John Wiley and Sons, 1995. [87] M. Rahimo and S. Klaka, “High voltage semiconductor technologies,” in Proc. 13th European Conf. Power Electronics and Appli- cations EPE, 2009. [88] P. Kollensperger, M. Bragard, T. Plum, and R. W. De Doncker, “The dual GCT - A new high-power device using optimized GCT technology,” in Proc. IEEE Industry Applications Conf., 2007, pp. 358–365. [89] B. K. Bose, Modern power electronics and AC drives. Prentice Hall, 2002. [90] C. Meyer and R. W. De Doncker, “Power electronics for modern medium-voltage distribution systems,” in Proc. Int. Power Elec- tronics and Motion Control Conf., vol. 1, 2004, pp. 58–66. [91] R. Sastry Edam and M. S. Sarma, Power quality - VAR compensation in power systems. CRC Press, 2009. [92] ABB, “Eagle Pass - Piedras Negras BtB Light,” Project reference brochure, 2000. [93] J. Harlow, Electric power transformer engineering. CRC Press, 2004. [94] IEEE P1409 Distribution Custom Power Task Force, “Custom power technology development list,” 2000. [Online]. Available: http://grouper.ieee.org/groups/1409 [95] N. G. Hingorani and L. I. Gyugy, Understanding FACTS: Concepts and technology of ﬂexible AC transmission systems. Wiley-IEEE Press, 1999. [96] R. W. Erickson and D. Maksimovic, Fundamentals of power electronics, 2nd ed. New York, NY, USA: Kluwer, 2004. References 155

[97] P. A. Laplante, Electrical engineering dictionary. Boca Raton, FL, USA: CRC Press, 2000.

[98] CIGRE WG 38.01, Task Force No. 2 on SVC, “Static var compensators,” 1986.

[99] P. Dusan, “Use of HVDC and FACTS,” Proc. of the IEEE, vol. 88, no. 2, pp. 235–245, 2000.

[100] ABB, “SVC Light - A breakthrough in power quality,” Project reference brochure, 2000.

[101] S. M. Ramsay, P. E. Cronin, R. J. Nelson, J. Bian, and F. E. Menendez, “Using distribution static compensators (D- STATCOMs) to extend the capability of voltage-limited distribution feeders,” in Proc. Rural Electric Power Conf., 1996, pp. A4/18–A4/24.

[102] E. le Pelleter, S. Bacha, R. Caire, and R. Guiraud, “Active and reactive power ﬂow control based on D-SSSC for looped and meshed distribution grids,” in Proc. CIRED, 2007.

[103] M. Roggen, “Slimme transformator: SmartTrafo verbetert kwaliteit en spaart kosten (in Dutch),” Energietechniek, vol. 5, 2004.

[104] G. G. Karady and P. Parihar, “Integrated PWM and transformer switching technique for AC voltage regulation,” in Proc. IEEE Applied Power Electronics Conf. and Exp., 1994, pp. 961–967.

[105] J. W. Schwartzenberg and R. W. De Doncker, “15kV medium voltage static transfer switch,” in Proc. IEEE IAS Industry Appli- cations Conf., Thirtieth IAS Annual Meeting, 1995.

[106] T. Ise, M. Takami, and K. Tsuji, “Hybrid transfer switch with fault current limiting function,” in Proc. Int. Conf. Harmonics and Quality of Power, vol. 1, 2000.

[107] S. G. Johansson, L. Carlsson, and G. Russberg, “Explore the power of HVDC Light - A web based system interaction tuto- rial,” in Proc. IEEE PES Power Systems Conf. and Exp., vol. 2. IEEE, 2004, pp. 839–842.

[108] ABB, “HVDC & SVC Light - Reference list,” Project reference brochure, 2005. [109] M. Weinhold, H. Werner, and M. Erath, “Power under control with SIPLINK,” in Proc. CEPSI, 2000, pp. 1–8. 156 References

[110] T. Larsson, A. Petersson, A. Edris, D. Kidd, R. Haley, and F. Aboytes, “Eagle Pass back-to-back tie: A dual purpose application of voltage source converter technology,” in Proc. IEEE PES Summer Meeting, 2001.

[111] SDG&E, “Talega STATCOM/BTB FACTS project,” Project reference brochure, 2003.

[112] Siemens, “Solution for utilities: city network Ulm ”SIPLINK for peak load exchange”,” Project reference brochure, Germany, 2003.

[113] J. Flottemesch and M. Rother, “Optimized energy exchange in primary distribution networks with DC links,” in Proc. IEEE Int. Conf. Electricity Deregulation, Restructuring and Power Technolo- gies, 2004.

[114] N. Okada, H. Sakai, and S. Katoh, “Development of a 6.6kV - 1MVA transformerless loop balance controller,” in Proc. IEEE Power Electronics Specialists Conf., 2007, pp. 1087–1090.

