dti

HIGH-TEMPERATURE SUPER­ CONDUCTING FAULT-CURRENT LIMITER

Optimisation of Superconducting Elements

CONTRACT NUMBER: K/EL/00304/00/00

URN NUMBER: 04/1488 The DTI drives our ambition of ‘prosperity for all’ by working to create the best environment for business success in the UK. We help people and companies become more productive by promoting enterprise, innovation and creativity.

We champion UK business at home and abroad. We invest heavily in world-class science and technology. We protect the rights of working people and consumers. And we stand up for fair and open markets in the UK, Europe and the world.

ii High-temperature Superconducting Fault-current Limiter - Optimisation of Superconducting Elements

K/EL/00304/00/REP URN 04/1488

Contractor

VA TECH T&D UK Ltd

The work described in this report was carried out under contract as part of the DTI Technology Programme: New and Renewable Energy, which is managed by Future Energy Solutions. The views and judgements expressed in this report are those of the contractor and do not necessarily reflect those of the DTI or Future Energy Solutions.

First published 2004 © V A Tech copyright 2004

ii i EXECUTIVE SUMMARY

Introduction

This project was undertaken with VA TECH T&D UK Ltd as lead partner, in conjunction with the Interdisciplinary Research Centre in Superconductivity (IRC) of the University of Cambridge and Advanced Ceramics Limited (Stafford) as subcontractors. The project was initiated to continue with work started in 1995 under a DTI - LINK Collaborative Research Programme: Enhanced Engineering Materials "Enhancing the Properties of Bulk High Temperature Superconductors and their Potential Application as Fault Current Limiters". This project culminated with the testing of a full-scale 11kV prototype limiter in 1999.

Fracture and burn-back of certain of the Bi-2212 elements used in the full-scale demonstrator had occurred and this was investigated prior to commencing the current project. It was concluded that:

• Three-dimensional mathematical modelling of the quench process was required to optimise the design of the elements. No such model, which would need to calculate the evolution over time of thermal, electric and magnetic fields simultaneously had yet been developed in three dimensions. The modelling would need to work on volumetric elements of a small size to allow localised material characteristics to be incorporated to make provision for in-homogeneities of the characteristics of the superconducting material.

• An improved method of testing elements following manufacture would be needed to maintain the critical current densities within acceptable limits. Modelling could be used to evaluate more precisely what the acceptable limits are.

• Resistive shunting of each element was desirable to help prevent the development of localised quench sites leading to hot spots during operation.

• Rapid cooling of the elements during initial immersion in the cryo-coolant had caused high levels of strain in the elements which were bonded to a substrate having a different thermal conductivity from that of the superconducting material. Precautions were required to prevent this.

The investigation which yielded these conclusions was undertaken in 2001 at the expense of VA TECH T&D, in order to help to design a programme of work to optimise the design, manufacture, testing, handling and mounting of the superconducting elements. The programme was then used as the basis of a proposal to the DTI for a funded project which was initiated in December 2002 and completed in March 2004. The project, initially conceived with a twelve month timescale, was extended to fifteen months due to delays in the manufacture of the superconducting elements, which is still a difficult and labour-intensive process. It was decided also to

iv incorporate other work to help to assess the feasibility of a practical distribution-scale Superconducting Fault-current Limiter (FCL). The additional uncertainties were:

• The quantity of heat generated within and entering the cryo-enclosure would need to be better evaluated to assist with the design and costing of the enclosure and cooling provisions. This would depend on the element design and the arrangements for mounting and connection of the elements and the connection of the limiter structure to the external environment. Various options would need to be investigated.

• The installation costs associated with placing an FCL in a substation would need to be estimated.

• Target specifications for FCLs for various applications were required.

This report presents the results of the investigations into the above areas. It also contains a detailed, carefully researched survey of the potential market for distribution-scale FCLs in the UK to form the basis of a business plan for commercialisation of the technology. The market study was funded by VA TECH T&D and was not included in the project proposal.

Results

An overview is given here of the results of investigations into each of the above areas of uncertainty:

3D Modelling program suite

Modules were coded for input, calculation and output of numerical data. Graphical output of each aspect of interest is provided. The program suite is user interactive, enabling input data to be changed quickly and easily. Calculation speed depends on the number of volumetric elements defined within a given volume of material and has been found adequate for the purposes of this work. Results of the modelling have been confirmed in laboratory tests.

Testing method and acceptance criteria

Equipment for characterising the superconducting material is currently in use in the laboratory. Recommendations for adapting this to the factory environment are presented. A parallel programme of testing and modelling been used to determine the required accuracy of measurements to ensure stable quenching of series- connected superconducting elements.

Resistive shunting

Bi2212 elements with internal and external silver layers were designed, manufactured and tested.

v Shunted elements of both types have been subjected to repeated quench cycles without degradation.

Controlled Cooling

Bi2212 elements have been cooled for a period of time, by immersion in gaseous nitrogen at marginally above the boiling point, prior to introducing nitrogen in the liquid phase. This is sufficient to prevent excessive strain in the material.

Thermal Losses

Extensive calculation of thermal losses for various arrangements has been carried out. These show that, whilst active cooling is desirable at modest levels of load current, the power consumption of the cooling equipment is within acceptable limits.

Installation Costs

These have been determined for the most likely initial application as bus-tie in a distribution substation at 11kV.

FCL Specifications

Rating requirements for a range of applications have been considered.

Market Survey

Published data detailing the plant ratings and local fault levels in most of the primary distribution substations in the UK were studied and it is concluded that there is an evolving need for action to prevent inadequate plant rating from becoming a problem for DNO's. Currently, around 10% of these locations will need switchgear upgrades or other action within a time period varying from months to a few years. The potential market size for FCLs has been evaluated on this basis.

Conclusions

By addressing all of the above issues, this work brings the concept of a fault-current limiter using high-temperature superconducting material in liquid nitrogen closer to practical implementation at full scale.

The next logical step is to design, construct and test a unit for field-trial / demonstration in a distribution network.

vi Acknowledgements

Thanks are due to all of the many contributors, notably Dr. Bernhardt Zeimetz, Dr. Tim Coombes, Professor Archie Campbell, Professor Jan Evetts, Ahmed Kursumovic, Dr. Milan Majoros, Kartek Tadinada and Matthew Hills at Cambridge IRC and Materials Science; Dr. Peter Malkin, Paul Barnfather and Herbert Piereder at VA TECH T&D.

vii Contents Page Executive Summary ii Introduction ii Results iii Conclusions iv Acknowledgements iv

Final Report 1

1 INTRODUCTION 1 1.1 Technical Background 1 1.2 Aims and Objectives 1 1.3 Benefits 2 1.4 Programme of Work 2

2 COMPUTER MODELLING OF QUENCH PROCESS 4 2.1 Introduction 5 2.2 Computer Model 5 2.3 Results 7 2.3.1 Thermal instability 7 2.3.2 Dynamic Response 11 2.3.3 Observation of 'hot spots' 12 2.3.4 Current - Voltage scaling 14 2.4 Conclusions 15

3 COMPOUND ELEMENT 17

4 MULTI-ELEMENT MODEL 17

5 FCL DESIGN 18 5.1 Distribution Bus-coupler 18 5.2 Embedded Generator Connection 19 5.3 Larger Generator Connection 19 5.4 Hazardous Area Safety 19 5.5 Interconnection to Fault-prone Network 19

6 THERMAL LOSS EVALUATION 23 6.1 Introduction 23 6.2 AC losses in a single bar 23 6.3 AC Losses in several bars parallel to each other 24 6.4 AC losses in resistive joints 34 6.5 AC losses in currents leads 35 6.6 Cooling power 36

7 TEST EQUIPMENT AND SCHEDULE 38 7.1 Mounting 38 viii 7.2 Cooling 38 7.3 Electrical Supplies and Data Collection 38 7.4 Data Collection and Analysis 39

8 OPTIMISED ELEMENT 40 8.1 Introduction 40 8.2 V-I Characteristics 40 8.3 Single Element Simulation 41 8.4 Thermal Design 42 8.5 Homogeneity 43 8.6 Experiments 46 8.7 Conclusions 48

9 INSTALLED COST DATA 50

10 UK MARKET STUDY 51

11 PROJECT SUMMARY 51

Appendices

Appendix 1 Development of Optimised Current Elements for a Resistive Fault Current Limiter - Stage I Programme: Evaluation of First Demonstrator

ix 1 INTRODUCTION

1.1 Technical Background to the Project

This project was initiated to continue with work started in 1995 under a DTI - LINK Collaborative Research Programme: Enhanced Engineering Materials "Enhancing the Properties of Bulk High Temperature Superconductors and their Potential Application as Fault Current Limiters".

The LINK programme led to the development of a production process for bars made from high-temperature superconducting material (Bi-2212) which were incorporated in prototype Fault Current Limiters (FCLs). A low-voltage "bench-top" prototype FCL was used to demonstrate successfully the principle and to gather data for further design work. A full-scale single-phase prototype was then constructed and tested. The results of these tests indicated that further work was required in two fundamental areas. The first of these was to develop a sophisticated model of the process of "quenching" (the transition from superconducting to normal conducting states which the material undergoes when large currents are present) so that various design options could be examined theoretically. The second area of work entailed improvements to the design and manufacture of the superconducting elements to improve the uniformity and predictability of the quenching process.

This project forms the first stage of a programme intended to deliver a full-scale fault- current limiter for installation in an electrical distribution network for endurance testing, trial and demonstration purposes.

1.2 Aims and Objectives of the Project

The technical feasibility of fault-current limiters using the relatively new high- temperature ceramic superconductors has been amply demonstrated1. The main advantage of the new materials is that they can be cooled using low-cost and readily available liquid nitrogen at 77K. Earlier superconductors required cooling in liquid helium at 4K, which is much more expensive both to obtain and to store. It has not yet however been possible to produce a FCL at a cost that would make it a viable industrial product. The remaining barriers to commercialising the technology are uncertainties in the following areas:

1.2.1 Is it possible to design a range of limiters for various network applications using standard superconducting element design(s) in appropriate configurations e.g. series / parallel connection - using magnetic triggering if necessary? If so, what is the optimum superconducting element design?

1.2.2 What are the likely lifetime costs, in particular for coolant replenishment or cryocooling, for various designs of thermal envelope?

1 “6.4MVA Resistive Fault Current Limiter Based on Bi-2212 Superconductor” submitted by ABB to EUCAS 2001

1 1.2.3 What additional local protection and monitoring arrangements are likely to be required by network owners / operators and what are the likely additional costs?

Satisfactory answers to the above questions will provide the confidence required to move towards a commercially viable FCL design and manufacturing facility.

