SIMULATION OF HARMONIC PROBLEMS IN THE HVDC STATION

Pål Otto Eide Roger Fredheim Jan Spånberg Bernt Bergdahl Lars-Erik Juhlin

Norwegian Power Grid Company ABB Power Systems AB , Ludvika, Sweden

Key Words: Modelling and simulation, Harmonic inter- 300 kV action, Core instability, HVDC. Pole 1 Feda SVC Filter 200 MVA BP1 Abstract - Harmonic problems in the Kristiansand HVDC Pole 2 T1 FT1 station have been analysed and identified by simulations. BP2 The most effective tool has been time domain simulation, i.e. 140 MVA, SC simulation by EMTP, EMTDC and analogue HVDC simula- 132 kV HP1 tor. For investigation of harmonic behaviour a substantial 17 kV Krossen Electrode part of the ac system was modelled. The investigated case HP2 Steinsf. of core instability was found to involve transformer core KB1 KB2 KB3 Pole 3 saturation in both ends of the HVDC link. The effectiveness of counteracting the core instability by damping the dc side Fig. 1. The Kristiansand HVDC Station. 50 Hz current has been verified in the HVDC simulator. Third harmonic resonance phenomena have also been simulated. intended to initiate and analyse the oscillations. Simulations using EMTP, EMTDC and an HVDC simulator have been 1. INTRODUCTION performed in order to analyse the events and to understand the mechanism behind and the conditions for the distur- Kristiansand is the Norwegian station of the Skagerrak bances. HVDC transmission between Denmark and Norway. One event that has been investigated is how switching Kristiansand is located in an area with major generation of in of a shunt reactor in Denmark could cause trip of the 300 hydro power in . The local load is up to kV line feeding the Norwegian Kristiansand HVDC converter around 200 MW, of which 80 percent is industrial load. The station via the line earth fault protection. EMTP and simula- capacity of the Skagerrak HVDC transmission is at present tor investigation show that it is due to a core saturation 1040 MW. The original rating of the link is 500 MW (two phenomenon including interaction between the transform- poles). The third pole was commissioned in 1993. Installed ers in both ends of the HVDC link. Although the resonance filter capacity on the Kristiansand 300 kV bus is 346 Mvar, frequency of the dc side, due to the long cable, is well below comprising four filterbanks. Additional 101 Mvar shunt ca- the fundamental, interaction between the dc side, trans- pacitors are installed on the tertiary of a 300/132 kV trans- former saturation on both sides and the impedance of the former T1, see Fig. 1. For reactive power control an SVC of network on both sides, gave a somewhat negative damping ±200 Mvar has been in operation since 1995. The 140 MVA of 50 Hz dc side oscillation. The triggering of the line fault synchronous compensator was not in operation at any of protection was caused by the zero sequence fundamental the events with harmonic problems, which have been simu- current component due to the asymmetrical transformer satu- lated. ration. The dc side consists of three line poles, pole 1 and pole The phenomenon has also been replicated in an HVDC 2 of the same polarity. Each line pole consists of 28 km over- simulator, although the ac networks had to be simulated by head lines on the Norwegian side, 130 km submarine cable somewhat extreme impedances. An additional damping func- and 85 km overhead line on the Danish side. On the Danish tion in the HVDC control has been proven to effectively side pole 1 and pole 2 are connected to the 150 kV system in introduce a positive damping for this core instability phe- the Tjele sub-station while pole 3 is connected to the Tjele nomenon. 