HVDC MULTI-INFEED PERFORMANCE

Transcription

1 MULTI-INFEED PERFORMANCE Göran Andersson, Paulo Fischer de Toledo and Göte Liss ABB Power Systems S Ludvika, Sweden 1 Abstract The number of power system configurations with two or more links feeding power to different points in the same network area is increasing as the total number of links in the world increases. This gives rise to a greater interest in studying the performance of such applications. A very special and extreme case of multi-infeed is the asynchronous interconnection of the Scandinavian (the NORDEL) network to Continental Europe, today by 6 cable poles, but probably using more poles in the future. A number of technical aspects specifically related to multiinfeed configurations must be considered: - Is coordination of recovery control needed? - How can DC power modulation for stabilization be established? - Infeed by two or more DC transmissions to one and the same network area often means that a large amount of power is fed into a network that is relatively weak, thus risking voltage instability. - The consequence of feeding a large amount of power into a weak network must be considered. - How effective is the electrical separation between inverters in reducing the risk of mutual commutation failures? 2 Introduction In a few places in the world, two or more transmission systems terminate electrically close to each other in the same network. The most extreme example of this kind of transmission system is of course multiple bipole infeed, not only to the same network but to the same place in the network. This can be found in Brazil, where the Itaipu transmission constitutes a double bipole scheme transmitting 6300 MW in all from the Itaipu power station to Ibiuna outside São Paulo. It is quite clear that there has to be mutual interference between the two inverters in this case and an network disturbance which causes commutation failures in one bipole affects the other bipole in the same way. The situation is of course the same for a normal bipole. However, if the DC links terminate at separate places, i.e., if the inverters are electrically separated by impedances in the network, the interaction will always be less than in the previous case. This is what we normally mean by multi-infeed, and problems are then reduced compared with multiple bipoles ending up at the same place. 3 The Northern Europe Cable Schemes The most obvious example of multi-infeed networks are to be found in Scandinavia and between Scandinavia and Continental Europe; and more are to come. U.K. NETHER- LANDS NORWAY DENMARK GERMANY SWEDEN POLAND FINLAND Figure 1. Existing and planned links in Northern Europe. Fig. 1 shows a map of the area mentioned above with the existing links, as well as those which are to come in the near future. It is clear that the interesting geographical areas from the point of view of multi-infeed are the north of the Danish peninsula, Southern Norway and the European continent. The five cables terminating in the north of Jutland, namely Skagerrak I, II and III from Norway, and Konti-Skan I and II from Sweden, have been in operation for a number of years. The two cables crossing each other in the Baltic Sea are the Kontek project from the island of Zealand in Denmark to Rostock in the Germany and the Baltic Cable from the far south of Sweden to the Lübeck area, also in Germany. Both have come into operation recently. Finally, there are three extremely long cables in Fig. 1 running from Norway to the north-west of Germany and the Nederlands. This scheme will be implemented in the near future. As can be seen in Fig. 2, two of the Skagerrak cables are connected to the 150 network, while the third terminates in the 400 network. Thus there is a transformer with its impedance between the two converter stations. Already here we can see a multi-infeed configuration. Further, the two Konti-Skan cables are connected to the two different systems in Denmark with a transformer in between. The Konti-Skan terminals and the Skagerrak terminals are interconnected by lines with their impedances.