1.3 Benefits

The potential benefits of fault-current limiters have long been recognised. These include improved safety, security and availability of plant and network, improved flexibility of network configuration, improved power quality and reduced cost of network reinforcement. Additionally, in the last 12 months it has been recognised that the availability of cost-effective and reliable limiters would have a positive impact on the feasibility of installing new generation connected to the distribution networks - known as "Embedded" or Distributed" Generation. This new generation is seen as a necessary element of Government initiatives to meet obligations to reduce CO2 emissions and to increase utilisation of renewable energy sources. The new generation will include combined heat and power (CHP), wind, landfill gas, biofuels and other emerging technologies, with outputs typically in the range from 1 to 10MW. Fault current limiters used in conjunction with new generation connections at distribution voltage would minimise the impact of a connection on the network fault level and stability. This could in many instances remove the need for upgrading switchgear and cables, the fault ratings of which would otherwise be exceeded. A limiter would also enable new plant of lower fault rating than would otherwise be required to be installed at the connection. These factors could dramatically reduce the cost of new connections, as well as permitting connections to be made at locations which would otherwise be regarded as unsuitable.

The development of this technology in the UK is entirely appropriate if radical changes in network configuration are to be achieved and will provide a UK lead in technology and expertise to facilitate such changes elsewhere.

1.4 Programme of Work: Project Activities

1.4.1 Activity Description

(a) Quench Modelling - IRC in Superconductivity at Cambridge

(i) Advanced numerical modelling of current, voltage and temperature within a superconducting element, with applied magnetic field and shunting.

3 months, to start at commencement

(ii) Extension of the model to deal with elements connected in series / parallel.

1 month, to start at completion of a(i)

2 (iii) Modelling various configurations of FCL.

8 months, to start at completion of a(ii)

(iv) Optimisation of design to target specification for a field-trial unit.

8 months in parallel with a(iii)

(v) Evaluate Thermal Losses under Normal Conditions for each Design Option.

8 months, in parallel with a(iii) and a(iv)

(b) Materials Development and Testing / Screening of Superconducting Elements Department of Materials Science / Device Materials Group at University of Cambridge with Advanced Ceramics Limited, Stafford

(i) The incorporation of thin conducting layers of a suitable metal either on the surface of the elements or within the element layer structure will be explored both by modelling and by making detailed voltage drop tests on prototype test elements during ac testing. These layers will be contoured and patterned at the 'green' stage, either as metal foil or laminates incorporating metal powder with binders and plasticisers. Extensive work demonstrating this process using silver has already been undertaken.

3 months, to start at commencement

(ii) Examine the effects of changes to the geometry of the current element to exploit the anisotropy of the critical current. The aim is to develop element geometries that result in a rapidly changing magnetic field environment during a quench.

9 months, to start at completion of b(i)

(iii) Develop methods for characterisation of elements as quality control measures for mass production.

12 months, to start at commencement

(c) Determine the likely costs of installation and ownership for a complete system - VA TECH T&D UK Ltd

This includes the limiter itself, supervisory systems for monitoring its condition and additional protective equipment including series circuit-breaker and protection.

3 12 months, to start at commencement

(d) Project Management and Reporting VA TECH T&D UK Ltd

4 2 COMPUTER MODELLING OF QUENCH PROCESS

This portion of the work constitutes the major effort made within this programme and has led to a modelling suite able to predict the quenching (transition from superconducting to normal conducting state) behaviour of superconducting material. Electric. magnetic and thermal fields are examined in three dimensions at a physical scale which can be chosen by the user. Rates of heat flow from the surfaces, which can be coupled to a solid substrate or directly to a cryocoolant are taken account of. Inhomogeneities in the critical current density Jc of the material can be specified locally, or a semi-random spread of varying Jc can be "scattered" throughout the material of an element.

The following paper, published in the Institute of Physics Journal of Superconducting Science and Technology, Volume 17 (2004) pages 657-662, gives a detailed account of the modelling methods and application of the model to predicting the behaviour of Bi2212 bars of typical design.

5 THERMAL INSTABILITY AND CURRENT-VOLTAGE SCALING IN SUPERCONDUCTING FAULT CURRENT LIMITERS

B. Zeimetz1,3, K. Tanida 2, D. E. Eves2, T. A. Coombs 2, J. E. Evetts1,3 and A. M. Campbell 2,3

^Department of Materials Science and Metallurgy, Cambridge University, Pembroke Street, Cambridge CB1 3QZ, United Kingdom 2Department of Engineering, Cambridge University, Trumpington Road, Cambridge, .U.K. 3IRC in Superconductivity, Cambridge University, Madingley Road, Cambridge, U.K. ^corresponding author; e-mail [email protected]

ABSTRACT

We have developed a computer model for the simulation of resistive superconducting fault current limiters in three dimensions. The program calculates the electromagnetic and thermal response of a superconductor to a time-dependent overload voltage, with different possible cooling conditions for the surfaces, and locally variable superconducting and thermal properties. We find that the cryogen boil-off parameters critically influence the stability of a limiter. The recovery time after a fault increases strongly with thickness. Above a critical thickness, the temperature is unstable even for a small applied AC voltage. The maximum voltage and maximum current during a short fault are correlated by a simple exponential law.

2.1 INTRODUCTION

Superconducting Fault Current Limiters (SFCL) are one of the most attractive applications of High Temperature Superconductors in electrical power engineering [1]. They could replace the slow and resistive mechanical switches which are being used in high/medium voltage transmission & distribution networks, thereby enhancing power quality and network efficiency. Two distinctive design concepts have evolved for SFCL: Resistive current limiters utilise the nonlinear resistivity of superconductors, while inductive SFCL are based on their magnetic shielding properties.

In order to design SFCL for practical applications, knowledge of SFCL parameters such as maximum load, response time and thermal recovery time is essential. On the other hand, the electromagnetic and thermal response of a SFCL to a fault involves very high voltages and currents at very short times, and is therefore a formidable challenge for experiments. Computer simulations are obviously helpful, since they allow research on arbitrary time scales and power levels.

This article presents recent results of SFCL simulations. In line with recent and current SFCL developments in Cambridge [2], it is focussed on resistive SFCL. The simulation software is based on earlier simulation work at Cambridge University [3][4], which recently has been significantly extended.

6 2.2 COMPUTER MODEL

Our simulation uses a finite-difference approach to calculate the electrical and thermal evolution of a SFCL. The superconducting material is modelled electrically as a tetragonal network of nonlinear resistors, and thermally as a three-dimensional array of tetragonal cells, cf. Fig. 2.1.

Fig. 2.1: schematic representation of the model geometry and resistor network; top left: resistors in an individual cell branching out from the centre; bottom: tetragonal cells form a slab-like geometry with applied voltage and current along x direction

Each cell contains 6 resistors branching out from its centre. The global current lFCL is flowing in x direction, and its magnitude is calculated as function of an applied AC voltage VFCL. The current-voltage characteristic of a resistor is defined as > o II DC for I < lCRi (1a)

VR = V0R*(l/lCRi-1)n for I > lCRi (1b) where V0R, lCRi (i = xy-plane or z-axis) and n are parameters defined locally for each cell, and as function of temperature and magnetic field. This allows the study of inhomogeneous, anisotropic superconductors.

The calculation of the current distribution involves a Newton-Raphson iteration to minimise the resistor network loop voltages as function of the loop currents [5]. By explicitly utilising the cubic symmetry of the model, the calculation speed can be dramatically increased, and the required memory size minimised, so that even large model sizes can be tackled on a standard personal computer. (For example, the calculation shown in Fig. 2.7 with 2000 cells and 1000 time steps takes ca. 150 min on a 400 MHz PC).

7 The thermal response of the FCL to the current is governed by locally defined, temperature dependent specific heat c v and anisotropic thermal conductivity . For the boundary conditions, i.e. the heat flow through the surfaces, one can choose between thermal insulation, thermal sink and boil off of a cryogenic liquid. The heat transfer properties of the boiloff model are shown in Fig. 2.2. The data for the boiloff power were estimated from experimental data given in Ref. [6]. These experimental data scatter over a wide range, up to an order of magnitude. The boiloff data are sensitive for example to the sample geometry and its surface roughness. Therefore our boiloff model can only be seen as a rough estimate, unless boiloff power for a specific sample can be measured.

SO 90 100 110 120 130 140 Temperature [K]

Fig. 2.2: nitrogen boiloff power per surface area against temperature, estimated from [1], as used in the simulation for cooling at surfaces of SFCL; dotted lines indicate transition temperatures and plateau heights; note the semi-logarithmic scaling

The second important feature of the nitrogen boiloff model shown in Fig. 2.2 is the existence of a maximum, which leads to thermal instability, as will be discussed below. The maximum originates from the behaviour of the nitrogen gas on the surface, which at low temperatures exists as individual bubbles ('nucleation boiling'), but then evolves into a continuous gas layer with poor heat transfer ('film boiling') [6],

Using Bi2Sr2CaCu208 (Bi2212) as a model material for our studies, we extracted the electrical input parameters V0R(T,B), lCR(T,B) and n(T,B) from our own measurements on Bi2212 polycrystalline CRT bars [7][8] carried out in a pressurised liquid nitrogen vessel [9][10]. However, reliable c-axis critical current data were not available. Owing to the imperfect texture of the Bi2212 grains in CRT bars, we estimated the critical current anisotropy to be 10 and used this value for our simulations. Thermal data for Bi2212 were extracted from measurements on single crystals [11-13]. In this paper we

8 consider mainly slab-like conductors modelled on CRT bars with length L, width W, thickness D, and D< W< L, see Fig. 2.1.

2.3 RESULTS

2.3.1 Thermal instability in normal (non-fault) operation

The plateau and subsequent decrease in the boil-off heat transfer at the surface leads to thermal instabilities or meta-stable states. The most straightforward case is that of a DC current applied to an ohmic conductor with temperature-independent resistivity. The simulated temperature-time dependence of this scenario is shown in Fig. 2.3, for different voltages. At low voltages, the conductor reaches an equilibrium temperature at which the heat generated by the electrical current is equal to the boil-off power at the surface. If however the temperature exceeds the plateau onset (TB1) of the boil-off function (TB1 = 89 K in our model, cf. Fig. 2.2), no such equilibrium exists, and we observe a thermal 'runaway' of the conductor which in an experiment would lead to its destruction.

time [ms]

Fig. 2.3: simulated temperature against time of a silver bar, with various applied DC voltages as indicated, dotted line indicates temperature of 89 K, where the bar temperature becomes unstable (cf. Fig 2.1); a second upturn can be seen at around 98 K

We have also carried out simulations with a more realistic temperature-dependent resistivity. If the resistivity increases with temperature, the thermal instability is somewhat mitigated, but the qualitative behaviour is very similar. Similarly, replacing the DC with an AC voltage only results in an additional temperature oscillation around the equilibrium values, but otherwise no change in stability behaviour.