400 kV system. The configuration in Tjele is indicated in Fig. One difficulty with this type of simulation is that it re- 4. quires very long computing time. A large system has to be Occasionally significant low order harmonic distortion simulated in detail as the phenomenon involves interaction has been noted in Kristiansand. At some occasions the con- between harmonics both from the HVDC converters and sequence has been unintended tripping of power lines by from transformer saturation. Also, saturation of the trans- zero sequence sensitive earth fault protection. In other cases formers is built up slowly requiring long simulation time, the system transformer T1 has been tripped by overload several seconds. Furthermore, the number of parameters, protection. It has been hard to find the causes as it has not which can be varied, is large and the ac network harmonic been possible to recreate the same disturbances at field tests impedance is not very well known. In addition, transient disturbance recordings from early events only show some line/cable. Besides, for low order harmonics, the converter hundreds of milliseconds, which for these phenomena are control will interact and contribute to the converter harmonic like snap shots, which do not show the dynamics. There- response. Furthermore, both the converter characteristics fore, a technique with a combination of analytic analysis and the converter control are non-linear. and simulation has been used rather than scanning for criti- cal cases. 2.2 Transformer Cores Fundamental current harmonic on the dc side is trans- 2. HARMONIC INTERACTIONS formed to dc current on the ac side, thereby causing trans- former saturation. As the dc current is injected phase-to- An important point regarding harmonic behaviour of an phase without earth mode current, the transformers will be HVDC system is the interaction between the ac and the dc saturated unsymmetrically resulting in different generation side of a converter. The main characteristic for this interac- of harmonic currents in the different phases. As a conse- tion is that the converter transforms a frequency on one quence of transformer saturation, the network will be loaded side to other frequencies on the other side due to the com- by saturation current, which will have both a fundamental mutation process. Also transformer cores have harmonic content and a content of all harmonic orders. The mix of interaction, especially that a dc voltage may result in large harmonics is depending upon the saturation level. An im- harmonic current due to core saturation. Also other compo- portant factor is that the saturation current also is a mix of nents in the system have other harmonic properties than positive sequence , negative sequence, and zero sequence might be expected in a first glance. current. Especially it must be noted that the resulting zero sequence fundamental current can be significant. 2.1 HVDC Converters The interaction between the ac and the dc side has been 2.3 Capacitors Behind Transformers quite well analysed in the literature [1][2][3]. The principle A capacitor behind a transformer may not be seen as a for the frequency conversion between the dc and ac side is capacitor from the converter ac bus. The principle is demon- found in table I. It must be noted that the conversion is strated in Fig. 2, which shows the simplified configuration different for positive sequence harmonics and negative se- and the impedance/frequency plot. The capacitor C is lo- quence harmonic on the ac side. The zero sequence har- cated behind the transformer T. L represents an inductive monics are not interacting at all. load on the secondary side of the transformer. The losses are represented by resistors in the figure. In the frequency TABLE I. Harmonic Transformation. plot, at increased frequency, there is at first a parallel reso-