2 The Skagerrak/Konti-Skan system is obviously a sophisticated multi-infeed arrangement which was planned both in the simulator and through computer simulation using a transient stability program. NORWAY this occurence, an existing telecommunication link was used to send an order from one station to the other to increase the commutation margin reference just before an intentional switching operation was to be carried out. The detrimental consequence of the strong electrical coupling between the two converter sites was reduced significantly by this measure. This solution will not help where there is protective switching of apparatuses but such cases are of course much rarer than intentionally controlled switchings and can be accepted. Kristiansand SWEDEN 5 Simulator Study for Skagerrak I, II & III and Konti-Skan II Skagerrak DENMARK (Jutland) Tjele 150 Vester Hassing Konti-skan to Germany Gothenburg 400 An simulation study of the dynamic performance of the multi-infeed scheme with simultaneous operation of the Skagerrak III, Skagerrak I & II and Konti-Skan II transmissions has been performed. The main purpose of the study was to determine any restriction that had to be imposed for the configuration. An equivalent of the system model representing the Tjele and Vester Hassing area was derived in the transient stability program. This includes the main busses and synchronous machine model as shown in Fig. 3 and was developed with the criterion of providing similarity to the results of a complete network model. Figure 2. Detail map for the Skagerrak/Konti Skan arrangements. 4 Characteristics of the Skagerrak/Konti-Skan Systems In conjunction with the fundamental frequency overvoltage study for this configuration, performed with the transient stability program, it was found that, in extreme conditions, the Jutland network suffered from voltage instability if a large amount of power was fed into the Danish peninsula. The fault case when this actually occured was a fault on and loss of the 400 line from Tjele, the converter site for the Skagerrak cables, southwards to Germany. This caused a transition from a relative strong network on the peninsula to a very weak one. It would have been necessary to reduce the total amount of transmitted power on the DC line in order to achieve a successful recovery after the fault. Although the configuration discussed here is a typical multi-infeed arrangement, the phenomenon mentioned whould, of course, also have happened if all the DC power was delivered to the same point in the Danish network. There has never been any reason to study this eventuality. The fault case is question was also studied in the simulator and the same behaviour was found. However, the critical level of transmitted power was slightly higher here, %, than in the computer study. The fact that the four infeed points are separated by impedances makes the situation rather less critical than if the power had been fed into one and the same point. The last section of this report deals with some theoretical aspects concerning how the risk of voltage instability depends on the degree of electrical coupling between the inverters in a multi-infeed scheme. A more typical multi-infeed phenomenon experienced in this scheme is the mutual electrical coupling of disturbances between the converter sites. It was observed that switching a filter or transformer in the Konti-Skan converter station could cause a commutation failure in the Skagerrak station. It should be noted that reinsertion resistors were not installed in the switches. To avoid Figure 3. A five-bus equivalent of the Jutland network for the simulator. The results of the study showed that the systems can be safely operated with no restrictions where there is a normal network configuration. Fig. 4 shows a typical recovery of each transmission system after a three-phase solid fault in the Tjele 400 network. In this model the converters in Norway and Sweden are rectifiers, and Tjele and Vester Hassing are the inverters. Nominal power transmission was assumed for this case. In the case where the inverter network is assumed to be very weak the study has indicated high risk of repeated voltage collapses and commutation failures. Operating restrictions were imposed on this configuration. 6 The Baltic Cable/ Kontek/Great Belt Schemes Baltic Cable and Kontek are two DC cables crossing each other in the south of the Baltic Sea. Baltic Cable has one converter station in the south of Sweden and the other close to Lübeck in the former West Germany. Kontek runs between the island of Zealand in Denmark and Rostock in what was previously East

3 Figure 4. Simulator recordings of the recovery after a solid 3- phase fault at the 400 bus in Tjele. Germany. Zealand is connected to Sweden by a number of cables and is, accordingly, synchronous with Sweden, Norway and Finland. The interconnection between Lübeck and Rostock is normally a 400 line and the two places are not strongly separated electrically. Here again we see a multi-infeed system; however, less complex than the previous one. A project which is probably going to be implemented in the near future is the Great Belt link connecting Zealand to the other large Danish island, Fyn. The latter is synchronous with the Jutland network and the converter station here will have a powerful connection, via a 400 line, to Northern Germany. 7 Simulator Study for the Baltic Cable/ Kontek/Great Belt Schemes This study was carried out in order to identify and investigate possible system interactions between the system and the three links which connect the south of Denmark, southern Sweden, northern Germany and the Jutland peninsula. Special attention was given to the risk of commutation failures in each of the systems. A demonstration of the performance during recovery of system faults was made and verification was also carried out of the overall stability of the system in some special weakened network configurations. A suitable network equivalent, including 23 buses and 8 generators provided by the concerned utilities, was used and is shown in Figs. 5 and 6. The models included the actual plant data for the Baltic Cable and Kontek projects. Regarding the Great Belt link, due to the similarity of this project to the Kontek transmission, the same data were used, but with the appropriate figure for the cable length. An important general result of the study was that the system has proved to have an overall acceptable performance. Both Kontek and Baltic Cable performed as expected during system disturbances. Thus the basic control and parameters were found to be adequate in these projects. With respect to the Great Belt link, some adjustments of the controls had to be made in order to ensure safer operation. In this project the DC cable is shorter than those of the Kontek and Baltic Cable links. Due to this characteristic, the system presented a rather fast recovery time from system faults and also the capability of transmitting a significant amount of power during asymmetric faults. As a result of these observations, some necessary changes to the settings of the current control amplifier and voltage dependent current limiter were introduced. In general, the three systems presented a number of mutual interactions. In quite a few configurations that were elec- Figure 5. A 13 bus equivalent for the synchronous Nordel network with Baltic Cable, Kontek and the Great Belt project. trically closely connected, an fault in the vicinity of one inverter bus can produce commutation failures in the neighbouring transmission link. It was also observed in an another load flow configuration that a trip of one link can produce commutation failure in the other ones. This is a consequence of a sudden load rejection with an accompanying voltage phase shift in the network which can make the converters lose their commutation margin. Figure 6. A 10 bus equivalent for the synchronous Northern UCPT network with the same projects as in Figure 4. In a particular network configuration, the result of the study showed that during the recovery of the systems from faults in a weak system, with a consequential split of a bus in the Nordel area, the systems displayed a tendency towards dynamic instability, see Fig.7. In this configuration, where all three transmission systems feed rated power into the Nordel equivalent (power direction: north/east), the fault case proved to be very severe. It has consequently been verified that a damping controller function,