In the case of superconducting materials, the situation is complicated by the strong temperature dependence of the electrical properties, and especially by the existence

9 of the critical temperature Tc . Fig 2.4 shows our DC simulation for Bi2212 material, where the value of Tc is ca. 90 K and therefore very close to the boiloff plateau temperature (TB1 = 89 K). (The vicinity of the two temperatures makes it even more crucial to obtain reliable experimental data for the boil-off parameters). As in the ohmic case, at low DC voltages the temperature converges against an equilibrium value. If the voltage is raised so that the temperature just exceeds TB1, the temperature rises to Tc and then oscillates around this value. When the DC voltage is increased further, the temperature diverges as in the ohmic case.

Bi2212 single layer DC voltage

100 mV

50 mV 90 mV

10 mV

Fig. 2.4: simulated temperature against time of a Bi2212 bar, with various applied DC voltages as indicated, dotted and dashed lines indicate temperatures of 89 K and 91 K, respectively

The simulations discussed above and shown in Fig s 2.3 and 2.4 were based on a two-dimensional network model, with only a single layer along the z-axis. This allowed us to focus on the boil-off properties and neglect the internal heat flow inside the conductor. However, the thermal conductivity in Bi2212 is lower by a factor of 100 compared to conventional metals, and it is also anisotropic with a lower value along the crystallographic c-axis [12]. Therefore, in fully three-dimensional simulations, we find a markedly higher temperature gradient in simulations with Bi2212 material, and we also find a strong dependence of thermal behaviour on sample thickness. Fig s 2.5 and 2.6 show simulation data from a fully three-dimensional calculation, in this case using AC applied voltage. The bottom z plane was thermally insulated, and it therefore represents the centre of a symmetric slab of dimensions LxWxD= 20x10x4 mm. All other surfaces were cooled by nitrogen boil-off. Here the surface temperature converges against a value below Tc , while the core temperature is heated above Tc . Apart from the oscillations caused by the AC current variation, shown in the inset, the temperature distribution reaches a stable, partially quenched state. Fig. 2.6 shows the temperature distribution of the central layers in the 'equilibrium' state as a grey scale, and the corresponding local current distribution, indicating local current density by the length of the arrows. The current flows predominantly along the surface layers

10 which are still superconducting. We have verified that this result is independent of the number of z-layers used, i.e. it is not a numerical effect of the discretisation.

Similar to the one-layer DC case (Fig. 2.4), we found that the equilibrium temperature distribution depends on the applied voltage. More interestingly, the equilibrium temperature distribution also depends on the thickness D of the slab: Because the surface-per-volume ratio decreases with increasing thickness, the equilibrium temperature increases with D. This is shown in Fig. 2.5 which includes also the temperature-time data of a simulation with thickness D = 0.2 mm (short line near bottom), at the same voltage 20 mV and same other parameters. In this case, all 5 layer temperatures practically overlap, and the temperature gradient is very small.

Bi2212 20x10x4 mm 50 Hz AC i 20 mV /

surface

Q_ 85"

£2 30"

time [s]

Fig. 2.5: temperature-time dependence calculated in a simulation of a Bi2212 slab using 5 layers along z; showing temperature maxima in each layer; inset: zoomed view of surface temperature showing oscillation; line marked 'D = 0.2 mm' is a simulation with this thickness, here all 5 temperatures practically overlap

11 Fig. 2.6: temperature and current distributions in a simulated B12212 slab (LxWxD = 20x10x4 mm) after 5 sec at 20 mV / 50 Hz AC (cf. Fig 2.4); top: view on bottom x-y layer; bottom: side view on central x-z layer; grey scale represents temperature between 77 K (black) and 92 K (white); arrows represent currents with arrow length and thickness indicating current density

We can conclude that, above a critical thickness of the superconductor, even a small voltage leads to a (partial or complete) quench of a SFCL after a few seconds. While this is not very relevant for quench protection in SFCL itself, which can be backed up by mechanical circuit breakers, a low level of dissipation is inevitable in AC applications due to AC losses. This dissipation might be sufficient to trigger a partial quench during normal, i.e. non-fault operation in SFCL consisting of thick slabs or other bulk superconductors, making them inherently unstable. In the next section we show that also the thermal recovery time after a fault strongly depends on the thickness.

2.3.2 Dynamic Response and Post-Fault Recovery Time

We now consider the behaviour of a SFCL during and after a fault. Our 'model' fault consists of a short high voltage peak, typically 10 ms long and several volts high. An example is shown in Fig. 2.7a. Before and after the fault, a low 'base voltage' of 20 mV/50 Hz AC is applied. (Due to numerical problems, we cannot simulate a truly superconducting state with zero voltage and finite current) Also shown in Fig. 2.7 are the overall current through the SFCL, and the maximum temperatures of the layers along z-axis. The highest temperature is reached in the core, which also is slowest to recover to its equilibrium temperature.

12 fault voltage (10 ms pulse)

'base' voltage; 20 mV AC

0.0-

layer temperatures . centre surface

time [s]

Fig. 2.7: SFCL model calculation using a homogeneous bar with Bi2212 material data. The applied voltage V(t) is the simulation input, in this case using a 50 Hz base voltage of 20 mV, and a sinusoidal pulse (50 Hz, 10 ms) to simulate a fault. The SFCL current l(t), and layer temperatures are calculated as function of time. Also indicated are peak voltage (VP), current (lP) and temperature(TP)

One of the most important characteristics of a SFCL is its recovery time after a fault. We define the recovery time tR as the time at which the maximum temperature is reduced to 1/e of the difference between peak and equilibrium temperatures, as indicated in the inset to Fig. 2.8. Using this definition, we have calculated the recovery time as function of slab thickness, and the result is shown in the main plot of Fig. 2.8. The calculated result lies somewhat in between linear and quadratic dependencies, which are indicated with dashed and dotted lines, respectively.

13 0 8

pulse: 2 V /10 ms T 0.2 3 FCL thickness [mm]

Fig. 2.8: recovery time against slab thickness, in logarithmic scaling squares: simulation data for a Bi2212 slab after a 2 V/10 ms pulse; dashed line: linear dependence; dotted line: quadratic dependence (cf. equation (2) in text); inset: schematic temperature-time dependence with definition of temperatures TP, TF and Tr (cf text)

A quadratic dependence tR(z) is found in one-dimensional heat transfer along z-axis, if the heat source is located at z0 = 0 [14]:

(2) If we identify z0 = 0 with the centre of the SFCL, and z with its surface, we get tR = {DU)2 * c v / c (3) where c is the thermal conductivity along the c-axis. This is the tR(D) dependence plotted as a dotted line in Fig 2.8 (using c v(77K) and C(77K)). However, this model with heat source at origin is obviously is too simplistic to explain our simulation data, because in the simulation the ohmic heating is spread across the entire volume. Nevertheless, the thickness is clearly a crucial factor determining the recovery time of a SFCL.

2.3.3 Observation of 'hot spots' at Contacts and Inhomoaeneities Up to this point we have only discussed SFCL simulations of homogeneous superconductors and metals. However, in our model we can define materials parameters locally for each cell, and thereby study effects of inhomogeneity, contacts and geometric constrictions.

A simple example is shown in Fig. 2.9. Here a defect was defined near the central part of a BI2212 slab. In the defect volume, which is marked by the black rectangles, the critical current density was set to be reduced by 50 %. The current flows around the defect as expected. Perhaps more surprising is the temperature distribution, which is shown in grey colour scale: The hottest point is not afthe defect, but next to it where the current density is highest.

14 Fig. 2.9: current and temperature distribution in a Bi2212 slab with a defect volume where Jc is reduced by 50%; top: view from top, bottom: side view; current density indicated by arrow length and thickness, temperature indicated with grey scale from 77 K (black) to 90 K (white); defect volume indicated with thick black line

The metal contacts of a Bi2212 CRT bar consist of silver sheets which are partially embedded in the superconducting bar. After a heat treatment, this results in very low contact resistance [8]. The open part of the silver sheet can be clamped to current leads. This contact geometry was modelled in our simulation as shown in Fig. 2.10a. The arrows in Fig. 2.10a represent a typical current distribution in normal SFCL (non- quench) mode (VFCL. = 20 mV AC). One can see that the largest part of the injected current is transferred to the superconductor at the end of the bar. This finding coincides with the isothermal finite element calculations carried out by Kursumovic et al. [8].

Fig. 2.10b shows the current and temperature distribution at the onset of a quench (VFCL. = 1 V AC). In this case, a large part of the injected current flows through the entire length of the silver contact. As a result, a 'hot spot' forms in the superconducting material just behind the end of the metal sheet. A second, weaker hot spot is present in the bottleneck of the metal sheet just out side of the Bi2212 bar. This suggests that by careful design of the contacts, hot-spot formation can be controlled and mitigated.

15 silver insulator Bi2212

a)

b)

Fig. 2.10: a) model geometry of the contact area of a BI2212 CRT bar with embedded silver contacts; in side view, L = 40 mm, D = 2.1 mm (height expanded for clarity); white cells :2212, light grey cells: silver; dark grey: insulating gap between bar and metal clamp (cf. text); arrows: typical current distribution in non-fault operation (V = 20 mV AC) b) current and temperature distribution with applied voltage of 1 V AC, after 25 ms at onset of quench; temperature indicated as grey scale with black = 77 K, white = Tmax = 81 K; current density indicated by length and thickness of arrows

2.3.4 Current - Voltage scaling The time development of voltage, current and temperature in a simulation using a CRT model geometry is very similar to that in a homogeneous slab as was discussed above and shown in Fig. 2.7. We can observe in Fig. 2.7 that the peaks in voltage, current and temperature all occur at different times. It is therefore surprising that the peak voltage and peak current are related by a simple exponential law, as is shown in Fig. 2.11 for a CRT bar. This result was confirmed for a number of different parameter sets, and in particular also for inhomogeneous materials, similar to that shown in Fig 2.9. If this exponential form

K>,fcl - Vp0*exp(l p FCL/lp0) (4) is indeed a universal one, which has to be confirmed experimentally, the current- voltage response of a SFCL would be determined by only two parameters Vp0 and lp0, which in turn could be found from only two measured data points. This would significantly simplify design considerations. It should also be noted that equation (4) is valid independent on whether the highest temperature TpFCL is below or above the superconductor's critical temperature Tc. This temperature is usually reached at a later time compared to 1/FCLand /p FCL.

16 However, equation (4) does not predict the destruction point of the SFCL, i.e. the highest voltage which can be applied before the SFCL suffers permanent damage. In our simulation, we defined this point rather arbitrarily, by defining a maximum possible temperature at 300 K. The peak temperature is plotted against peak current in Fig. 2.12. TP{ Vp) first increases gradually up to Tc, where it reaches a plateau, but then increases very rapidly. This functional behaviour can be explained by the variation of boil-off efficiency shown in Fig. 2.2 and discussed in Section 2.3.1. We found that the point at which the destruction temperature is reached depends critically on the input parameters. A more quantitative investigation is planned for the future.