nance between L and C at frequency f1. At frequency f2 DC Side AC Side Harmonic Order there is a series resonance between T and C in parallel with Harmonic order Pos. sequence Neg. Sequence L. It is noteworthy that a capacitor behind a transformer is 0, DC 1,Fundamental --- seen as a capacitance from the system only at frequencies 1) 1, Fundamental 2 0, DC f1

nDC nAC = nDC + 1 nAC = nDC - 1 former can not be treated in the same way as shunt capacitor banks on the main bus, when estimating the system reso- 1) The dc current is in phase-to-phase mode, but is neither nance frequency. negative nor positive sequence current. Typical examples of capacitors behind transformers are SVC:s and shunt capacitor banks and filter banks on tertiary In principle current harmonics on the dc side are trans- windings. In the case that there is no parallel inductive load, formed to harmonic current on the ac side with the frequen- the curve in Fig. 2 starts at -j∞ at f=0. The total impedance is ∆ cies given by Table I. Besides there are smaller sideband thus capacitive for f

4.2 SVC in Kristiansand 5. IDENTIFICATION OF CORE INSTABILITY The Kristiansand SVC has in the time domain simula- tions been modelled with actual configuration including The main objective of the simulations was the find the transformer, thyristor switched capacitor, thyristor controlled cause and the mechanism that actually caused the unin- reactor and control equipment. The control functions and tended trip. Otherwise it is not possible to be certain that the dynamics have been the same as in the plant. For the fre- real problem, and not a hypothetical one, is countermeasured. quency domain analysis the transformer impedance has been considered for the zero sequence scheme only. 5.1 The Unintended Trips In August 10, 1993, the Skagerrak HVDC station was 4.3 Transformers and AC Filters connected with the Norwegian grid via the Kristiansand - The transformers connected to the 300 kV system have Arendal 300 kV ac line only, due to maintenance work. At been modelled considering connection type, transformer that time only Pole 1 and Pole 2 were in operation in bipolar reactance, and tap changer position. For the time domain operation mode. In conjunction with switching-in of a 400 studies, saturation of transformers has been modelled with kV shunt reactor in the Danish station Tjele, the an external non-linear inductance on the high voltage side Kristiansand-Arendal line was tripped by the earth fault pro- of the transformers. The characteristics have been adapted tection, sensitive for zero sequence current. As the last ac to the transformer saturation characteristics. line tripped, the HVDC converters also tripped, which initi- The 300 kV filters have been modelled in detail. Shunt ated a start of the transient recorders. Besides the quite capacitor banks connected to the tertiary of the 300/132 kV distorted ac voltages, the recordings showed about 35 A 50 transformer T1 have been modelled in their actual configu- Hz current superimposed on the dc current. Some hours ration. Also the ac filters on the Danish side have been later the same day an identical trip of the ac line occurred, modelled in detail. but without relation to any known ac system event. With this background the event was simulated in EMTP. Each 4.4 DC Line mechanism was to some extent isolated in order to get a For the dc line models of the cables and overhead lines better understanding. have been used in order to get an accurate response for the harmonics under investigation. The dc side smoothing re- 5.2 Comparison with Protection Trip Criteria actors have been represented with linear inductances. The One principal question was: Could 35 A 50 Hz current on dc filters in Kristiansand and Tjele have been modelled in the dc side of the converter cause zero sequence current on detail. For the actual investigations only the characteristics the ac side, sufficient for tripping the line protection? Simu- related to low order harmonics are of significance. It may be lation showed that this was the case. The simulated con- mentioned that the first resonance of the dc side alone is at figuration is shown in Fig. 3. A 35 A 50 Hz current was super- around 35 Hz. imposed on a constant dc current. The short circuit power at Kristiansand 300 kV bus was simulated to be 2940 MVA, 4.5 AC System which corresponds to the situation at the trip event. The Modelling of the ac system has been the most difficult impact of the line capacitance was simulated by a shunt part. The problem is not to find resonances, but to find the capacitance in parallel with an inductive impedance so there realistic harmonic impedance of the actual system configu- was no impact on the 50 Hz short circuit power level. A ration. In simulations intended for verification of a principle shunt capacitance of about 400 Mvar gave resonance at the second harmonic. This capacitance corresponds to the shunt 160 Mvar capacitance of about 1200 km 300 kV ac lines. Saturation of 300 kV 150 kV the converter transformers was modelled with their satura- 2940 MVA 2000 MVA tion characteristics.

AC System Equivalent 350 Mvar 2940 MVA DC 50 Hz 400 kV, 4795 MVA

R C Fig. 4. Connection of Shunt Reactor in Tjele. L 350 Mvar 50 Hz DC the Kristiansand converter transformers was about 18 A peak Line Capacitance AC Filter value and increasing with a rate of 10 - 15 A per second. The 50 Hz current superimposed on the Kristiansand dc current Fig. 3. Injection of 50 Hz Current. was about 7 A. Even if the 50 Hz current superimposed on the When the simulation stopped after 1.1 s of 50 Hz current Kristiansand dc current was only 20 % of the value recorded injection, the magnetising current of the converter trans- in the trip, it was concluded that the simulation verified that former was in the order of 900 A, peak value, and still in- shunt reactor switching in Tjele can cause significant trans- creasing. The unsymmetrical transformer saturation caused former saturation in Kristiansand. However, the final ampli- a zero sequence current in the ac system, i. e. the tude may be depending on the network conditions. Also Kristiansand-Arendal line, which is higher than the protec- other simulations verified that saturation of transformers in tion trip level, see Table II. Thus, it was verified that the dc one station could be transferred to transformer saturation in side 35 A 50 Hz current was sufficient for causing the trip of the other station. It also was considered that a significantly the ac line via transformer saturation. It was also concluded longer simulation time was needed for reaching the final that the trip was due to fundamental frequency current rather state. than harmonic current. After the event, the setting of the One observation is that as transformer saturation in both protection has been increased. end of the HVDC link is interacting, the deviation between the frequencies in the two systems may have some impact. TABLE II. AC Line Zero Sequence Current. Especially as the time for build-up seems to be quite long. This aspect was not analysed any further, and generally the Frequency 50 Hz 100 Hz frequency deviation is low.

