Power failure in Eastern Denmark and Southern Sweden on 23 September 2003 Final report on the course of events.

Transcription

1 Power failure in Eastern Denmark and Southern Sweden on 23 September 2003 Final report on the course of events 4 november 2003

2 Contents 1. Introduction 2 2. Summary and continuing work Summary of the course of events Possibilities for action The continuing work 6 3. Review of the course of events The initial situation on 23 September hours unit 3 at Oskarshamn trips At about 1235 hours short circuit Ringhals units disconnected from the grid Events from 1235 to 1237 hours Black start Restoration Bornholm The unsupplied energy The state of the system at the end of September Special issues Restoration of voltage to the transmission grid Connecting consumers Public communication DC connections 21 Annex 1 Production and voltage at Avedøre Power Station, unit 2 24 Published by Elkraft System Lautruphøj 7 DK-2570 Ballerup Tel Fax elkraft@elkraft.dk ISBN November Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

3 1. Introduction On Tuesday, 23 September 2003, Eastern Denmark and Southern Sweden suffered an extensive power failure. In Eastern Denmark, this meant that about 2.4 million people were without electricity from 1237 hours. The first consumers on Zealand had power restored to them at 1347 hours, and the vast majority had electricity back a little after 1900 hours. As a result of the power failure, about 1,850 MW consumption was disconnected and about 8 GWh electricity was not supplied as planned in Eastern Denmark. On 24 September, Elkraft System held a press conference at which the course of events was reviewed. The course of events has also been described in a preliminary report from 2 October. This report describes and evaluates the course of events with a view to determining what can be done to reduce the risk of similar situations in the future. The report has been prepared by Elkraft System after a dialogue with Svenska Kraftnät, Energi E2, Østkraft and the East Danish grid companies. Svenska Kraftnät has prepared a similar report on the course of events in Southern Sweden. The report is available at Svenska Kraftnät's website 2 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

4 2. Summary and continuing work 2.1. Summary of the course of events The power failure on 23 September occurred as a consequence of a number of faults in the South Swedish power system. The main cause of the power failure was a double busbar fault at the substation in Horred in Southern Sweden, which caused four 400 kv lines and two units at Ringhals Nuclear Power Station to trip. Prior to that, unit 3 at Oskarshamn Nuclear Power Station had tripped. As a result, the voltage collapsed in Southern Sweden and Eastern Denmark. Before the trips in Southern Sweden, both the South Swedish system and the East Danish system were within the normal limits for system operation. The tripping of unit 3 at Oskarshamn at 1230 hours was handled in accordance with the design rules. However, the double busbar fault at Horred Substation and the consequent tripping of four 400 kv lines in the western part of Southern Sweden and units 3 and 4 at Ringhals at 1235 hours were events that the power system was not designed to handle. Immediately after the short circuit in Horred, there was practically no production left in Southern Sweden and the area was briefly supplied from the Zealand power stations via the Øresund connection and from the eastern transmission links from Central Sweden until the transmission links from Central Sweden were disconnected by protection relays. The Øresund connection and the connection with Bornholm were not similarly disconnected, and Eastern Denmark remained connected to the South Swedish power system until the voltage in the whole of the area fell to zero. The entire course of events from short circuit to zero voltage lasted about 90 seconds. During the course of events great power swings and voltage fluctuations occurred, which affected the large power stations connected to the transmission grid. In practice, it was only the East Danish power stations that were affected because the large South Swedish power stations had been disconnected before that. The power swings and fluctuating voltage damaged unit 5 at Asnæs Power Station, which is Zealand's largest station. Owing to the damage, the unit was unable to help restore the power supply. The unit is still out of operation. Other power stations were also damaged, although to a lesser extent. In the event of big drops in frequency, the East Danish power stations are designed to disconnect from the grid and go over to stable house-load operation, in which the plant is kept in operation without supplying electricity to the grid. However, the situation in the East Danish transmission grid during the voltage collapse meant that the East Danish power stations did not go into stable house-load operation, and this delayed the restoration of the power supply. As a result, both the grid and power stations in Eastern Denmark were without voltage shortly after the short circuit in Horred. After the voltage collapse the transmission grid was prepared for restoration of voltage, which took place via the Øresund connection at 1347 hours. This connection then contributed about 200 MW power for a period of time, making it possible to reconnect the first Zealand consumers to the grid shortly before 1400 hours. Consumers were then reconnected to the grid in step with the reconnection of the East Danish power stations to the grid, and just after 1900 hours Elkraft System gave permission for the grid companies to reconnect the last consumers. However, owing to local conditions, a few consumers did not get electricity back until around 2200 hours. 3 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