[115] N. Okada, H. Kobayashi, T. Ishikawa, K. Takigawa, and K. Kurokawa, “Simulation of isolated operation in fault condition by loop power ﬂow controller,” Control Engineering Practice, vol. 13, no. 12, pp. 1637–1543, 2005.

[116] G. C. Schoonenberg, A. R. A. Pikkert, and P. Voulon, “Upgrading of compact load break switches to a circuit breaker function,” in Proc. CIRED, 2001.

[117] EATON HOLEC, “Magneﬁx 3.6 - 15kV insulation enclosed switchgear - product focus,” Product Brochure, 2008.

[118] A. von Meier, Electric power systems: a conceptual introduction. IEEE Press, 2006.

[119] E. Kreyszig, “Chapter 21: Unconstrained optimization, linear pro- gramming,” in Advanced engineering mathematics, 7th Ed. John Wiley and Sons Inc., 1993.

[120] N. Okada, “A method to determine the distributed control setting of looping devices for active distribution systems,” in Proc. PowerTech, 2009.

[121] M. P. Kazmierkowski and L. Malesani, “Current control techniques for three-phase voltage-source PWM converters: A survey,” IEEE Tr. Industrial Electronics, vol. 45, no. 5, pp. 691–703, 1998. References 157

[122] C. Houpis and G. Lamont, Digital control systems. McGraw-Hill, 1992.

[123] G. Franklin, J. Powell, and M. Workman, Digital control of dynamic systems. Addison-Wesly, 1998.

[124] S. Buso and P. Mattavelli, Digital control in power electronics. Morgan & Claypool, 2006.

[125] A. Lopez de Heredia, H. Gaztanaga, I. Etxeberria-Otadui, S. Bacha, and X. Guillaud, “Analysis of multi-resonant current control structures and tuning methods,” in Proc. Conf. IEEE In- dustrial Electronics Soc. IECON, 2006, pp. 2156–2161.

[126] D. N. Zmood and D. G. Holmes, “Stationary frame current regulation of PWM inverters with zero steady-state error,” IEEE Tr. Power Electronics, vol. 18, no. 3, pp. 814–822, 2003.

[127] M. P. Kazmierkowski, R. Krishnan, and F. Blaaberg, Control in power electronics - Selected problems. Academic Press, 2002.

[128] H. Akagi, E. H. Watanabe, and M. Aredes, Instantaneous power theory and applications to power conditioning. IEEE Press, 2007.

[129] F. Wang, J. L. Duarte, and M. Hendrix, “Control of grid- interfacing inverters with integrated voltage unbalance correc- tion,” in Proc. IEEE Power Electronics Specialists Conf., 2008, pp. 310–316.

[130] “IEC61000-4-30:2008 Electromagnetic compatibility (EMC) - Testing and measurement techniques - Power quality measurement methods.”

[131] F. Provoost, J. M. A. Myrzik, and J. F. G. Cobben, “Advantages of cables regarding network performance,” in Proc. IEEE Young Researchers Symp., 2004.

[132] F. Wang, J. L. Duarte, and M. Hendrix, “Active and reactive power control schemes for distributed generation systems under voltage dips,” in Proc. IEEE Energy Conversion Congress and Exp., 2009, pp. 3564–3571.

[133] F. van Overbeeke, “Fault current source to ensure the fault level in inverter-dominated networks,” in Proc. CIRED, 2009.

[134] A. Oudalov and A. Fidigatti, “Adaptive network protection in microgrids,” International Journal of DER, vol. 5, no. 3, pp. 1– 24, 2009. 158 References

[135] M. C. Benhabib and S. Saadate, “A new robust experimentally validated phase locked loop for power electronic control,” EPE Journal, vol. 15, no. 3, pp. 36–48, 2005.

[136] M. Ciobotaru, R. Teodorescu, and V. G. Elidis, “Oﬀset rejection for PLL based synchronization in grid-connected converters,” in Proc. IEEE Applied Power Electronics Conf. and Exp., 2008, pp. 1611–1617.

[137] F. Wang, M. C. Benhabib, J. L. Duarte, and M. A. M. Hen- drix, “High performance stationary frame ﬁlters for symmetrical sequences or harmonics separation under a variety of grid conditions,” in Proc. IEEE Applied Power Electronics Conf. and Exp., 2009, pp. 1570–1576. [138] M. Bollen and O. Samuelsson, “Controlled island operation of part of the 50 kV grid in southern Sweden,” in Proc. PowerTech, 2009.