I0 in the simulation 46 A 58 A

Protection trip level (I0) 25 A 73 A 5.4 Theoretical Analyses A theoretical analysis showed that the rectifier ac net- 5.3 Connection of Shunt Reactor work impedance approximately can be transformed to a 50 Another sub-task was to verify that connection of a 400 Hz dc side impedance in accordance with Equation (1) - (3). kV shunt reactor in Tjele could cause significant 50 Hz cur- The impact from the converter control and saturation are rent on the dc side of the Kristiansand converters. The model disregarded. used for the simulation is shown in Fig. 4. The configuration 9 '' ' corresponds to the actual situation at the trip event. The =++()(1) =55M;15555π 2 11 400/170 kV transformer in Tjele was modelled as an ideal transformer. The switching-in instant of the shunt reactor ''=+++ααXXΦ was selected for obtaining a full dc offset in one phase. In 5=12552 cos( ) cos( 2 5 ) (2) the simulation the dc offset was damped out with a time 22 constant of 3 a 4 seconds. (The shunt reactor switching was ''=+++ααXXΦ performed during commissioning of the breaker ;=12552 cos( ) sin( 2 5 ) (3) synchronising unit.) 22 In the simulation, the dc offset of the 400 kV reactor Where: caused saturation of the converter transformer connected •Z1R is the rectifier ac network impedance seen at 50 Hz to the 170 kV system. The saturation caused a spectrum of from the dc side harmonics, of which the positive sequence second harmonic •Z´2R is the rectifier ac side second harmonic impedance, current resulted in a positive sequence second harmonic transformed to the valve side of the converter trans- voltage. Positive sequence 2nd harmonic voltage on the ac formers side is transformed to 50 Hz voltage on the dc side resulting •R´R is the rectifier ac side dc resistance, transformed to in 50 Hz current. After 5.5 seconds, when the simulation the valve side of the converter transformers stopped, the magnetising current in the Tjele converter trans- •F2R is the phase angle of the rectifier ac side second formers was 700 A, peak value. The magnetising current in harmonic impedance • α is the converter delay angle .5,67,$16$1' 7-(/( • u is the overlap angle. DC side 50 Hz current DC side 50 Hz current For the inverter the impedance is given by the corre- AC network AC network sponding Equations (4) - (6). AC side DC current AC side DC current 9 =++'' ' AC side DC voltage AC side DC voltage =55M;1,,,,2 ()11 (4) π CONTROL CONTROL Transformer saturation Transformer saturation

AC side 100 Hz current AC side 100 Hz current '' XX ,,=++−2 cos(γγ ) cos(Φ , ) (5) 5=1222 2 AC side 100 Hz voltage AC side 100 Hz voltage DC side 50 Hz voltage DC side 50 Hz voltage