5 An attempt was made to black-start Kyndby Power Station, but this proved impossible because of faults in both the station's black-start units. The island of Bornholm, which has a power station in Rønne and a 60 kv cable to Southern Sweden, was also affected by the voltage collapse. Østkraft Power Station was able to black-start within 12 minutes by means of the diesel plant in Rønne. With this, 40 per cent of the consumption could be covered while Bornholm was in island operation. At about 1600 hours the island was reconnected to Sweden, whereby the remaining consumers on Bornholm could be reconnected. The other units on Bornholm were not in operation before the power failure, and because of the long start time from cold plant, it was not possible to get them going again before the connection with Sweden had been re-established. To summarise, it must be concluded that, with the present design of the power system and the current criteria for system operation and protection of grid and plants, it was not possible to prevent the power failure once the short circuit between the two busbars in Horred had happened Possibilities for action The following factors must be evaluated as an element of minimising the risk of similar power failures in the future: 1. The fault that triggered the course of events on 23 September was mechanical failure in an isolator. The fault focuses general attention on design, inspection and preventive maintenance of the transmission system, particularly at vulnerable points with major consequences in the event of a fault. 2. The tripping of the 400 kv connections on the west coast of Southern Sweden, combined with the loss of 3,000 MW production, violently overloaded the eastern transmission link between Southern and Central Sweden. As a consequence of this, a big drop in voltage occurred in the eastern part of Southern Sweden. Consideration should be given to including voltage-controlled disconnection of electricity consumption as an active possibility in the handling of such system disturbances. This applies particularly in areas with a strained power situation due to a shortage of production capacity. 3. Just before the final voltage collapse, the protection relays in the Swedish transmission grid disconnected Southern Sweden from the rest of Sweden. This happened because the relays registered the fault as a far-off short circuit and reacted after a short, planned delay by disconnecting the transmission links, both to protect the components in the transmission grid against damage and to isolate the fault. It would have been useful for Zealand to have been disconnected at the same time as the internal Swedish transmission lines. That did not happen because the relays on the Øresund connection did not similarly measure imbalance in the system, and this connection was therefore only disconnected when the voltage reached zero. Disconnection of the Øresund connection at the same time as the internal Swedish transmission links would have required more advanced measuring and control systems, in which information from several areas and system conditions were integrated. Such systems have not yet been fully developed and also require an integrated evaluation of the entire design and system operation strategy for the Nordic synchronous area in order to avoid unplanned disconnections. 4 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

6 4. Like the relays on the Øresund connection, the relays at the East Danish power stations did not register the unstable situation in the South Swedish power system. The power stations remained connected to the grid and helped to keep the voltage up on Zealand and in the western part of Southern Sweden. From this point of view, it was good that the power stations did not disconnect from the grid at that point of time. However, production at the Zealand power stations was far from sufficient to maintain the supply in the whole of the South Swedish/East Danish area, and towards the end of the course of events, the large power stations were heavily overloaded. The unstable situation at that point made it difficult for the power stations to change to stable house-load operation, and at the same time, serious damage occurred to unit 5 at Asnæs Power Station. In view of this, it would have been useful if the power stations had disconnected from the grid earlier to go into stable houseload operation, even though that would have hastened a voltage collapse. The situation reflects the conflicting considerations that have to be made - and made allowance for when establishing the protection strategy. On the one hand, the power stations are needed to keep the system in operation and to re-establish stable operation, but on the other, they must disconnect from the grid early enough to avoid damage and be able to go into stable house-load operation. In the actual course of events, the protection relays at the power stations were unable to register the collapse in the grid until a very late stage, in part because the frequency, voltage and current did not change dramatically on Zealand in the initial stage of the course of events. There is therefore a need to develop protection systems that are also able to detect voltage collapse in the entire system. Here, too, the protection strategies must be seen together with the entire strategy for regulation of the power system. 5. The first step in restoring the electricity supply is to prepare the transmission grid for voltage to be restored to it. Elkraft System's control room is in charge of the preparation, which is carried out via the grid companies' control rooms. On 23 September the preparation work took around 40 minutes, which is only to be expected since some of the preparation had to be done manually. It would be useful to be able to reduce the preparation time by various means, including establishing relays that disconnect the underlying grid etc. in the event of zero voltage. 6. On 23 September, voltage was restored to the transmission grid via the Øresund connection. As mentioned, the attempt to restore voltage from Kyndby failed because of faults at both black-start units. Energi E2 has already taken steps for better monitoring of these units' start reliability. Besides that, consideration must be given to how restoration of voltage to the transmission grid on Zealand can be improved, including the possibility of restoring voltage from power stations in house-load operation. Lastly, consideration must be given to establishing special plants dedicated to voltage restoration from dead grid. 7. The electricity supply was restored in step with the reconnection of Zealand's power stations to the grid because only limited power could be obtained from Sweden. The restoration was therefore relatively long drawn-out. It would have taken less time if some of the power stations had gone into stable house-load operation or the Kontek Link with Germany had been in operation during the power failure. The grid companies had no serious problems in reconnecting consumers, but Elkraft System will go through the principles for restoration with the grid companies and the authorities with a view to ensuring the right order of priority for disconnection and reconnection of consumers. 5 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