4 tion will very briefly review these concepts and propose an extension of these that is applicable to multi-infeed systems. Figure 7. Simulator recordings for the Baltic Cable / KONTEK / Great Belt study. similar to that existing in the Kontek link, if implemented in the Great Belt link, would have a positive impact in increasing the damping of the generators connected in the Nordel area. Another alternative and possible complementary solution considered in the study was to introduce some kind of run back of DC power in the Great Belt controls. There is also one more network configuration where an asymmetric fault close to the Great Belt rectifier converter bus (ELS bus) may cause commutation failure in the Kontek link during the fault, or during the recovery of the system, when the fault is cleared. The configuration used in this fault case assumes rated power flow from the Nordel area, which means that all three links have a common rectifier equivalent. This result again shows an example of a strong interaction between the systems. In this particular fault case, it has been observed that the Great Belt link is well able to operate and transmit power from an unsymmetrical network. This affects the receiving network in that it accepts a significant injection of harmonics. To avoid this interaction, i.e., to avoid an fault being transmitted between areas through the links, the power flow through the Great Belt link was limited during faults. Further control changes were also made to avoid power overshoots during the clearance of the fault. The basic conclusion from this study verification was that the Nordel system presents a satisfactory performance during transients, although there is some interaction between the systems. With appropriate control tuning and control strategy implemented in the Great Belt link, the obtained interactions will be minimized to an acceptable level without severe restrictions. In the Kontek and Baltic Cable links, there will be no special change in the controls. The performance here is adequate. 8 Analysis of Voltage/Power Stability As the complexity of the interconnected /DC power systems increases, there is a need for a systematic approach to analyze the various stability limits of the system. One stability limit that is of great importance, particularly when the system is weak as compared with the rating of the converters, is the voltage/power stability limit. For a single converter feeding into an power system, the risk of voltage/power instability has been extensively studied over the years, and a comprehensive overview of this topic is given in [1]. One of the main results presented in [1] consists of the analytical criteria for voltage/power stability represented by the Effective Short-Circuit Ratio (ESCR), or the Voltage Sensitivity Factor (VSF). This sec- 8.1 Analysis of Single-Infeed System In the analysis of the voltage/power stability in the considered time frame, the system can be represented by an impedance determined by the short circuit capacity. A motivation for and discussion of the validity of this model can be found in [1]. It can be shown that when the system short circuit capacity is lowered, i.e., the equivalent impedance is increased, the system ultimately becomes unstable. This stability limit is dependent on parameters of the converters, e.g. commutation reactance and commutation margin, but also to a large extent on the control strategy. It is found that constant power control with constant commutation margin control encounters the stability limit first, i.e., for the highest ESCR value, but that this limit can be extended to lower ESCR values by use of more sophisticated controls. However, these more sophisticated controls require operation with higher commutation margins and the converters, converter transformer and reactive comensation must be upgraded accordingly, which involves extra cost. Typical values where instability is reached for standard controls are for ESCR values of around 1.5 (SCR» 2) and lower. A useful measure of the voltage/power stability is given by the Voltage Sensitivity Factor (VSF) defined by: VSF = U Q where Q is a small reactive power injection at the commutation bus and U the corresponding change in the magnitude of the voltage at the commutation bus. A small and positive value of VSF indicates stable operation and an increasing value of VSF corresponds to a decreasing stability margin. When VSF becomes infinite, transition to unstable operation occurs, and a negative value of VSF is found in the unstable region. The fact that VSF goes to infinity at the stability limit is unattractive from an analytical point of view and therefore 1/VSF is often studied instead. 8.2 Analysis of Multi-Infeed Systems The above analysis will now be extended to multi-infeed systems, as reported in [2]. To demonstrate the method, a system with two converter stations according to Fig. 8 will be used, but the method can be applied to systems with any number of converter stations. In this paper, only the main results of the method are given and the reader is referred to [2] for a more detailed derivation of and motivation for the method. In analogy with VSF, the following relationship between reactive power injections at the commutation buses and voltage magnitude changes can be derived as: Q 1 Q 2 = J R U 1 U 1 U 2 U 2 where J R is the 2 by 2 reduced Jacobian matrix. It is easily seen that J R corresponds to 1/VSF as defined above. In (2) the relative voltage changes have been introduced. Equation (2) can be written in compact form as: Q = J R U U where Q and U now are vectors with two components each. The stability of the system is determined by the eigenvalues of J R : (1) (2) (3)