[139] G. Iwanski and W. Koczara, “Synchronization and mains outage detection for controlled grid connection of the wind driven vari- able speed power generation system,” in Proc. Int. Conf. Clean Electrical Power, 2007. [140] J. M. A. Nuijten, M. J. M. van Riet, A. Geschiere, F. L. Baldinger, and W. M. van Buijtenen, “Coupling of networks with a new system for predicting dynamic equalizing currents,” in Proc. CIGRE, 2006. Abbreviations, symbols and notations

Abbreviation Meaning AC alternating current B2B back-to-back CHP combined heat and power DC direct current DG distributed generation DNO distribution network operator DSM demand side management DSP digital signal processor D-FACTS distribution system FACTS D-SSSC distributed static synchronous series compensator DVR dynamic voltage restorer EMC electromagnetic compatibility EMVT elektromagnetische vermogenstechniek (electromagnetic power technology) ENTSO-E European Network of Transmission System Operators for Electricity ETP European Technology Platform EU European Union FACTS ﬂexible AC transmission systems FPGL Flexible Power Grid Lab GPS global positioning system GTO gate turn-oﬀ thyristor HTS hybrid transfer switch HV high voltage HVDC high voltage direct current IGBT insulated gate bipolar transistor IGCT integrated gate commutated thyristor IN intelligent node Continued on next page 159 160 Abbreviations, symbols and notations

Abbreviation Meaning IOP innovatiegerichte onderzoeksprogramma’s (innovation oriented research programmes) IPFC inter-line power flow controller ITIC Information Technology Industry Council LB load-break switch LDC line drop compensation LV low voltage MCT MOS controlled thyristor MOS metal oxide semiconductor MOSFET MOS field-effect transistor MV medium voltage OHL overhead line OLTC on-load tap changer OSG orthogonal system generator P proportional PE power electronic PI proportional integral PLL phase locked loop PMU phasor measurement unit PR proportional resonant PQ power quality PST phase shifting transformer PV photovoltaic PWM pulse width modulation QB quadrature booster (transformer) RES renewable energy source r.m.s. root mean square SiC silicon carbide SSLTC solid-state load tap-changer SSSC static synchronous series compensator STATCOM static synchronous compensator STATCON static condensor STS static transfer switch SVC static var compensator TCSC thyristor controlled series capacitor TSO transmission system operator UCTE Union for the Co-ordination of Transmission of Electricity UPFC unified power flow controller Abbreviations, symbols and notations 161

Symbol Meaning δ angle between sending and receiving end voltages (rad) φ angle between active and apparent power vectors (rad) ω frequency (rad/s) C capacitance (F) f frequency (Hz) i time domain current (A or p.u.) I current amplitude (A or p.u.) I complex current (A or p.u.) Ki integral gain Kp proportional gain L inductance (H) P active power (W or p.u.) Pst short-term ﬂicker Q reactive power (var or p.u.) R resistance (Ω) S complex apparent power (VA or p.u.) Sk short circuit power (VA) v time domain voltage (V or p.u.) V voltage amplitude (V or p.u.) V complex voltage (V or p.u.) x distance, length (m) Z impedance (Ω)

Notation Meaning α,β α or β axis of stationary αβ reference frame a,b,c phase a, b or c ac alternating current cc current controller dc direct current rms root mean square vac ac voltage controller 0 zero sequence − negative sequence + positive sequence ∗ reference value ∗ complex conjugate

Appendix A

DC current of AC/DC converter

To assess the impact of active and reactive power exchange on the AC side of the converter on the DC bus voltage, an analysis is made of the current on the DC side due to current on the AC side. First, the analysis is made for a single-phase converter, followed by the same analysis for a three-phase converter, in which both the balanced and the unbalanced situation are treated. In the analysis it is assumed that the DC bus voltage vdc(t) of the DC bus is constant and equal to Vdc, and that the switching frequency of the converter is high enough and the converter control good enough so that the currents and voltages on the AC side are sinusoidal.

A.1 DC link current in single phase voltage source converter

Since the converter switching elements themselves do not have any energy storage components, there is an instantaneous power and energy balance between the AC and DC side:

vdc(t)idc(t) = v1(t)i1(t) (A.1) which results, for a sinusoidal voltage and a load with power factor cos(φ), in a DC current equal to 1 i (t) = √2V sin(ωt)√2I sin(ωt φ) dc V 1 1 − dc (A.2) V I = 1 1 (cos(φ) cos(2ωt φ)) Vdc − − The DC current consists of a constant part, corresponding to the amount of active power taken on the AC side, and a pulsating component with a frequency that is twice the AC system power frequency.

163 164 Appendix A

A.2 DC link current in three-phase voltage source converter

We repeat the same analysis for a three-phase converter, for which with the instantaneous power balance expressed as

vdc(t)idc(t) = v1(t)i1(t) + v2(t)i2(t) + v3(t)i3(t) (A.3)

In the following paragraphs Equation (A.3) is evaluated for the AC current consisting of positive, negative and zero sequence components.