'' XX =++−2 cos(γγ ) sin(Φ ) (6) ;=12,,22 2 , Fig. 5. Core Instability Interaction. 5.5 Active Damping Verification in HVDC Simulator Where: Normally the converter control provides significant damp- ing for the fundamental current [9]. However, the built-in •Z1I is the inverter ac network impedance seen at 50 Hz from the dc side, etc. damping for the fundamental currents seems not to be suffi- • γ is the inverter commutation margin angle. cient to prevent core instability in all situations in the Kristiansand HVDC station. In order to specifically counter- In the Skagerrak transmission both α + u/2 and γ + u/2 act the core instability, a special 50 Hz damping control cir- are around 25 degrees. This means that when the phase cuit was developed. The damping control is sensitive for dc angle of the rectifier ac network second harmonic imped- side fundamental current and is fully activated if the 50 Hz ance is above 65 degrees it contributes with a negative damp- current exceeds a pre-set limit. A similar function has been ing. So does the inverter ac network when the phase angle used in another scheme [10]. The function of the damping of the ac side second harmonic impedance is below - 65 control has been verified in an analogue HVDC simulator, degrees, which normally is not the case. Consequently, the set-up for Skagerrak Pole 3. The configuration used is shown most critical situation regarding negative damping of the dc in Fig. 6. side 50 Hz oscillations is when the second harmonic imped- 1760 MVA 0.96 1.2 ance of the rectifier ac network is high and with a quite high 4000 MVA 80 degree p.u. p.u. impedance phase angle, and the second harmonic imped- 83 degree ance of the inverter ac network is low. This means a situa- tion when the rectifier network is rather weak with a reso- Rating nance somewhat above the second harmonic and the in- 436 Mvar 440 MW 290 Mvar verter network is strong. The critical situation corresponds well to the situation at Fig. 6. Simulator Set-up for Core Instability Study. the trip event, see Fig.4. The resonance frequency for Kristiansand ac network, without considering the ac line In order to obtain core instability in the simulator, the capacitance is 145 Hz. The ac line capacitance will reduce conditions have to be rather extreme and worse, compared the resonance frequency somewhat. The inverter ac net- with the ones used for other Skagerrak 3 simulations. The work is quite strong with a short circuit power twice that of capacitance of the Kristiansand 300 kV bus was increased, the rectifier system. Thus, it can be expected that the damp- getting close to second harmonic resonance. Also the knee ing of 50 Hz oscillations was low at the trip event. It also point for transformer saturation was reduced to 0.96 p.u. of must be remembered that even if the 50 Hz impedance damp- nominal ac voltage. However, the circuit was still stable in ing is not negative, the additional driving 50 Hz voltage steady state, even without the damping control. But, at dis- caused by the non-linear loop (DC side 50 Hz current => turbances driving the transformers into saturation, the cir- Transformer saturation => Second harmonic positive se- cuit became unstable and a 50 Hz current on the dc side quence current into the ac network => AC side positive se- grew up. The stimuli used for initiating core instability were: quence second harmonic voltage => DC side 50 Hz voltage) temporary superimposed 50 Hz signal on the dc current or- can cause instability. That is the phenomenon recognised der, increased ac voltage, and ac faults. In all cases the 50 Hz as core instability. The complete interaction is shown in Fig. damping control was found to be very effective. 5. At a superimposed temporary 50 Hz signal on the dc More about theoretical analyses can be found in the current order, an injection duration of 200 ms was sufficient literature [1][6][7][8]. The article [8] by Burton et al. treats for triggering the core instability. However, it took about 10 the transformation of impedances from the ac side to the dc second until the 50 Hz dc side current started to grow to a side and states criteria for instability. However, the impact of significant value. After about additional 2 seconds, the final transformer saturation in both stations is not considered. value of 0.14 p.u. was reached, nominal dc current as p.u. base. With the damping control activated no core saturation tion that the conditions for core instability are fulfilled. occurred. The phenomena involve transformer saturation in both At increased ac voltage it took 80 - 85 seconds until the Kristiansand and Tjele. 50 Hz dc current started to grow to a significant value. How- • In order to obtain core instability in the HVDC simula- ever, the time to reach the final value from the start of the tor with actual control and dc configuration, it has been noticeable growth was only a few seconds. The damping necessary to use extreme parameters for network controller prevented core instability even when the voltage conditions and transformer saturation characteristic. A was significantly higher than the voltage needed for core specific damping control for dc side 50 Hz current has instability with the damping control inactivated. been verified as an efficient countermeasure for core At an ac system fault the core instability situation was instability. reached immediately. Fig. 7 shows the dc voltage and dc • Similar tripping is realised only a few times for the current at a 70 ms three phase fault applied in Kristiansand. Skagerrak scheme in conjunction with unusual network The damping control is deactivated and consequently at situations. It has not been possible to replicate the real fault clearing the dc current is superimposed with a 50 Hz event even with the best estimate of the conditions for current. The amplitude is considered to be limited by the event. It seems there are too many uncertain nonlinearities in the converters and converter control. The parameters involved. time scaling is in seconds and the amplitude scaling is in p.u. of the nominal value. With the damping control the core 6. THIRD HARMONIC RESONANCE instability is eliminated, see Fig. 8. The superimposed 50 Hz current is damped out. At the fault clearing, the transformers One of the problems to identify harmonic problems in become somewhat saturated. Therefore, disturbance in dc the Kristiansand HVDC station is that there seems to be current can be seen until the saturation ceases. However, more than one mechanism for the harmonic resonance phe- there is no build up and no subsequent core instability. nomena. Based on recording from another trip event simula- tion has been performed for simulation of a third harmonic DC VOLTAGE problem too.