5 * Positive eigenvalues indicate stability and if any eigenvalue becomes negative the unstable regime is encountered. ESCR2 It can be shown that, for the considered system, positive eigenvalues correspond to a situation where a reactive power injection in any of the commutation buses results in voltage magnitude increases at the buses in the system. z 1 θ 1 P 1, Q 1 U 1 δ 1 Q d1 P d1 I d1 U d ESCR1= 1.06, PBR= 0.5 p x x Pz 12 Stable x P ESCR Unstable E 1 ψ 1 P 12,Q 12 b c z 12 (p.u.) E 2 ψ 2 z 12 θ 12 P 21,Q 21 Q d2 P d2 I d2 U d2 Figure 10. Stable and unstable regions of the system in Figure 9. (PBR is the ratio between the rated powers of stations 1 and 2.) z 2 θ 2 P 2, Q 2 U 2 δ 2 Figure 8. Model of interconnected /DC system with two converter stations. 8.3 Practical Application The above method will now be demonstrated for a practical case. Consider the situation in Fig. 9. z 12 Fig. 9 illustrates a situation where an subsystem, subsystem 2, exists in the vicinity of /DC. A new link is planned for and with that an interconnection with /DC is also envisaged, perhaps to export the then excess power of /DC to system 1. Here, the system planner would be interested to know the relationship between the coupling impedance, z 12, and the ESCR of /DC for stable operation of the integrated / DC system. The rating of station 2 is twice that of station 1. In the z 12 -ESCR2 plane the border line between stable and unstable operation is shown in Fig. 10. b c2 /DC Planned interconnection /DC Figure 9. Planning of a new link and interconnection. When calculating the graph in Fig. 10 the following assumptions have been made: First, that there is constant power control with constant commutation margin in both stations. Furthermore, parts of the original are included in z 12, which explains the relatively low value of ESCR1, i.e., From Fig. 9 it is clearly seen that the stronger the interconnection is, i.e., the lower the value of z 12 is, the lower is the value of ESCR2 needed for stable operation. (Note that the ESCR2 value in Fig. 9 is calculated from the power rating of station 1. If calculated based on station 2, the value will be half of that indicated.) Depending on the costs and other parameters of the system, the system planner can decide which choice of ESCR2 and z 12 in the stable region gives the optimum solution. Simulations involving a transient stability program with a detailed representation of the stations and their controls verify the stability regions in Fig. 10; i.e., the point indicated by p is stable whereas pescr and pz12 are unstable. 8.4 Concluding Remarks The method described above is a generalization of the method of analysis used for single-infeed systems. The method has recently been proposed and work is currently going on concerning the implications and use of the method. It is the intention that it should be a powerful and useful method for the system planner in the future when analyzing and designing multi-infeed systems. References [1] Guide for Planning DC Links Terminating at System Locations Having Low Short-Circuit Capacities, Part I: /DC Interaction Phenomena, Cigré-IEEE Joint WG, ( ), Cigré Technical Brochure 68, Paris, France, [2] Denis Lee Hau Aik and Göran Andersson, Voltage Stability Analysis of Multi-Infeed Systems. To be presented at IEEE SPM 1996.

6