Positive sequence current In case of a balanced three-phase current, only a positive sequence current exists on the AC side of the converter, and the DC current is written as

√2V1 sin(ωt)√2I1 sin(ωt φ) 1 ◦ − ◦ idc(t) = +√2V2 sin(ωt 120 )√2I2 sin(ωt 120 φ) Vdc  − ◦ − ◦ −  +√2V3 sin(ωt 240 )√2I3 sin(ωt 240 φ)  − − −  V1I1(cos(φ) cos(2ωt φ)) 1 − − ◦ = +V2I2(cos(φ) cos(2(ωt 120 ) φ)) V ◦ dc  +V I (cos(φ) − cos(2(ωt − 240 ) − φ))  3 3 − − −  3 cos(φ) cos(2ωt φ)  V I+ = ac − cos(2ωt − 240◦ φ) V ◦ dc  − cos(2ωt − 120 − φ)  − − − V I+   = ac 3 cos(φ) Vdc (A.4)

+ with Vac and I the r.m.s. values of the balanced phase-to-ground voltage and positive sequence phase current. Substituting cos(φ) = 0 (only reactive power supply or consumption) in Equation (A.4) results in zero current taken from the DC bus. With the mentioned assumptions, reactive power supply or consumption does not aﬀect the DC bus voltage and can be controlled independently for each converter. Active power exchange results in a constant current in the DC bus.

Negative sequence current In the unbalanced situation where the AC current contains a negative sequence component, the DC current is DC current of AC/DC converter 165 equal to

√2V1 sin(ωt)√2I1 sin(ωt φ) 1 ◦ − ◦ idc(t) = +√2V2 sin(ωt 120 )√2I2 sin(ωt 240 φ) Vdc  − ◦ − ◦ −  +√2V3 sin(ωt 240 )√2I3 sin(ωt 120 φ)  − − −  V1I1(cos(φ) cos(2ωt φ)) 1 ◦ − − = +V2I2(cos(120 + φ) cos(2ωt φ)) V ◦ dc  +V I (cos(240 + φ) − cos(2ωt − φ))  3 3 − − − cos(φ) cos(2ωt φ)  VacI ◦ − − = + cos(120 + φ) cos(2ωt φ) V ◦ dc  + cos(240 + φ) − cos(2ωt − φ)  − − V I−  = − ac 3 cos(2ωt φ) Vdc − (A.5)

− with Vac and I the r.m.s. values of the balanced phase-to-ground voltage and negative sequence phase current. The DC current consists only of a pulsating component with a frequency that is twice the AC system power frequency.

Zero sequence current For the unbalanced situation where the AC current consists of only a zero sequence component the DC current is equal to

√2V1 sin(ωt)√2I1 sin(ωt φ) 1 ◦ − idc(t) = +√2V2 sin(ωt 120 )√2I2 sin(ωt φ) Vdc  − ◦ −  +√2V3 sin(ωt 240 )√2I3 sin(ωt φ)  − −  V1I1(cos(φ) cos(2ωt φ)) 1 ◦ − − ◦ = +V2I2(cos( 120 + φ) cos(2ωt 120 φ)) V ◦ ◦ dc  +V I (cos(−240 + φ) − cos(2ωt − 240 − φ))  3 3 − − − −  cos(φ) cos(2ωt φ)  V I0 = ac + cos( 120◦ + φ) − cos(2ωt − 120◦ φ) V ◦ ◦ dc  + cos(−240 + φ) − cos(2ωt − 240 − φ)  − − − − = 0   (A.6)

0 with Vac and I the r.m.s. values of the balanced phase-to-ground voltage and zero sequence phase current. A zero sequence component in the AC current does not aﬀect the DC bus: the DC current due to this current is equal to zero.

Appendix B

Simulations and experimental results practical set-up

Figures B.1 AC current and AC voltage controller responses. . . . 168 B.2 AC and DC voltage controller responses...... 169 B.3 Three-phase connection of unloaded Area 2...... 170 B.4 Three-phase connection of loaded Area 2...... 171 B.5 Connection of phase B of unloaded Area 2...... 172 B.6 Connection of phase B of loaded Area 2...... 173 B.7 Connection of phase C of unloaded Area 2...... 174 B.8 Connection of phase C of loaded Area 2...... 175 B.9 Three-phase disconnection of unloaded Area 2. . . . . 176 B.10 Three-phase disconnection of loaded Area 2...... 177 B.11 Disconnection of phase C of unloaded Area 2...... 178 B.12 Disconnection of phase C of loaded Area 2...... 179 B.13 Disconnection of phase B of unloaded Area 2...... 180 B.14 Disconnection of phase B of loaded Area 2...... 181 B.15 Mitigation of 12, 15 and 20 % voltage dips...... 182 B.16 Mitigation of 30 and 50 % voltage dips and 17 % swell. 183