50

2.5 3.0 3.5 4.0 4.5 5.0 5.5 peak current [kA]

Fig. 2.11: peak voltage against peak current reached during a 2 V /10 ms fault in a CRT bar; for different materials parameters as indicated; in semi-logarithmic scaling; full symbols indicate destruction point defined by TP > 300 K (see Fig 2.12)

n isotr. V an is. Z 250 —A— 2-dim o defects

3.0 3.5 4 0 4 5 5.0 peak current [kA]

17 Fig. 2.12: peak temperature against peak current reached during a 2 V / 10 ms fault in a quarter CRT bar; for different materials parameters as indicated and detailed in Fig. 2.11

2.4 Conclusions

Our simulations of normal (low-voltage) operation and quench in SFCL show that a number of parameters have a critical influence on the stability of SFCL: The boil-off efficiency is strongly temperature dependent, which can lead to instabilities. The thickness of a resistive SFCL strongly influences its recovery time, and thick slabs can be thermally unstable even at low voltages. 'Hot spots' form predominantly near defect points, but also close to embedded metal contacts. The maximum voltage and current reached during a short fault are correlated by a simple exponential law.

The results of our simulations so far are mainly of qualitative nature, due to the uncertainty in some of the input parameters, namely the boil-off efficiency function and c-axis critical currents. An urgent task for future work is therefore to obtain more reliable data. This should then allow us to make quantitative predictions of SFCL behaviour, and to test these against experimental results from Bi2212 CRT bars and other materials.

Acknowledgements We thank D. Klaus and P. Barnfather for many valuable discussions, and Dr. J Loram for generously supplying specific heat data of Bi2212. This work was funded by VA TECH T&D and the U.K. Department of Trade and Industry

References

[1] W. Paul, M. Chen, M. Lakner, J. Rhyner, D. Braun, W. Lanz, Physica C 354, 27 (2001) [2] P. Malkin, D. Klaus, IEE Review March 2001 [3] T. Coombs, Physica C 372, 1602 (2002) [4] T. Coombs, ASC 2002 [5] Numerical Recipes in C++, 2nd Edition, 2002 [6] G. Haselden: Cryogenic Fundamentals, Academic Press, London (1971) [7] D. R. Watson, M. Chen and J. E. Evetts, Supercond. Sci. Techn. 8, 311 (1995) [8] A Kursumovic, R. P. Baranowski, B. A. Glowacki and J. E. Evetts, J. Appl. Phys. 86, 1569 (1999) [9] B. A. Glowacki, E. A. Robinson and S. P. Ashcroft, Cryogenics 37, 173 (1997) [10] B. Zeimetz, B.A. Glowacki, Y.S. Cheng, A. Kursumovic, E. Mendoza, X. Obradors, T. Puig, S.X. Dou and J.E. Evetts, Inst. Phys. Conf. Ser. 167, 1033 (2000) [11] M. F. Cronnie and A. Zettl, Phys. Rev. B 41, 10978 (1990) [12] M. F. Cronnie and A. Zettl, Phys. Rev. B 43, 408 (1991) [13] J. W. Loram, personal communication [14] Textbook on thermodynamics, discussing 1d heat flow and erfc(t/tR) solution

18 3 COMPOUND ELEMENT

This involved the design of superconducting elements with integral metallic shunts, buried within the superconducting material. Practical tests, described in Appendix A of this report "Development of Optimised Current Elements for a Resistive Fault Current Limiter - Stage I Programme: Evaluation of First Demonstrator" showed the benefits that these could offer. The primary benefit is that the shunt provides an alternative route for current during quenching, reducing the current which must be carried by the superconductor. If local quenching (which would result in hot spots in the superconductor) occurs, the voltage drop at the local quench forces current into the shunt, reducing the temperature of the superconducting material and potentially preventing bar failure. This work package was to examine the design of such shunts.

Various types of shunt were modelled using the modelling software and bars were manufactured accordingly. Further details are given in Section 8 of this Report.

4 MULTI ELEMENT MODEL

The complete, distribution-voltage FCL will be constructed using a number of superconducting elements connected in series. This work package was to allow elements having different values of Jc to be modelled and was designed into the programme before the final design of the modelling software had been decided, when it was not clear that cells with individually specifiable Jc could be incorporated. The need for this has been satisfied by the core modelling program so a separate calculating module is not required.

19 5 FCL DESIGN

This work package examined the basic design requirements for FCLs to be used in specific types of application. The technical feasibility of each application is examined in Figure 5.1. The desirable characteristics and mode of connection are summarised for the most important application areas in Figures 5.2-5.6

Application ^ Disturbing Technical challenge Distribution Transmission Generation loads

Technically achievable M a rin e MW T ra c tio n C h e m ic a l

Technically possible s te m tie tie A rc A e ro s p a c e B u s - B u s - 5 M W C le a n p o w e r 5 0 M W c o u p le r S y s te m c o u p le r fu rn a c e 9 / Technically unlikely 5 0 0 W e ld in g

Inherent stability: <15kA element electrical & physical design design S ta b ility >15kA

Number of 400A elements:

cost oi current C u rre n t 1600A u

Number of 15kV elements:

costo: voltage v o lta g e 150kV

Thermal mass: Off­ physical size load Cooling system: On­ capacity R e c o v e ry load □ Open

Cooling system: loop capacity, reliability Closed C o o lin g loop ■ Estimated sales (€m)^> 1.5 75 5 100 2.5 50 100 1 10

Figure 5.1: Basic Characteristics of FCLs for each application area with estimates of technical feasibility

5. 1 Distribution Bus-coupler

For this application the FCL is connected at a point where two sections of network, each with an associated infeed, are interconnected. It is assumed that one of the incoming supplies is new and would cause network fault-level limits to be exceeded. Normal FCL current would be low, the device serving primarily to reduce network source impedance and therefore enhance voltage stability. If one incoming supply were to fail, the FCL would have to carry all of the current associated with the load on this side, but because the fault level would now be reduced and again within network limits, the FCL could be bypassed using an additional circuit-breaker (not shown) fitted with appropriate interlocking. These factors make this application attractive for a first implementation (e.g. demonstrator) in a distribution network.

20 5.2 Embedded Generator Connection

For smaller embedded generators, the bus-coupler approach (above) can be used and the operating conditions will be similar to those described in Section 5.1. Faults beyond the bus-coupler FCL will have little impact on the generator, but faults on network directly connected to its terminals will be fed by it and may cause ride- through difficulties depending on the type of machine and associated protection. In this case, a series connection (see Section 5.3) may be used.

5.3 Larger Generator Connection

In the case of larger clusters of small generators, e.g. wing farms or single larger generators e.g. CHP, a series-connected FCL between the generator terminals and the rest of the network brings significant benefits, including:

• Reduces DC contribution, reducing stresses on generator and local circuit-breakers and cabling and potentially shortening protection time-settings • Improving network stability by preventing pole-slipping • Enhancing the chance of riding through network faults by reducing the fault- contribution of the generator and reducing voltage-drop at its terminals

5.4 Hazardous Area Safety

In mines, petrochemical and chemical plant and other environments where flammable or explosive materials may accumulate, electrical faults can trigger destructive events. The energy released at the point of the fault is critical here and reducing the fault-current reduces the risk. For this application, an FCL with a large limiting ratio (prospective current / limited current) is required.

5.5 Interconnection to Fault-prone Network

Where loads sensitive to poor power quality (i.e. voltage dips) such as process machinery are connected it is sometimes necessary to remove links to adjacent network such as that containing overhead lines. While this can reduce the incidence of voltage dips, it makes supply more critical on the availability and quality of the principal infeed. Using an FCL to join the network sections reduces this criticality and isolates the load from the effects of local faults.

21 Product 1: Distribution Bus Coupler

■ 1000A continuous (3 hour) rating • 12kV 3 phase • >33kA prospective fault current • Limit to 33kA peak • Limit to 12kA RMS • Response time <10ms • 200mS fault duration • 5min recovery time • Impedance during quench > 1C1 • Dimensions approx. 1.5m x 1.5m x 2m

Figure 5.2: Distribution bus-coupler application, retrofit connection

Product 2: Embedded generator connection

• 300A continuous rating • 12kV 3 phase

• Limit to 5kA peak New • Limit to 3kA RMS Embedded Generation • Response time <10ms • 100mS fault duration • 1 sec recovery time T

Figure 5.3: Embedded generator application, bus-tie connection preferable for smaller EG

22 Product 3: Large generator connection

• 1000A continuous rating • 25 kV3 phase • Limit to 30kA peak New • Limit to 12kA RMS £■>; Embedded • Response time <5ms • 50ms fault duration • 1sec recovery time I

Figure 5.4: Embedded generator application, series connection preferable for larger EG

Product 4: Hazardous area/safety

• 300A continuous rating • 12kV 3 phase • Limit to 5kA peak • Limit to 3kA RMS • Response time <10ms • 100ms fault duration • 5min recovery time

Figure 5.5: Feed into hazardous area application

23 Product 5: Power quality

30OA continuous rating 12kV 3 phase Limit to 30kA peak Limit to 12kA RMS Response time <10ms BUS Tie 100ms fault duration 1sec recovery time

—r X

Figure 5.6: Link with fault-prone network application

24 6 THERMAL LOSS EVALUATION

One of the most important aspects affecting the cost and volume of a practical FCL are the thermal losses which occur when the limiter is in service, carrying normal load current. These losses determine the amount of energy required to maintain the cryocoolant in the liquid phase in a closed system design, or the frequency of cryocoolant top-up in an open system and a knowledge of the losses can help to determine whether an open or closed cryosystem is required. Losses fall into three distinct categories, each one of which has been examined:

- Ohmic losses in normal resistive conductors, i.e. bushings and connectors.

- Thermal losses due to the ingress of heat from the ambient environment, through the cryoenclosure itself, and through bushing conductors.

- AC (alternating current) losses due to currents induced into the superconducting elements by the magnetic fields generated by the load current. These can be significant and are difficult to predict.

The following abstract from unpublished work by Milan Majoros at Cambridge IRC "An estimate of ac losses in a superconducting fault current limiter (SFCL) made of Bi- 2212 bars prepared by composite reaction texturing (CRT) method" presents full details of the loss estimations.

6.1 Introduction AC loss assessment was done for the following size of the limiter: rated normal current 630 Arms, normal state resistance 1 Ohm. As the elements Bi-2212 superconducting CRT bars 2 cm wide, 3 mm thick and 25 cm long have been considered. We assumed that the limiter at fault conditions reach a temperature of 300 K. The resistivity of a CRT bar at this temperature is 2 pQm. To achieve a resistance of 1 Ohm we need a length of the superconductor 30 m, which is 120 bars. If we assume that in normal conditions the conductor should work at a current of 90% of its critical current Ic , then from the rated current 630 Arms one obtains Ic =990 Apeak.