6.1 The Trip Events From September 1995 up to July 1996 the transformer T1

0510feeding the 132 kV system in Kristiansand was tripped with- DC CURRENT out any identified fault or fundamental current overloading. As the trips did not affect the HVDC link, the HVDC tran- sient fault recorders did not start. Only in one case a tran- sient recorder for the Kristiansand 300 kV system started. In that case about 120 A third harmonic zero sequence current could be found in 300 kV side of the transformer T1. As also Fig. 7. AC fault, without Damping Control. the overcurrent protection of the capacitor banks connected to the tertiary winding of transformer T1 had started, it is

DC VOLTAGE likely that the third harmonic positive/negative sequence current was significant too. Recording of 300 kV voltages and line currents showed a significant higher third harmonic distortion in phase T than in the other phases. The trans- former was tripped by the transformer overcurrent protec- 0 5 10 tion in phase T. DC CURRENT In none of the other cases the capacitor banks on T1 tertiary winding were connected. Another trip was also caused by the transformer T1 overcurrent protection in the same phase. The cause for the third trip is not known. All three trips occurred during evening hours when the load Fig. 8. AC fault, with Damping Control was rapidly reduced and consequently the number of hydro generators in operation was reduced with a fairly high rate. 5.6 Conclusive Results 6.2 Third Harmonic Simulation The conclusive results from these investigations are: For simulation and analysis of the expected third har- • The 50 Hz current superimposed on the dc current in monic resonance problem a substantial part of the ac net- the transient recordings from the trip is a very strong work in the vicinity of Kristiansand was modelled in EMTDC. evidence for core instability. The configuration is shown in Fig. 9. Besides the complete • All links from connection of the reactor in Tjele up to HVDC system with the three separate poles with complete the trip of the Kristiansand-Arendal 300 kV ac line have control and the Kristiansand SVC were modelled in detail. been verified by digital simulation under the assump- The Danish station Tjele was modelled in a manner similar to It was also found that the resonance peak for the zero the Kristiansand station with the ac network represented by sequence system in Kristiansand was quite high and sharp. an equivalent circuit. Due to the size of the modelled system An injected current of 1 A caused a total current of 33 A in and the time needed for resonance build up, each simulation the four 300 kV bus filters together. For the positive se- shot took a fairly long time. With a 433 MHz DEC Alpha quence system the corresponding gain was 25. The gain for workstation the initiation up to steady state took about one transformer current in the simulation was significantly lower. hour and 1 s simulated time took 5-10 minutes. With the modelled configuration the resonance frequencies were always well above 150 Hz. The reason seems to be the Steinsfoss Kristiansand capacitance of the surrounding ac network. Another critical

BP BP HP HP ~ G 1 1 2 1 2 input parameter was capacitance behind transformer imped- T 1 Y Y ance, see clause 2.3 above. However, the resonance fre- D FT Krossen Kb 1 1 Y Y Y Y Y Y Y Y quency in the simulations was found to be in accordance D Y D Y D Y D Y D Kb 2 HVDC HVDC HVDC SVC with the impedance of the subsystems modelled. G Pol 1 Pol 2 Pol 3 ~ 1 Kb 3 During all the three events the load was rapidly chang- Arendal ing and so was the impedance. Thus a large number of con-