167 168 Appendix B ∗ ∗ ∗ I I I I V V 0.07 0.08 0.07 0.08 0.07 0.08 0.06 0.06 0.06 0.05 0.05 0.05 (s) (s) (s) t t t 0.03 0.04 0.03 0.04 0.03 0.04 0.02 0.02 0.02 0.01 0.01 0.01 0 0 0

0 0 0 1 1 1

-1 -1 -1

(p.u.) V (p.u.) I (p.u.) I (b) Step change of active power set-point (meas.). (d) Step change of reactive power set-point (meas.). (f) Step change ofsurement). voltage amplitude set-point (mea- ∗ V V ∗ ∗ I I I I 0.07 0.08 0.07 0.08 0.07 0.08 0.06 0.06 0.06 0.05 0.05 0.05 (s) (s) (s) t t t 0.03 0.04 0.03 0.04 0.03 0.04 Figure B.1: AC current and AC voltage controller responses. 0.02 0.02 0.02 0.01 0.01 0.01 0 0 0

0 0 0 1 1 1

-1 -1 -1

(p.u.) V (p.u.) I (p.u.) I (a) Step change of active power set-point (sim.). (c) Step change of reactive power set-point (sim.). (e) Step change ofulation). voltage amplitude set-point (sim- Simulations and experimental results practical set-up 169 ∗ V V 0.07 0.08 0.3 0.35 0.06 0.25 0.05 0.2 (s) (s) t t 0.03 0.04 0.1 0.15 0.02 0.05 0.01 0 0 0 1 1 -1

1.04

1.02

0.98

(p.u.) V dc (p.u.) V (b) Connection ofment). 0.9 p.u. resistive load (measure- (d) Connection ofment). 0.67 p.u. resistive load (measure- ∗ V V 0.07 0.08 0.3 0.35 0.06 0.25 0.05 0.2 (s) (s) t t Figure B.2: AC and DC voltage controller responses. 0.03 0.04 0.1 0.15 0.02 0.05 0.01 0 0 0 1 1 0 -1 -1

0.5

-0.5

dc (p.u.) V (p.u.) P (a) Connection of 0.9 p.u. resistive load (simulation). (c) Connection of 0.67ment). p.u. resistive load (measure- 170 Appendix B 0.2 0.2 Grid-side Converter-side 0.15 (s) (s) 0.1 0.15 t t 0.05 0.05 0.1 0 0 (b) Converter current (measurement). 1 1 0 0 -1

0.5 0.5

-0.5

abc,rms ab (p.u.) I (p.u.) V (d) Load-break switch voltages (measurement). PSfrag replacemen 0.2 0.2 Grid-side Converter-side 0.15 (s) (s) 0.1 0.15 t t 0.05 0.1 0.05 0 0 Figure B.3: Three-phase connection of unloaded Area 2 with public LV network. (a) Converter current (simulation). 1 0 1 0 -1

0.5

0.5 (c) Load-break switch voltages (simulation).

-0.5

bc (p.u.) V rms (p.u.) I PSfrag replacemen Simulations and experimental results practical set-up 171 2 0.2 1 1.5 (s) (s) 0.1 0.15 t t 0.5 0.05 0.2 Grid-side Converter-side 0 0 0.15 (b) Converter current (measurement). 1 1 0 0

0.5 0.5

abc,rms abc,rms (p.u.) I (p.u.) I (d) Converter current,ment). longer timescale (measure- (s) t 0.05 0.1 2 0.2 0 1 0 -1

0.5

(e) Load-break switch voltages (simulation).

-0.5

bc (p.u.) V 1 1.5 (s) (s) 0.1 0.15 t t 0.5 0.05 PSfrag replacemen Figure B.4: Three-phase connection of loaded Area 2 with public LV network. 0 0 (a) Converter current (simulation). 1 1 0 0

0.5 0.5

rms rms (p.u.) I (p.u.) I (c) Converter current, longer timescale (simulation). 172 Appendix B 0.2 0.2 Grid-side Converter-side 0.15 0.15 (s) (s) 0.1 t t 0.05 0.1 0.2 0 0 0.05 0.15 (b) Converter current (measurement). 1 0 0 -1

0.5 0.3 0.4 0.1 0.2

-0.5

bc (p.u.) V (p.u.) I (d) Load-break switch voltages (measurement). − (s) 0.1 t PSfrag replacemen 0.2 − abc,rms I I Grid-side Converter-side 0 0.05 0.15 0.15 0.2 1 0 -1 0.5

-0.5

(e) Load-break switch current (measurement). (p.u.) I (s) (s) 0.1 t t 0.05 0.1 0.05 0 0 Figure B.5: Connection of phase B of unloaded Area 2 with public LV network. (a) Converter current (simulation). 1 0 1 0 -1