6.2 AC losses in a single bar

AC losses P (in Watts per unit length l in a single bar can be estimated from an analytical expression using the critical state model of an elliptical cross-section carrying an ac transport current [1]

2 +f- 2 7 = 7^ 1 - lnV1 - f (1) where

p _7_ fll .^4 _ 4 (2) l 2n and

25 3 P_ fie << (3) 1 6n

(Io- peak value of the transport current, f- frequency) It was found experimentally (see e.g. [2], [3]) that equation (1) gives losses in a reasonable agreement with experiment. The results for a CRT bar 30 m long at f=50 Hz are shown in Fig. 6.1. AC losses at rated current 630 Arms are 155.68 W. Considering the surface of the bar, this gives 112.81 W/m2 , a value, which causes an insignificant overheating of the superconductor in liquid nitrogen bath (less than 0.1 K).

6.3 AC Losses in several bars parallel to each other

Because it is required to have a superconducting bar 30 m long it is more practical to use several bars parallel to each other and carrying mutually opposite currents. We considered two basic geometries - an x-array and a z-stack (Fig. 6.2). The losses were calculated in full penetration regime, when the bars carry their critical currents, using the second Norris method [1]. The results obtained for an x-array are shown in Fig. 6.3.

1000 F

Loss (W)

100

10

0.0001 f

10.-5 I I I I 10 100 1000

I (A ) o peak

26 Fig. 6.1: AC losses of Bi-2212 CRT bar 30 m long in dependence on applied ac current at 50 Hz (lc =990 Apeak, rated current 1=891 Apeak=630 Armj.

Fig. 6.2: Arrangements of the bars carrying antiparallel currents ( a) x-array, b) z- stack). {a- bar width, b- bar thickness, Ax, Az- bar separations)

-----«— Degree of coupling (%)(2 bars)(x-array) • Degree of coupling (° o)(4 bars)(x-array)

Ax (mm)

a) b)

Fig. 6.3: Degree of coupling D of x-array bars plotted against the bar separation Ax \r\ mm (a) and against Ax/a (b) (red symbols - 2 bars, blue symbols - 4 bars).

We defined the degree of coupling D as the percentage increase of ac losses with respect to bar separation Ax= (infinitely distant bars, i.e. mutually independent). AC losses in an x-array are higher than those of infinitely separated bars. The reason for this is the fact, that magnetic fields generated by the nearest neighbouring bars carrying opposite currents are of the same direction at their edges. Next to the nearest neighbours, however, generate opposite fields, therefore the ac loss increase is somewhat smaller for higher number of bars at a moderate separation (Fig. 6.3). This is schematically shown in Fig. 6.4. A real field map for 4 bars with Ax= 2 mm is shown in Fig. 6.5.

27 Fig. 6.4: Magnetic field directions at the edges of the bars carrying antiparallel currents - nearest (black lines) and next to the nearest (red line) neighbours effects (schematic - not to scale).

28 Fig. 6.5: A real field map for 4 bars with Ax= 2 mm.

29 • -D (%)(2 bars)(z-stack) • -D (%)(4 barsj(z-stack) • -D (%)(6 bars)(z-stack) • -D (° o)(8 bars)(z-stack)

Az (mm)

------•— -D (%)(2 bars)(z-stack) • -D (°b)(4 bars)(z-stack) • -D (%)(6 bars)(z-stack) ------•— -D (%)(8 bars)(z-stack)

Az/b

a) b) Fig. 6.6: Degree of coupling -D (note that it is negative) of z-stacked bars plotted against the bar separation Jz in mm (a) and against Az/b (b) (red symbols - 2 bars, blue symbols - 4 bars, black symbols - 6 bars, green symbols - 8 bars).

30 Fig. 6.7: Magnetic field directions around the bars carrying antiparallel currents - nearest (black lines) and next to the nearest (red line) neighbours effects (schematic - not to scale). We found that a z-stack has lower losses than those of infinitely separated bars and the degree of coupling D is negative. The results are shown in Fig. 6.6. The reason for this is the fact, that magnetic fields generated by the nearest neighbouring bars carrying opposite currents are of the same direction at their faces but of opposite direction at their edges, which is where the main loss occurs. Next to the nearest neighbours decrease both fields further and ac losses decrease with increasing number of bars (Fig.6.6). This is schematically shown in Fig. 6.7. A real field map for 8 bars with Az=3 mm is shown in Fig. 6.8.

31 b)

32 If we accept as a practical bar separation 9 mm (i.e. 3-times the bar thickness) then ac losses depend on number of the bars as shown in Fig. 6.9. There is a saturation at the number of the bars >12 and the ac loss reduction is approximately 30%. Then ac losses in CRT bars of the total length 30 m arranged as a z-stack of 30 bars, 1 m long each, with 9 mm separation at 50 Hz and at rated current 630 Arms will be 0.7x155.68

35

-D (%)

30 ■

15 ■■

10 -

Number of bars W=108.976 W.

Fig. 6.9: AC loss reduction (degree of coupling -D) in dependence on number of bars arranged as a z-stack with 9 mm separation (i.e. 3 times the bar thickness).

As a last case we analysed ac losses in a circular arrangements of the bars as shown schematically in Fig. 6.10.

33 Fig. 6.10: A circular arrangement of 4 bars. The results for 4 bars are shown in Fig. 6.11 together with those of an x-array and z-

----- •----- D (%)(4 bars )(circular arrangement) ♦ D(%) (4bars)(x- array) —■------D (%)(4 bars)(z-stack)

Ax, Az (mm) stack.

Fig. 6.11: Coupling coefficient of 4 bars arranged circularly and in an x-array and a z- stack (note that in the z-stack the coefficient is negative, i.e. ac losses are lower than those of 4 individual bars infinitely separated from each other).

34 AC losses of a circular arrangement are higher than those of infinitely distant tapes, but they are lower than the losses of an x-array. This is due to the geometry of the magnetic fields - the fields at the edges of 2 nearest neighbouring tapes are of the same direction but they are less perpendicular to the edges in comparison with x- array and are weaker at the outer edges (Fig. 6.12). A real field map of 4 bars circularly arranged at a distance Ax=8 mm is shown in Fig. 6.13.

Fig. 6.12: Magnetic fields around the bars in a circular arrangement.

35 a)

b)

Fig. 6.13: A real field map of 4 bars circularly arranged at the distance Ax=8 mm.

6.4 AC losses in resistive joints

CRT bars are supplied with contacts made of Ag sheets which are partly embedded and sintered into the bar (Fig. 6.14). It is useful to subdivide the contact resistance into 2 parts: the resistance due to current transfer inside the bar between Ag and superconductor (Rc1 ) and the ohmic loss from the Ag sheet and connecting metal outside the bar (Rc2 ). The current transfer inside the bar can be treated as a

36 transmission line problem if the operating current is less than the critical current of the superconductor [4], and if the two sheets at each end of the bar are treated as independent contacts. With the dimensions from Fig. 6.14 and the 'apparent contact resistivity' [4] pc =0.25 pOcm 2 the 'transfer contact resistance' was calculated to be Rc1 =0.3 pQ. At currents l>lc the transfer contact resistance is no longer constant but increases strongly. For example at l=3lc the resistance increases 10 times (estimated from [4]). If the bar is placed in a magnetic field Rc1 increases strongly even at low currents due to the decreased critical current of the superconductor. In a perpendicular field of 50 mT, Rc1 increases by one order of magnitude (estimated from [4]).

Ag Superconductor \ ' 02 ' \ 02 } 3

\ / ------> 15

Fig. 6.14: Cross-section of a CRT bar along its length (numbers in mm).

In the LINK prototype SFCL the bars were connected in two ways: by soldering the open sections of the silver sheets together (contact area 20 x 20 mm2, total thickness cca 2-3 mm) and by clamping them to copper plates (60 x 20 x 2 mm3). The resistances of these ohmic sections were estimated to be 1 pQ for the soldered contacts and 2 pQ for the clamped contacts. The overall contact resistance Rc (in zero magnetic field and at operating currents below lc ) is then Rc =2Rc1 +Rc2 =2.1 pQ per an 'average' contact and the corresponding power loss at rated current 630 Arms is 0.8335 W per contact. For a set of 120 bars the total loss of 120 contacts would be 100 W.

6.5 AC losses in currents leads

We considered a current lead made of copper. From practical experience it is known that copper can carry a current with a current density J=6 A/mm2 in an open air and with J=10 A/mm2 in liquid nitrogen without a significant overheating. For a rated current 630 Arms of the limiter, this requires a cross-section 105 mm2 at room temperature and 63 mm2 in liquid nitrogen temperature. We considered a current lead 1 m long made of a copper plate 10 mm thick of a width of 6.3 mm in liquid nitrogen and of 10.5 mm at room temperature with a linear increase of the width as shown in Fig. 6.15.

37 10.5 mm

6.3 mrr

1000 mm

Fig. 6.15: A copper current lead cross-section 1 metre long of thickness 10 mm (perpendicular to the plane of the drawing) with linearly increasing width.

We considered a linear increase of temperature along the current lead in a cryostat. Then we determined a resistance of the current lead as an integral

(4)

where p(T(x)) is a temperature dependence of the resistivity along the length / of the copper lead and S(x) is its cross-section dependence. As a result we obtained a resistance of one current lead 1 m long R=24.2886 pQ. For 2 such current leads at 630 Arms this gives a loss 19.28 W. A heat influx into liquid nitrogen due to the two current leads will be 14.59 W.

6.6 Cooling power

Total loss of the limiter: 422 W

This determines the refrigeration rate required:- 422 W The minimum power requirement is given by the second law of thermodynamics

where Q0 is the refrigeration rate in watts, and Tamb and T0 are the ambient and operating temperature respectively. For Tamb =300 K and T0=77.3 K we obtain Pmin=1 -216 kW.

38 The required power can be desribed by the efficiency n

P = — P act n min

The efficiency of a LN liquifier is about 0.1, which gives the required cooling power 12.16 kW. If instead of a cooler liquid is supplied the 422W loss will boil off 200 litres per day

References

[1] W. T. Norris, J. Phys. D 3 (1970) 489 [2] M. Majoros, B. A. Glowacki, A. M. Campbell, Z. Han, P. Vase, Applied Superconductivity Vol. 1 (1999) p. 843 (Presented at EUCAS'99, Sitges 14 - 17 September (1999), Spain) [3] M. Majoros, B. A. Glowacki, A. M. Campbell, M. Leghissa, B. Fischer, T. Arndt, Applied Superconductivity Vol. 1 (1999) p. 895 (Presented at EUCAS'99, Sitges 14 - 17 September (1999), Spain) [4] A. Kursumovic et al, J. Appl. Phys. 86 (1999) 1569

39 7 TEST SCHEDULE

When FCL manufacture is conducted on an industrial scale, it is necessary to be able to characterise each superconducting element in a simple, quick and reliable manner. This work package was incorporated to examine possible methods of achieving this and to provide an optimised design of factory test.