C s Porsgrunn C s figurations were scanned. The likelihood that one of these C sh configurations gave resonance very close to 150 Hz was T 1 T2 Y Y Lista Y Y ~ G 2 D D considered to be quite high. Due to the long computer cal- Kb G Kb P 4 3 1 Z Z 1 Z Feda Q ~ 1 Q 2 culation time it was not practically possible to scan a large Kb5 Kb2 P Z 1 Z Q 1 Q 2 number of configurations, especially as the impact of the Kb Kb3 6 capacitance of the connected system was rather uncertain. Solhom

Y Y Y Y Y 6.4 Countermeasures D T 1 D T 2 ~ G 3 D T 12 D T34 D T 5 ~ G 4 The reason for the third harmonic problem was found to ~ G ~ G Z ~ G ~ G ~ G Z 1 2 1 2 3 be network tuning to third harmonic, combined with low damping, especially in the zero sequence system. As it was Stokkeland Kjelland Ana- considered practically impossible to avoid network tuning P P T T Z 2 Z 12 34 P , Q Q Y G 1 Y 1 1 2 Q Y D T to the third harmonic, a more general approach was applied. Y ~ Y Z 1 D D Kb Kb 1 3 At start of a protection related to high harmonic current, a ~ G 1 Kb Kb 2 4 filter bank is temporarily tripped by the reactive power con- trol system. The temporary trip of the filter will detune the Fig. 9. Configuration for Resonance Simulation. system resonance during a time considered necessary for the system to be detuned. As the configuration in Kristiansand was somewhat dif- Possible additional specific damping measures, if needed, ferent for the three trip events, the main objective of the are under consideration. The evaluation also includes con- simulations was to identify the mechanisms behind the trips. sideration of possible interaction with the new HVDC con- For analysis of the frequency response of both the com- verter stations under planning in the same area. plete system and different components a small “white noise” current was injected in the system and the response was 6.5 Conclusive Results analysed with FFT. Injection of 1 A of each multiple of five The conclusive results are: harmonic from 5 to 500 Hz (5, 10, 15 ... 500 Hz), each in differ- • There is no evidence that the third harmonic resonance ent phase position was found to give sufficient amplitude problem is related to core instability. for analysing, without disturbing the operation of the con- • All HVDC converters and FACTS devices contribute to verters and SVC too much. Injection was performed alterna- the positive sequence third harmonic due to network tively with positive, negative and zero sequence current. asymmetry. For each sub-system the obtained impedance/frequency plot • Detuning of the system by temporary filter trip is a was compared with the expected values, based upon input general countermeasure. parameters, for verification of the models. • No equipment failures that can explain the third harmonic problems have been identified. 6.3 Main Findings As the ac lines are not fully transposed, high loading of 7. CONCLUSIONS the lines causes asymmetrical voltage in the station. As a consequence of the negative sequence voltage the HVDC Recordings from a disturbance are of utmost importance converters and SVC in Kristiansand responded with posi- for being able to replicate the event and to understand the tive sequence third harmonic current. In the simulation, the mechanism so that proper countermeasures can be taken for total third harmonic current was about 25 A with the con- avoiding recurrence of the disturbance. figuration in Fig. 9. In a more critical configuration, with the Time domain simulation is the only tool covering the Feda - Kristiansand 300 kV line open, the simulation gave a complete interaction, including the non-linearities and the total generation of about 50 A. transformations in HVDC converters, FACTS, transformer saturation, and transmission lines. The drawback with time 8. REFERENCES domain simulation is that only one configuration can be analysed in each run. [1] E.V. Person, “Calculation of Transfer Functions in Grid- By adjusting input parameters it is quite easy to obtain controlled Converter System, with Special Reference to any type of instability and resonance phenomena in digital HVDC Transmissions”, IEE Proceedings, Vol. 117, No 5, simulations. The problem is to make sure that the studied May 1970, pp. 989-997. problem is the real one and not a hypothetical one. When [2] M.P. Bahrman