0.5 0.5

-0.5

bc (p.u.) V (p.u.) I PSfrag replacemen Simulations and experimental results practical set-up 173 2 0.2 Grid-side Converter-side 1.5 0.15 1 (s) (s) t t 0.05 0.1 0 0 0.5 (b) Converter current (measurement). 1 0 0 -1

0.5 0.3 0.4

0.1 0.2

-0.5

bc (p.u.) V (p.u.) I (d) Load-break switch voltages (measurement). − PSfrag replacemen 0.2 − abc,rms I I Grid-side Converter-side 0.15 0.15 0.2 (s) (s) 0.1 t t 0.05 0.1 0.05 Figure B.6: Connection of phase B of loaded Area 2 with public LV network. 0 0 (a) Converter current (simulation). 1 0 1 0 -1

0.5

0.5 (c) Load-break switch voltages (simulation).

-0.5

bc (p.u.) V (p.u.) I PSfrag replacemen 174 Appendix B 0.45 2 0.4 Grid-side Converter-side 0.3 0.35 0.25 1 1.5 (s) (s) t t 0.2 0.5 0.1 0.15 0.05 0 0 (b) Converter current (measurement). 1 0 1 0 -1

0.5 0.5

-0.5

bc rms (p.u.) V (p.u.) I (d) Load-break switch voltages (measurement). 0.2 0.2 Grid-side Converter-side 0.15 (s) (s) 0.1 0.15 t t 0.05 0.1 0.05 0 0 Figure B.7: Connection of phase C of unloaded Area 2 with public LV network. (a) Converter current (simulation). 1 0 1 0 -1

0.5

0.5 (c) Load-break switch voltages (simulation).

-0.5

bc (p.u.) V rms (p.u.) I PSfrag replacemen Simulations and experimental results practical set-up 175 0.45 2 0.4 Grid-side Converter-side 0.3 0.35 0.25 1 1.5 (s) (s) t t 0.2 0.5 0.1 0.15 0.4 0.45 0.05 0 0.35 0 (b) Converter current (measurement). 1 0 1 0.3 0 -1

0.5 0.5

-0.5

bc abc,rms (p.u.) V (p.u.) I (d) Load-break switch voltages (measurement). (s) t 0.2 0.25 0.15 0.2 0.2 0.1 Grid-side Converter-side 0 0.05 0.15 0 1 -1

(e) Load-break switch current (measurement).

(p.u.) I (s) (s) 0.1 0.15 t t 0.05 0.1 0.05 Figure B.8: Connection of phase C of loaded Area 2 with public LV network. 0 0 (a) Converter current (simulation). 1 0 1 0 -1

0.5

0.5 (c) Load-break switch voltages (simulation).

-0.5

bc (p.u.) V rms (p.u.) I PSfrag replacemen 176 Appendix B 0.2 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 0.15 0.15 0.15 (s) (s) (s) 0.1 t t t 0.05 0.1 0.05 0.1 0 0 0 0.05 (a) Power frequency Area 2 (simulation). 1 1 0 0 -1 -1 51 50 49 0.5 0.5

(b) Load-break switch voltages (simulation).

-0.5 -0.5

50.5 49.5

bc ab (p.u.) V (p.u.) V (c) Load-break switch voltages (measurement). (Hz) f PSfrag replacemen PSfrag replacemen Figure B.9: Three-phase disconnection of unloaded Area 2 from public LV network. Simulations and experimental results practical set-up 177 0.2 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 0.15 0.15 0.15 (s) (s) (s) 0.1 of Area 2 is supplied by converter t t t 0.05 0.1 0.05 0.1 0 0.05 0 0 (b) Power frequency Area 2 (simulation). 1 1 0 0 51 50 -1 -1 49 0.5 0.5

(d) Load-break switch voltages (simulation).

-0.5 -0.5

50.5 49.5

ab bc (f) Load-break switch voltages (measurement). (Hz) f (p.u.) V (p.u.) V PSfrag replacemen PSfrag replacemen 0.2 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 0.15 0.15 0.15 (s) (s) (s) 0.1 t t t 0.05 0.1 0.05 0.1 0 0 0 0.05 1 1 (a) Power frequency Area 2 (simulation). 0 0 -1 -1 51 50 0.5 0.5 49

-0.5 -0.5

50.5 49.5

ab bc (p.u.) V (p.u.) V (e) Load-break switch voltages (measurement). (Hz) f PSfrag replacemen PSfrag replacemen Figure B.10: Three-phase disconnectionin of loaded ﬁgures Area a), 2 c) from and public LV e) network. and Load from the public LV grid in ﬁgures b), d) and f). 178 Appendix B 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 1.5 2 0.15 0.15 1 (s) (s) (s) t t t 0.5 0.05 0.1 0.05 0.1 0 0 0 (a) Converter voltage (simulation). 0 1 1 0 0

-1 -1 0.1 0.5 0.5

0.05 (b) Load-break switch voltages (simulation).