7.1 Mounting

The sample is clamped in the holder shown in fig. 7.1.

LlV ------►

Fig.7.1.

The base is tufnol on which are mounted two sets of copper clamps. The silver end contacts are held in these and the transport current is passed through them. A hinged bracket containing two voltage contacts is lowered onto the top. The contacts are spring-loaded pins as used for watch straps. The holder is attached to a stainless steel tube on which it can be lowered into a liquid nitrogen container.

7.2 Cooling

It has been found that the bars are liable to crack due to thermal stresses if cooled too quickly. Therefore the probe must be lowered to a level just above the liquid and left for ten minutes to cool down. It is then lowered further into the liquid.

40 7.3 Electrical Supplies and Data Collection

A block diagram of the test rig is shown in fig. 7.2. Transformer

Sample 250V 100 A

Computer Amplifier

Fig. 7.2

The high current transformer contains an electronic switch in the primary circuit which detects the zero of the waveform and can be set by the computer to pass a fixed number of cycles. A 0.02 Ohm resistor limits the current and is used to measure it. The voltage from the voltage contacts is amplified and also fed into the computer.

7.4 Data Collection and Analysis

The computer is set to pass one complete cycle and then displays a graph of voltage against current. Fig. 7.3. shows the shape expected. The procedure is repeated to ensure there has been no degradation.

Voltage

10 |LlV Current

Fig. 7.3. The Voltage Current Characteristic

41 For the purposes of the fault current limiter the critical current is defined as the current when the voltage reaches 10 microvolts (corresponding to an electric field of about 1 microvolt per centimetre).

The data are then drawn with logarithmic axes, and the gradient gives the value of n in the empirical V-I relation V proportional to I n. The values are then checked to make sure they are within the appropriate limits for the specific application. (It has been found that the relevant limits are extremely sensitive to design variables such as the source impedance so we do not give specific values here. This will be elaborated on in section 8 of this report.

8 OPTIMISED ELEMENT

It is necessary for elements to quench in the presence of fault-current and then to return to the superconducting state without degradation. This work package is required to demonstrate practically that this is achievable and has entailed a considerable investment in test equipment which has been installed in the Materials Science department at the University of Cambridge.

8.1 Introduction

A series of experiments and simulations were carried out in order to find the best design of element. Two software tools were developed. The first is a relatively simple programme which treats the element as a single lumped constant, the second a more sophisticated program which splits the element into a series of non-linear resistors linked by thermally conducting connections and with a variable heat flow to the nitrogen bath.

8.2 V-I Characteristics

The shape of the V-I characteristic is central to the operation of the limiter. Fig.1 shows the voltages obtained from a series of taps along a bar. The curves can be fitted to an equation of the form E

42 Voltage

Current Amps These are typical of a superconductor and it is worth noting the very steep rise in voltage with current, and the very low voltages in the flux flow state. In the normal state they are 100 times higher.

8.3 Single Element Simulation

This uses a single set of parameters applying to the whole element to calculate the rate of rise of temperature and resistance.

Electric Field

Current Density

Fig. 8.2: The V-l characteristics at various temperatures up to 300K

Figure 2 shows the V-l characteristics assumed from 77 K to 300 K. The arrow indicates the path taken when a fixed voltage is applied and the sample warms up to room temperature.

43 80 Temperature

Time msec

Fig. 8.3: Temperature as a function of time at 2 volts

Temperature

Fig. 8.4: Temperature as a function of time at 2.5 volts

The next two figures show how the sample warms up in the first cycle, in each case with a 2 milliohm series resistance. Figure 8.3 is for a 2 volt source and Fig. 8.4 is for 2.5

44 volts. This shows clearly how sensitive the system is to external conditions. For 2 volts the sample warms to 79. 5K while at 2.5 volts it reaches 350K.

8.4 Thermal Design

The finite element model allows us predict the temperature distribution in the element. Fig. 8.5 shows a finite element simulation over five cycles

Fig. 8.5 The temperature at various depths in the element

It can be seen that a temperature difference of 20 degrees builds up across the sample. This is more than enough to crack the sample. Figure 8.6 shows the effect of insulating the sample from the bath. It can be seen that the temperature difference is reduced to an acceptable level.

Fig. 8.6. Temperatures in an insulated bar

45 It is therefore recommended that a layer of insulation be put between the superconductor and the bath. This was confirmed by experiment. An alternative route would be to use much thinner bars, which would decrease the recovery time, but these may be too fragile and there was not enough time to make a set of bars to test this hypothesis.

8.5 Homogeneity

It is very important that the transition to the normal state occurs uniformly, both within an element and between elements. The extreme non-linearity of the process means that it is all too easy for one element, or part of an element, to heat up while the remaining sections remain superconducting. This will cause the element to burn out.

Temperature

800 -

600 -

400 -

200 -

Time Time msec

Fig. 8.7

Figure 8.7 shows the effect of putting two elements in series if there is a difference in Jc of 1 % between them. Fig. 8.8 shows the current through the limiter, and the critical currents of the two elements as a function of time.

46 1500 Current

1000

500 Icl(77)

0

0 5 10 15 20 Time msec

Fig. 8.8 The current as a function of time (red) and the critical currents of the two elements (green and blue)

The external parameters are a source impedance of 2 milliohms and a source voltage of 5 volts. It can be seen there is a difference in temperature of 200K between the elements but both are acting as limiting resistors as can be seen in fig 8.8 where the current drops sharply as both go normal. 2000

Temperature

1500

1000

500

0 5 10 15 20 Time ms

Fig. 8.9 The temperature of the two elements with a 3% difference in Jc

The effect of increasing the Jc difference to 3% is seen in Fig 8.9. One element goes up to well above its melting point while the other remains superconducting.

47 We can therefore say that for these external parameters samples must have a Jc within 1 % of the mean. However this figure is extremely dependent on the source voltage, its impedance, and the heat flow to the environment. It will also depend on the number of elements. It is therefore not possible to give a general prescription for the elements without an analysis of the details of the application of the limiter and its environment.

1500

Temperature

1000

500

0 5 10 15 20 Time ms Similarly the stability is very dependent on the n value. Fig 8.10 shows that the maximum variation allowable in the power of the V-l curve is 0.1.

8.6 Experiments

The experimental apparatus consisted of a high current transformer which could provide 20 volts and 7000A over a few cycles. The current was controlled through a computer and an electronic switch which detected the zeros of current and switched it off after a fixed number of cycles. Five voltage contacts on the sample were used to monitor the voltage at various points.

48 Time s

Fig. 8.11 Voltages for 1.5 cycles

Fig 8.11 shows the voltages for 1.5 cycles. The current was 7000 A. The large dotted curve is the current wave-form and the blue line the total voltage across the limiter. Voltages across the individual contacts are also shown. It can be seen that the waveform is that expected from the Bean model, but that one set of contacts is showing a much larger voltage drop than the others, corresponding to the 10% lower Jc seen in fig.8.1. To avoid the cracking due to temperature gradients a thin layer of insulator was put on the surfaces. A test to five cycles is shown in fig. 8.12. The element survived this and the current was cut down from 7000 A to 3000A as the sample warmed up so this is a promising result. However the imbalance between the voltages at different positions was much greater, as would be expected over longer times. This led to failure in a subsequent test, so there is a good deal more work to be done before we can say that we have a working element. It is clear that a simple shunt resistance will not be sufficient protection to avoid hot-spots. It will be necessary to provide thermal shunting by means of a layer of conductor in close thermal contact to the surface of the sample.

49 ------Voltage across contacts 1-2 ------Voltage across contacts 2-3 Voltage across contacts 3-4 ------Voltage across contacts 4-5 Voltage across shunt — Voltage across 0.18mQ

-10-

Time(s)

Fig. 8.12 The voltages after 5 cycles The current was limited to 3000A, but only one section was driven normal

8.7 Conclusions

1) Conditions can be found both in simulations and experiments in which a superconducting element can survive five cycles and limit the current. This demonstrates that the fundamental function for FCL applications is achievable and sufficiently robust.

2) Thermal strains are a problem both in the initial cooling, (which must not be too fast), and during faults. An insulating layer improves the resistance to thermal shock. However thinner samples may be a better proposition if they are strong enough. In the first implementations, insulating layers will be used; they offer the additional benefit of preventing the penetration of cryocooland into small fissures in the surface of the material, which can raise stresses during quenching.

3) A sample homogeneity of 10% in Jc is barely adequate, 5% will be safer. This is vital in controlling element manufacture and will be used in refining the manufacturing processes.

4) Homogeneity between samples must be better than this since there is no mechanism for transferring heat from one part to another, as can be done within an element.

5) For the conditions assumed elements must have a uniform Jc within 1 % and an n value within 0.1 of the mean (4 in our case). This and the previous point define the repeatability target for element manufacture.

50 6) Elements will require a thermal shunt as well as an electrical one (this can be the same layer of conductor in contact with the surface). Material will be selected to provide optimum thermal as well as electrical performance. 7) The thermal instability of the system was found to be much greater than anticipated at the start of this project, and also extremely sensitive to external factors such as the source impedance. This means that an optimised element can only be designed in conjunction with the complete system, including generator, load and type of fault expected. This is not unexpected and the work to characterise each application area gives some some distinction between the sets of desirable criteria. It will clearly be necessary to expand on this as real applications are approached.

51 9 INSTALLED COST DATA

In order to obtain an idea of the complete costs of an FCL installation, it was necessary to examine the entire scheme including auxiliary circuit-breakers, cabling, additional protection etc.

The costs for adding a superconducting FCL to an existing distribution primary substation have been estimated. Figure 1 shows a typical substation arrangement; the items associated with adding the FCL are highlighted in red.

In a pilot installation it will be necessary to ensure that the series circuit-breaker trips every time the FCL operates. The proposed method utilises two standard voltage transformers, one each side of the FCL. When a voltage drop across the FCL terminals is detected, a tripping pulse is delivered to the series circuit-breaker. This is a complicated and expensive solution; a simpler approach will need to be developed and when available this could be connected alongside the proposed protection arrangement in a trial installation.

It will also be necessary to monitor the cryocoolant pressure to detect an internal fault condition and to trip the series breaker if an abnormal internal pressure rise is detected.

33 kV Bus coupler in °PerU2Psitioti- 12/24MVA 33/11 kV 11kV £ X x :

-CD-I i----- Pressure detector

52 ITEM COST

1 60 metres of 11kV cable £600 2 60 metres of hand dug excavation and backfilling @ £3000 £50/metre 3 Small civil plinth without piles £2000 4 Earthing into existing earth mat and multicores £2000 5 Project management (50 hrs @ £50/hr sell price) £2500 6 Scheme Engineering, multicores, earthing, cabling etc £5000 (100 hrs at £50/hr) 7 Commissioning - 2 days at £450 - £500 / day. £1000 8 Craneage £500 10 Two busbar end boxes £4000 11 SAP time (£500/day, two busbar outages plus permits) £3000 12 Fitter time (two weeks for the busbar end boxes, all £6000 multicore terminations etc. two men @ £300/day each) 13 Voltage transformers at £3000 each £6000 14 Protection equipment £1100 15 Pressure detector £800 16 Misc Items £1000 TOTAL £38500

10 MARKET STUDY - UK

Whilst not included in the Project Work Programme, VA TECH T&D decided to undertake a detailed assessment of the market potential for FCLs in order to furnish essential data for the Business Plan which has been prepared to support investment in FCL development. The market studies are included in Appendix 2.