K.J. Peterson, R.H. Lasseter, “DC system modelling a large system, many input parameters are not resonance analysis”, IEEE Transactions on Power well known and it is therefore difficult to identify the con- Delivery, Vol. PWRD-2, No. 1, January 1987, pp 156-164. figuration and set of input parameter for exact replica- [3] X. Jiang and A.M. Gole, “A Frequency Scanning tion of the real event. Due to the long simulation time for Method for the Identification of Harmonic Instabilities complex circuits, even with today’s computer capacity, there in HVDC Systems”, IEEE Transactions on Power are practical limitations for the number of input parameter Delivery, Vol. 10, No. 4, October 1995, pp 1875-1881. variations that can be simulated. [4] A.M. Gole, O.B. Nayak, T.S. Sidhu, and M.S. Sachdev, With recordings from the actual disturbance event as a “A Graphical Electromagnetic Simulation Laboratory starting point, it has been possible to identify and verify for Power System Engineering Programs”, IEEE two different types of harmonic problems in Kristiansand, Transactions on Power Systems, Vol. 11, No. 2, May i.e. core instability and third harmonic resonance. The time 1996, pp. 599-606. domain simulation has been combined with analytic analy- [5] J. Arrillaga et al, Cigré Joint Task Force 36.05.02/ sis and frequency domain simulation. A common factor is 14.03.03, “AC System Modelling for AC Filter Design - the low damping of the Kristiansand station for low order An Overview of Impedance Modelling”, Electra No. harmonics, especially in the zero sequence system. Also 164, February 1996, pp. 132-151. countermeasures have been analysed. [6] R. Yacamini, J.C. de Oliveira, “Instability in HVDC For analysis of low order harmonic behaviour a signifi- schemes at low-order integer harmonics”, IEE Proceed- cant part of the connected ac system has to be modelled. ings, Vol. 127, No. 3, May 1980, pp. 179-188. Even with the substantial size of the model used, the impact [7] S. Chen, A.R. Wood, J. Arrillaga, “HVDC converter of the surrounding ac system may not have been properly transformer core saturation instability: a frequency modelled for correct resonance frequency. The missing fac- domain analysis”, IEE Proceedings - Generation, tors seem to be the impact of the capacitance of surround- Transmission, Distribution, Vol. 143, No. 1, January ing ac lines. 1996, pp. 75-81. The used models have as far as practically possible been [8] R.S. Burton, C.F. Fuchshuber, D.A. Woodford, A.M. verified, and it is confirmed that EMTDC is a reliable tool for Gole, “Prediction of Core Saturation Instability at an the performed harmonic simulation. However, the capacity HVDC Converter”, IEEE Transactions on Power of presently available computers makes this type of analy- Delivery, Vol. 11, No. 4, October 1996, pp. 1961-1969. sis very time consuming. [9] Å. Ekström, E. Borg, “D.C. Line Overvoltage in Connec- Regarding core instability, it has been found that core tion with Injection of Alternating Voltage into the saturation in the transformers in both ends of the HVDC Line”, Electra No. 53, July 1977, pp. 21-30. transmission interact, although the saturation only caused [10] E.A. Hammad, “Analysis of Second Harmonic problems in Kristiansand. With some extreme parameters, Instability for the Chateauguay HVDC/SVC Scheme”, core instability has also been re-simulated in an analogue IEEE Transactions on Power Delivery, Vol. 7, No. 1, HVDC simulator using the Skagerrak III set-up. The effec- January 1992, pp. 410-415. tiveness of a damping circuit has been demonstrated. One interesting aspect is why the problems only occur in Kristiansand but not in Tjele. There seems to be an un- known factor in Kristiansand causing unfavourable condi- tions resulting in the harmonic problems there in some quite rare situations. Furthermore, since commissioning of the first pole in 1976, the harmonic problems have occasionally oc- curred only a few times. Besides, the experience is that the studies and investigations, which in accordance with the normal praxis are performed during design and commission- ing of an HVDC scheme, normally ensures stable and reli- able operation over the complete operation range. The on- going recording in Kristiansand is expected to fully reveal the secret of unknown factor in Kristiansand.