-0.5 -0.5

(p.u.) V

bc bc (p.u.) V (p.u.) V (c) Load-break switch voltages (measurement). − PSfrag replacemen PSfrag replacemen Figure B.11: Disconnection of phase C of unloaded Area 2 from public LV network. Simulations and experimental results practical set-up 179 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 0.15 0.15 1.5 2 (s) (s) 1 (s) t t t ad of Area 2 is supplied by converter 0.5 0.05 0.1 0.05 0.1 0 0 0 (b) Converter voltage (measurement). 1 1 0 0 0 -1 -1 0.5 0.5 0.1

(d) Load-break switch voltages (simulation).

-0.5 -0.5

0.05

bc bc (f) Load-break switch voltages (measurement). (p.u.) V (p.u.) V (p.u.) V − PSfrag replacemen PSfrag replacemen 0.2 0.2 Grid-side Converter-side Grid-side Converter-side 1.5 2 0.15 0.15 1 (s) (s) (s) t t t 0.5 0.05 0.1 0.05 0.1 0 0 0 (a) Converter voltage (measurement). 0 1 1 0 0 0.1

-1 -1

0.05 0.5 0.5

-0.5 -0.5

(p.u.) V

bc bc (p.u.) V (p.u.) V (e) Load-break switch voltages (measurement). − PSfrag replacemen PSfrag replacemen Figure B.12: Disconnection ofin phase ﬁgures C a), of c) loaded Area and 2 e) from and public from LV the network. public Lo LV grid in ﬁgures b), d) and f). 180 Appendix B 0.8 2 2 0.7 1.5 0.5 0.6 1 (s) 1 1.5 (s) (s) t t t 0.4 0.3 0.5 Grid-side Converter-side 0.1 0.2 0 0.5 0 0 (a) Converter voltage (measurement). 1 0 1 1 -1

0.5 1.1 0.9

1.1 0.8 (b) Load-break switch voltages (simulation). 0.9

-0.5

bc (p.u.) V (c) Load-break switch voltages (measurement). bc (p.u.) V (p.u.) V + Figure B.13: Disconnection of phase B of unloaded Area 2 from public LV network. Simulations and experimental results practical set-up 181 0.8 2 2 0.7 1.5 0.5 0.6 1 1 1.5 (s) (s) (s) t t t 0.4 ad of Area 2 is supplied by converter 0.3 0.5 Grid-side Converter-side 0.1 0.2 0 0.5 0 (b) Converter voltage (measurement). 0 1 1 1 0

-1 1.1 0.9

0.5 1.1 0.8 (d) Load-break switch voltages (simulation). 0.9

-0.5

(p.u.) V bc bc (f) Load-break switch voltages (measurement). (p.u.) V (p.u.) V + 0.8 2 2 0.7 1.5 0.5 0.6 1 1 1.5 (s) (s) (s) t t t 0.4 0.3 0.5 Grid-side Converter-side 0.1 0.2 0 0.5 0 0 (a) Converter voltage (measurement). 1 0 1 1 -1

0.5

1.1 0.9 1.1 (c) Load-break switch voltages (simulation). 0.9 0.8 -0.5

bc (e) Load-break switch voltages (measurement). (p.u.) V bc (p.u.) V (p.u.) V + Figure B.14: Disconnection ofin phase ﬁgures B a), of c) loaded Area and 2 e) from and public from LV the network. public Lo LV grid in ﬁgures b), d) and f). 182 Appendix B 0.15 0.2 0.15 0.2 0.15 0.2 (s) (s) (s) 0.1 0.1 0.1 t t t 0.05 0.05 0.05 0 0 0 1 1 1 0 0 0

-1 -1 -1 0.5 0.5 0.5

-0.5 -0.5 -0.5

(p.u.) Q (p.u.) Q (p.u.) Q (b) Reactive power during 12 % voltage dip (meas.). (d) Reactive power during 15 % voltage dip (meas.). (f) Reactive power during 20 % voltage dip (meas.). 0.2 0.2 0.2 No Mitigation Mitigation No Mitigation Mitigation No Mitigation Mitigation 0.15 0.15 0.15 (s) (s) (s) 0.1 0.1 0.1 t t t Figure B.15: Mitigation of 12, 15 and 20 % voltage dips. 0 0.05 0 0.05 0 0.05 1 1 1