11 PROJECT SUMMARY

The objectives set for the project and outlined in Part 2, section 1.2 have been addressed. Much progress has been made in providing and refining robust theoretical tools to strengthen our ability to predict how a given design of SFCL will behave under various network conditions. Progress had been delayed by difficulties with sourcing the superconducting elements required to substantiate the work with practical tests, but these have been carried out during a 3 month extension to the programme.

The market potential for FCL products has been carefully examined.

Calculations of thermal losses indicate that reducing these will constitute the main area of difficulty in realising a commercial product. Further practical work in this area is therefore clearly needed and should be included in the next programme of work.

53 Appendix 1

Development of Optimised Current Elements for a Resistive Fault Current Limiter

Stage I Programme: Evaluation of First Demonstrator

1. Executive Summary

The original Workplan is attached as Section 2 below. All components of the agreed programme have now been completed. The current elements have been evaluated against available processing, test data and failure mode, a spreadsheet was used to sort the elements to assess significant correlations. However we do not have information on whether or not some elements were replaced after failure during the SFCL tests.

The main findings were:

(1) The correlation between ratio of the initial and final critical current values and the initial and final furnace treatments was good for elements receiving all heat treatments at ACL

(2) The tendency for elements to fail correlated with the final critical current. The correlation was strong but certainly not overwhelming. For instance 39 out of a total of 93 elements in the rig failed, and although 9 out of 14 elements with the lowest | values failed, only 2 out the 4 elements with the lowest critical currents had failed.

(3) Temperature measurements during 'cool-down' showed that the 'cool down' procedure, unless carried out in a controlled way, can subject the elements to a tensile strain of 0.1% to 0.2%. This could lead to a correlation between critical current and failure quite independent of the thermal stresses introduced by quenching in the SFCL tests.

(4) Almost all samples cooled in a controlled way showed the same Ic (to within 1%) as the final Ic measured by Advanced Ceramics Ltd prior to mounting in the FCL demonstrator.

(5) Samples showed uniformity in variation in IC of ±1% to ±4%. This is better than laboratory samples. However a higher resolution (non-contact) measurement needs to be developed.

(6) Only limited data on quench tests from the LINK programme are available. New experiments were performed and sections 3 and 4 of the Workplan combined.

54 (7) Use of resistive shunt protects the samples from catastrophic failure under condition where the unshunted sample failed, and also promoted faster quenching.

2. Workplan of the Stage I Programme (Materials Science)

(1) Complete the evaluation of all current elements in the demonstrator, study the nature of the failure mechanism and relate as far as possible to the position in the unit, the element critical current (Ic) as measured previously, the processing data recorded by Matthew Hills and Ron Henson and any other information we can discover by studying closely the unit or from people present at the test.

(2) Remove 5 undamaged elements from the unit for testing as follows, measure Ic to compare with the original pre-test Ic. Place a number (e.g. at 10 mm intervals) contacts along the length of the samples to test local voltage drops.

(3) Cut two of these elements in half along their length. Carry out quench tests in the existing ac pulse rigs. We prefer to use uncut elements however since this requires an additional transformer to increase the voltage we cannot guarantee this will be possible within the three months (If this proves possible tests ac quench tests will be made on uncut elements). The quench test used will be discussed with VAT and Archie Campbell.

(4) Provide data from previous measurements on quenching measurements with different shunts.

(5) Submit a short report summarising the results of the Stage I programme within one month of the end of the project.

(Sections 3 and 4 of the Workplan are combined in the report because much of the original work was repeated.)

55 Progress was made against the stages of the Workplan as follows:

(1) Complete evaluation of all current elements

The spreadsheet has been extended to include the most important data on each of the current elements. This so far includes the coding of the intermediate anneals, the number of intermediate anneals and the position of elements in the furnace cradle. Intermediate critical current values prior to optimising anneals have been tabulated, and the excellent overall consistency of the processing treatment has been assessed in terms of the ratio of the critical currents before and after annealing.

Very detailed highly professional processing histories for all current elements have been assembled with the assistance of Matthew Hills of Advanced Ceramics Ltd. For elements fully fabricated at ACL the reproducibility was very good and the correlation between ratio of the initial and final critical current values and the initial and final furnace treatments was very good. For example the elements 85, 88, 89, 102,114, 126, 127 received furnace treatments 'C followed by 'Q' and showed ratios of initial and final Ic values respectively 1.32, 1.28, 1.25, 1.27, 1.35, 1.28. All the data is available in the spreadsheet and further information and analyses can be provided as necessary.

However a major gap in our knowledge relates to the procedures leading up to the failures visible in the elements as returned to Cambridge. There is very little information on the demonstrator test schedules. This is a serious matter and it is essential to have at least an outline of the number and length of the pulses applied, and most importantly on whether tests where continued after the first element failed by replacing or shunting elements or otherwise. We also need to know what tests were applied before the elements started to fail and if the failures that we now see all occurred during one of the pulses. I have tried informally to get information from Dr Andrew Rowley who was present at the test; he reported the following:

" As far as the SFCL tests are concerned, I cannot remember too much about them (in terms of currents used etc.). All I do remember, is that the fault currentwas increased slowly through a number of tests. I think single "pulses" were used i.e. the short circuit current only flowed for 1 or two half cycles. In the tests where some elements failed the cryostat did seem to "jump". This was attributed to "explosive" boiling of the liquid nitrogen at the hotspots

56 where the elements failed. Clearly, this mechanical shock, could have weakened or damaged other elements - which then failed in subsequent tests (after the failed elements had been replaced). I do not believe that the forces between the elements could have been large enough to produce this "jump". I think we all (or almost all of us) agreed that some form of current shunt should have been used in the FCL to prevent this."

(2) Re-testing of undamaged elements

Elements 198 (900 A), 219 (870 A), 85 (840 A), 152 (810 A) and 133 (780 A) were removed from the demonstrator for routine DC testing (M. Hills), they were chosen to provide a full spread of critical currents. As a consequence they will have been 'stressed' to very different levels in terms of the important "fault current to critical current" ratio. They were prepared for re-measurement with a series of voltage taps arranged along the length of each lead at 10mm intervals. On testing these samples we were initially alarmed to see that the majority of the elements had unexpectedly low critical currents (Ic). The critical current was determined using the 1pV criterion.

Additional testing of undamaged elements Additional testing was done on another batch of elements with much the same results. Measured critical currents varied from <1A to the value measured just after element fabrication. This is shown in Fig. 1. Damaged zones were isolated by both voltage increase (lower Ic) and by locally increased nitrogen boiling off at higher currents. It was initially noted that "obvious" cracks were present at ~1/3 and 2/3 of the sample length. Detailed study showed a more complex behaviour. Some samples had rather uniform (±10% variation) Ic along the length. However, the slope, n, of the log(V)- log(I) curve decreased with increasingly "damaged" Ic, approaching n=1 for very small Ic. These results are illustrated in Fig. 2. It is evident that in addition to the visually inspected damage caused during the initial testing in the demonstrator (see Table 1), some invisible damage was occurring as well. However, such large number of elements with lowered critical current led us to conclude that some additional damage was being caused by the retesting procedure for current elements at Cambridge. We are concerned about thermal stresses involved during relatively fast cooling of the elements when immersed in liquid nitrogen. Normally, these elements, when free standing, can withstand repeated thermal shocks from high temperatures, as well in liquid nitrogen. However, the Bi-2212 elements used in the test are fixed with an epoxy compound to a fibre-reinforced plastic support rod of large thermal mass and low thermal conductivity. To investigate the temperature profiles, for various cooling procedures, sensitive thermocouples were placed inside the Bi-2212 elements and the composite support rod. Results are shown in Fig. 3. It can be seen that the elements cooled much quicker than the substrate, resulting in a large transient tensile stress. It was estimated that this stress was around 0.1-0.2 % which is significant for the brittle Bi-2212 ceramic material.

57 A closed polystyrene container was therefore constructed that allows controlled cooling in the cold nitrogen vapour before immersion in liquid nitrogen. The rest of the samples were treated in this manner. Over 75% samples now showed an /c identical within measurement error to the original specification (Fig. 4). Some of these "good" samples were retested by controlled cooling showing excellent repeatability. However, fast cooling of those samples very often (>50%) resulted in permanent damage (lower /c ). The samples were tested for /c uniformity by placing voltage contacts at closely spaced intervals. However, the lack of well-defined voltage contacts (silver paint was used) limited the resolution of the measurements. Also the high currents applied in these samples generally produce a high level of electrical noise reducing the resolution further. We found that 40-50 mm spacing was necessary for repeatable results for ac and dc testing. Some of the results on the /c uniformity along bars is shown in Fig. 5. The deviation from the average /c value ranges from +1 % to +4%. This is much better than on laboratory samples (FCL Progress Report No. 6). However, one has to keep in mind that this is an average over a relatively long length (40-50 mm). The hot spots, when they develop, tend to be on a millimetre (1-2 mm) scale. There is an urgent need to develop a quick method for reliable (non- contact) checking of the sample uniformity on this scale.

S)(£)KmtOO)S®S)00y $2 g? a £ 2 1 000 :

100 :

initial value ] final value

sam pie Fig. 1. Critical current of the FCL elements (bar numbers are indicated as labels) after CRT processing, and after testing with rapid cooling.

58 holder, Fig.

Fig. 3. Temperature (°C)

Thermal

-150 -100 when 2. -50

Variation

- — —

immersed

response

of

the

(and of Cooling

exponent

the

removed

current Time

n

with

from) elements 100 I

E (seconds)

/I — — — ------

-150 -100 59 critical A V O cO

— — — in

- - Bi-2212

near holder Bi-2212 liquid Heating

glued

current the ------

centre

Time

(detached)

glued

holder Bi-2212 Bi-2112 nitrogen. near holder

to

(seconds)

the the

properly

(V~ocl centre surface

glued (detached)

holder

composite

properly

c surface n )

1000 :

100 :

initial value final value reimmer. fast

sample

Fig .4. Critical current of the FCL elements (bar numbers are indicated as labels) after CRT processing, and after testing cooling them slowly.

440 -

420 -

400 -

380 - FCL element: V— 89-bottom A— 220-bottom •— 248-bottom 360 - O—248-top

segment (x50mm)

Fig. 5. Uniformity (along the length) of critical current for samples that were later tested by ac current as well. Note that critical current corresponds roughly to the half of the original value, since samples were cut alongside (see Section 2).

60 (3) Quench tests (including data from the LINK programme and new measurements on shunted samples)

Equipment

For the ac quenching tests a power supply that relies on the mains operated power transformer was used. The power of the transformer is about 5 kW, and the nominal voltage 20 V. Internal resistance of the transformer as about 2 mQ with extra resistance of R< 1 mQ from the current leads. Current was estimated by measuring the voltage across 0.17 mQ "standard" resistor. The necessary equipment was developed previously allowing simulation of the fault like conditions. It is possible to record semi-simultaneously up to 8 differential or 16 single-ended inputs at a total frequency of 300,000 samples/s. Full simulation of short circuit occurring in an electrical distributive network is made, having the superconducting CRT Bi-2212 current element (and the "standard resistor" with the current leads) as the only load. A sophisticated electronic system was developed that allows partial (via variable autotransformer) and full loading of the transformer (with the Bi-2212 current element) in a chosen number of current half-periods (cycles) of 50 Hz. Moreover, sensitive low-noise amplifiers were built for amplification of weak voltage signals both for V-1 and temperature rise measurements. As an illustration a quenching experiment on a sample of lc~100 A, using a ~3kW transformer, is shown in Fig. 6. The full explanation of the results can be found in the LINK Report 2. The short-circuit current "characteristics" during a virtual short circuit of the transformer, used in this study, is shown in Fig 7.

40

30

20 10 £ 0

-10

-20

-30 0.000 0.005 0.010 time (s) time (s' Fig 6. Power dissipation and temperature rise Fig 7. Shape of the current during quenching 1. Current is limited in about 4 variation in the first half cycle for a ms (less than % of the ac cycle). virtually short- circuited transformer used in this study.

2 2nd Annual Report on LINK-FCL Project 2nd Annual Report on LINK-FCL Project

61 Tests

In this report, ac testing was carried out by applying a half ac pulse from the 5kW transformer described above. Slow cooling to liquid nitrogen temperature was performed in order to avoid damaging the elements.

Two current elements (bars No: 220, 248) were cut in half. Detail of the cuts near the common contact of "220" and "248" bars is shown in Fig 8-a. These two (x2) elements were tested both for dc critical current measurement (Fig 5) and high ac current for quenching (i.e. going from superconducting to normal state). Voltage contacts arrangements are shown in Fig. 8-b. In the text and figures that follows "twin" bars are referred to as "top" and "bottom". There is a general rule that local quenching starts at places where the critical current, established by dc measurements (see Fig. 5) is lower. This indeed occurred in the case of the "248-bottom" sample, simultaneously, in segments 1 and 3.

However, even when the / is seemingly constant, as in the case of "220- bottom" and "248-top" (see Fig. 5), elements tend to develop non-uniform quenching. Transition to the "normal" state through the flux flow regime goes relatively uniformly (Figs. 9 and 10) up to about 10/c in these elements. Further current increase results in sudden rather non-uniform quenching that is enhanced by the local temperature increase (see earlier LINK Project Reports), which starts here3 at around 10/. This Bi-2212 bar just survived tests for lower power settings (<90% voltage via variable transformer), but not for the full load. The full power experiment (Fig 9) shows same non-uniform quenching with a part of the sample being blown off (Fig. 11). The destruction occurred in the segment 1, which has lowest critical current. As expected, the 248-bottom element had pronounced quenching in the regions of lower critical current (segments 1 and 3). Part of segment 3 was blown off at full power (Fig. 13).

Another illustration is in Fig. 12 where 220-bottom sample had non-uniform quenching as well. Again, the segment with lowest critical current was quenched first. However, there was an anomaly that segment 3 was quenched as well, although

62 it had the highest critical current. Meanwhile, segments 1 (roughly average critical current) and 4 (same critical current as 2) did not quench. This bar survived a couple of full power tests, and eventually formed a "weak" hot spot in segment 2. Bar 220- top had too low a critical current in segment 1 and was not tested.

However, in all three cases >50% of the sample went into the normal state. The total voltage on elements during quenching approached ~20V. The length of Bi-2212 superconductor is around 25 cm (including embedded contacts).

Fig. 8-a. Detail of the cuts near the Fig. 8-b. Voltage contacts common contact of "220" and "248" arrangements for V-l (dc) and V(l,t), l(t) bars. 3 ac testing.

3 Elements with lower critical current density would be quenched at higher ///c values

63 Fig. E (V/cm)

10. 0.01 0.02

Non-destructive

0.000 0.000 - - Fig. element

elem. 9.

Quenching current

248-top

248-top tests 0.005

under 0.005 time

results

(s)

just

time for ------• ------

0.010

•

pre-quenching 64

•

bar segment-2 • segment-4

segment- el. segment-3 (s)

"248-top".

field: 0.010 ■segment-1 el. segment-4 segment-3 segment-2 1

0.015

-

- field

conditions 0.4 0.6 LU | E 0.015 (up to l~9lc) for bar "248-top".

Fig. 11. Damaged part of the "248-top" element.

; element 220-bottom

i— current'

el. field:

segment-1_ q 4 lu segment-2 segment-3- segment-4

0.000 0.005 0.010 0.015 time (s)

Fig. 12. Quenching results on bar "220-bottom".

65 Fig. 13. Damaged part of the "248- bottom"

66 Data From Shunted Elements Previous data on shunted Bi-2212 elements were lost due to computer failure. Hence, additional experiments were carried out for this task. The samples prepared were of similar geometry as used previously (laboratory scale: roughly 1.5 mm thick, 5 mm wide and 60 mm long).). Two different shunts were used. Shunts were made with ohmic resistances (in both cases) of the order of the normal state (just above critical temperature ~90 K) resistance of the Bi-2212 conductor. The resistive shunt was made of a constantan tape 0.5 mm thick and 5 mm wide. The inductive shunt was a coil made of copper square-wire (1 x2.5mm 2) with 20 turns along the sample length (90 mm in diameter). However, a more detailed design-oriented approach should be made in order to access the full benefits of shunting. Voltage contacts were placed along three equal segments (V1, V2 and V3 in the figures). The current in all cases was measured as total current, i.e. through both Bi-2212 element and the shunt. The power source used was the 3 kW unit with 7V nominal rms voltage. The results are not discussed in great detail, but are presented as a set of graphs with brief comments.

Results for resistive shunt Comparison of the quenching experiment without and with the resistive shunt is shown in Fig. 14. The total "limited" current (through the shunt and the element) is, as expected, higher in the arrangement with the shunt. However, it can be seen that elements with shunts quench faster than without. The origin of this surprising behaviour is not clear at this moment. More experiments should be done with different shunt "geometries" in order to clarify contribution of the magnetic field from an apparently pure resistive shunts. A family of current limiting curves for different voltage settings (on the 3 kW transformer) is shown in Fig. 15. An important finding is that with the resistive shunt quenching can start in less than 2.5 ms (for higher power used) that is almost half the time previously established for our CRT Bi-2212 elements. In Figs. 16 and 17 E-i curves are shown for different applied voltages. This is comparatively shown in Fig. 18. It is important to note that the voltage rise, hence the power dissipated in the Bi-2212 element, is faster for segments with initially higher dissipation. This higher initial dissipation was due locally lower critical current (IC) values in those segments. Fig. 19 illustrates transition to the normal state in two segments where E-i curve becomes linear, i.e. Bi-2212 becomes normal ohmic resistor. It can be noted that the voltage increase (hence power dissipated) in segment V3 was twice as high as in segment V1, while segment V2 stayed in the flux flow regime without quenching.

67 through resistive Fig.

14. l(kA)

Comparison

both, the shunt. 0.000 -o - current, current

m- Note —

Bi-221

without resistive

of that

/

current (7 2

the -150 element

shunt

shunt measured

limiting A)

and 0.005 time

the

results current

shunt (s) 68

with ------

el. (I)

when

field, is

and

the without resistive

connected.

E without sum

(segment 0.010

shunt of

shunt

currents the

V3)

resistive shunt transformer setting ------100% ------80%

...... 70%

------60%

------50%

0.000 0.005 0.010 time (s)

Fig. 15. Family of current curves from current limiting experiments for different voltage settings (with the resistive shunt).

resistive shunt

E at segment ------V1 - V2 ------V3

current

- 0.0

0.000 0.005 0.010 time (s)

Fig. 16. Total current and voltages across different segments at 50% transformer voltage setting, Quenching occurs at "4.5 ms.

69 3 resistive shunt

— V1

E I LLI

0.000 0.005 0.010 time (s)

Fig. 17. Total current and voltages across different segments at 100% transformer voltage setting, for resistively shunted samples. Quenching occurs in less than 2.5 ms.

10 resistive shunt

slope=4 a - - ' A ' § 1

LLI slope=3.5

slope=l 0.1

segment ■ V1 • V2 ▲ V3 0.01 10

C

Fig. 18. Trends in voltage increase at different segments by increasing maximum applied voltage.

70 l(kA) Fig. 19. E-l characteristics during quenching (from the data in Fig. 17).

Results for inductive shunt A new sample was used in experiments with the inductive shunts4 . Figs. 20 and 21 show results on inductively shunted sample. A smaller proportion of this sample was "quenching" than in the previous case. The trend that quenching is faster with the shunt is found here as well (Fig. 20). A comparison between quenching with inductive and resistive shunts that were used here5 is shown in Fig. 22. It can be seen than resistive shunt was more effective than the inductive one. Quenching was almost 59% faster. At the same time ratio of quenching and critical current was much favourable in the case of resistively (///c«7) than inductively (///c«10) shunted Bi-2212 element. However, it should be kept in mind that a perpendicular field (5||c-axis) would give much bigger effect than the transversal one employed here.

4 First sample broke after a series of overloading conditions; hence another sample was used for inductive shunting. 5 Different shunt geometries and resistance/impedance values might give rather different results; ideally the same Bi-2212 sample should be used in all cases.

71 . current,/ E at segment V2 without shunt ■------without shunt- \ — induct, shunt induct, shunt

current - 0.5

- 0.0

0.000 0.005 0.010 time (s)

Fig. 20. Comparison of current limiting results (at 50% transformer voltage) with and without the inductive shunt. It quenched somewhat faster with an inductive shunt.

inductive shunt B\\a-b\\I

current [ E at segment

— - V2

0.000 0.005 0.010 time (s)

Fig. 21. Total current and voltages across different segments at 100% transformer voltage setting.

72 shunt shunt. Fig. l(kA)

22.

used. Quenching

Comparison 0.000

occurs

of

quenching

much

faster 0.005

time with

with

the (s)

the resistive 73

------inductive resistive current, resistive inductive resistive el.

and

field,

0.010 than the

/

shunt shunt E

shunt shunt

inductive the

inductive 2 4 0 3 1

m | E