1.2 1.2 1.2 0.8 0.8 0.8 0.6 0.6 0.6

(p.u.) V (p.u.) V (p.u.) V (c) Voltage during 15 % voltage dip (measurement). (e) Voltage during 20 % voltage dip (measurement). (a) Voltage during 12 % voltage dip (measurement). Simulations and experimental results practical set-up 183 0.2 0.15 0.15 0.2 0.15 0.2 (s) (s) (s) 0.1 0.1 0.1 t t t 0.05 0.05 0 0 0 0.05 1 1 0 0 1 0 -1 -1

-1 0.5 0.5 0.5

-0.5 -0.5

-0.5

(p.u.) Q (p.u.) Q (p.u.) Q (b) Reactive power during 30 % voltage dip (meas.). (d) Reactive power during 50 % voltage dip (meas.). (f) Reactive power during 17 % voltage swell (meas.). 0.2 0.2 0.2 No Mitigation Mitigation No Mitigation Mitigation No Mitigation Mitigation 0.15 0.15 0.15 (s) (s) (s) 0.1 0.1 0.1 t t t Figure B.16: Mitigation of 30 and 50 % voltage dips and 17 % swell. 0 0.05 0 0.05 0 0.05 1 1 1

1.2 1.2 0.8 0.8 1.2 0.8 0.6 0.6 0.6

(p.u.) V (p.u.) V (p.u.) V (c) Voltage during 50 % voltage dip (measurement). (a) Voltage during 30 % voltage dip (measurement). (e) Voltage during 17 % voltage swell (measurement).

Acknowledgements

Now that I am at the end of this PhD project, I have the pleasure to write some words of gratitude. Many people were involved directly or indirectly in my research. I am grateful to all of them; here I want to address some in particular, realizing it is impossible to be complete. The theoretical part of the work leading to this thesis was performed at the Electrical Energy Systems research group, while the practical part took place at the Electromechanics and Power Electronics research group of Eindhoven University of Technology. I thank my promotor Wil Kling for the trust he gave me to ﬁnd my own way in the research, for his scrutiny when reviewing this thesis and the papers we published together. I want to thank my copromotor Jorge Duarte for his vital support when starting and performing the practical part of this research and for reviewing all my work. Thanks to Johanna Myrzik for her eﬀorts in the beginning of this research project. Also thanks to Jan Blom for his support when it was most needed. I am grateful for the practical assistance of Wim Thirion, who helped turning my ideas into a working set-up. Thanks also to all the other colleagues of the EES and EPE groups for the help, discussions and the pleasant working atmosphere. I thank the students Daniel Persson, Wouter Bos and Frank van den Bergh for helping me with the theoretical and practical work and for the pleasant cooperation. Parts of their work have been included in this thesis. This project was part of the IOP-EMVT project ’Intelligent Power Systems’. I thank all members of the IOP-EMVT supervisory committee for the discussions and valuable feedback during the half year meetings. I appreciated the cooperation with all the other PhD students from Eindhoven and Delft University, and I thank them for sharing their experiences, for the fruitful discussions and for the pleasant company. I am grateful to all members of the PhD committee for being prepared to take upon them that role and for their valuable comments and questions which improved this thesis, and for participating in my defense ceremony. I thank Hans Overbeek from KEMA for challenging me to take upon

185 186 Acknowledgements me this research project and for providing me with the time to do so. Thanks also to AndréZeijseink for continuing this support which made it possible to finish the work. Also I want to thank all my colleagues at KEMA for the discussions and help, and for their flexibility in dealing with my limited availability to do ’normal work’. Especially, I thank my coach Peter Vaessen for his valuable support, review of this thesis and our papers and for his participation in the committee, and Johan Enslin for being my coach in the beginning of this work. Although less directly involved in this research project, I want to thank my family and friends for their interest in (the progress of) my PhD work. I do not know whether I was able to explain what it was all about, but they remained interested anyhow, thanks for that. Finally, a big ’thank you’ to Susana for her continuous support, en- couragement, patience and love. Curriculum vitae

Roald A.A. de Graaff was born in Waalwijk, the Netherlands, in 1975. He attended secondary school at Dr. Mollercollege in Waalwijk, where he graduated on Gymnasium B in 1993. He received the M.Sc. degree in Electrical Engineering from Eindhoven University of Technology, the Netherlands, in 1998. During his education, he joined Chalmers Univer- sity of Technology, Gothenburg, Sweden, for an internship concerning voltage dip immunity of electrical drives. After graduating, he worked at Eindhoven University of Technology as a research assistant in the field of electromagnetic compatibility in railway systems and public electrical power systems. Since 2001 he has been with KEMA in Arnhem, the Netherlands, where he is a consultant in electrical power systems and electromagnetic compatibility. In the end of 2004 he started as a PhD student in a part-time collaboration with Eindhoven University of Technology on the subject of flexible distribution systems through the application of multi back-to-back converters. This research project has led to this dissertation.

187