7 8. Communication between Elkraft System's control room and the control rooms at Energi E2, the power stations and the grid companies functioned without major problems, but the power failure revealed a number of deficiencies that should be remedied. Similarly, communication with the public on the day itself functioned reasonably well, whereas there were problems with communication between Elkraft System and the grid companies with respect to information to the consumers. This too should be adjusted. The power failure on 23 September stresses the importance of having sufficient production in the vicinity of the electricity consumption centres in the case in question, that means in Southern Sweden. In Eastern Denmark, Elkraft System has chosen to ensure that there is always a positive power balance. The power failure also demonstrates the importance of strong transmission connections. It is therefore necessary to consider how new transmission connections can help to improve security of supply in the area. The power failure also gives grounds for considering whether the development of the electricity market is producing changed conditions for system operation. There is thus a need to evaluate and, if necessary, adjust the entire package of technical specifications for grid and production plants and the Joint Nordic System Operation Agreement The continuing work The continuing work of reducing the risk of similar power failures in future is primarily going on in a cooperation between Elkraft System, the East Danish grid companies, the East Danish generators, the other Nordic system operators and the Danish authorities, including particularly, the Danish Energy Authority. The above-mentioned points will be among the matters taken up in this cooperation. In addition, the need for research and development will be evaluated. The following, in particular, is stressed: - In Nordel, the Nordic cooperation between the system operators, a decision has been taken to review the operating and planning specifications and agreements together in order to evaluate whether they need updating. - A working group headed by the Danish Energy Authority is at present analysing the framework for security of supply, including a Great Belt connection. Both the Danish system operators are participating in this working group. - A review of the East Danish transmission grid has been initiated, partly with a view to identifying and possibly strengthening vulnerable points. - In cooperation with the grid companies, Elkraft System will review the handling of the transmission grid and improve preparedness for major system disturbances in the power system, including the prioritisation of disconnection and reconnection of consumers. - Together with the generators, including particularly Energi E2, Elkraft System will also examine the entire regime concerning system operation and protection systems. A major element of this examination will be the requirements and expectations concerning the primary power stations' possibilities for house-load operation, restoration of voltage to the grid and their own protection against faults. This work is closely related to the work going on under Nordic auspices. - Elkraft System will review its level of preparedness for major system disturbances in the light of the experience from 23 September. This will include ensur- 6 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

8 ing that Elkraft System's website functions efficiently, even during power failures that affect Elkraft System's buildings. - The general status of work on security of supply will be given in the coming System Plan from Elkraft System. The status of the work on the protection systems in the transmission grid and at power stations will be given in next year s Transmission Plan. Lastly, these tasks must be seen in connection with the authorities' general work on Denmark's preparedness. 3. Review of the course of events 3.1. The initial situation on 23 September 2003 The initial situation was characterised by maintenance works on the transmission grid and at the power stations, which is normal for the time of year. o o o o o The DC connection between Zealand and Germany (Kontek) was out of operation for planned maintenance. In addition, several 132 kv lines and a 400 kv line in Eastern Denmark were out of operation for various reasons. None of these lines had any effect on the course of events. In Eastern Denmark, unit 4 at Asnæs Power Station was in mothballs. Unit 1 at Avedøre Power Station was undergoing an overhaul and unit 1 at Amager Power Station was out of operation because of damage to it. Both DC connections between Southern Sweden and the Continent (Swepol Link and Baltic Cable) were out of operation for planned maintenance. In the Swedish transmission grid, two 400 kv lines were disconnected for inspection and maintenance. The lines in question are the Strömma-Horred- Breared line on the west coast of Sweden and the Hallsberg-Kimstad line, which connects Central and Southern Sweden. As far as the Swedish power stations are concerned, unit 2 at Barsebäck and units 1 and 2 at Oskarshamn were out of operation owing to prolonged maintenance. In addition, Karlshamn Power Station was out of operation. These factors were taken into consideration in the operational planning, so the system was in a reliable operating condition before the system disturbance. Just before the power failure, the power stations in Eastern Denmark were producing about 1,800 MW, while wind turbine production was about 450 MW. This covered a consumption of about 1,850 MW in Eastern Denmark and a market-determined export to Sweden of about 400 MW. In Eastern Denmark, the total capacity ready to go into operation at that time was around 3,300 MW, which covered reserves of 775 MW besides consumption and export. There was thus plenty of production capacity in Eastern Denmark to cope with a random design trip 1. 1 The operation of the power system is planned in such a way that the power system can be brought back to a reliable operating condition within a reasonable time after a single, random grid or production unit has tripped (n-1). Here, reliable operating condition means that the next fault can be tolerated. This design criterion is the same as the Joint Nordic System Operation Agreement is based on. 7 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

9 In Southern Sweden, consumption before the power failure was 3,000 MW. According to Svenska Kraftnät, operating reliability in Sweden was also normal and the load on the grid was low, with big margins to the specified limits hours unit 3 at Oskarshamn trips Due to problems with a valve in the feed-water circuit at unit 3 at Oskarshamn, production at the unit had to be reduced, which rapidly resulted in the unit stopping altogether. With that, about 1,200 MW electricity production was lost. As a consequence of the unexpected shortage of production, the frequency in the synchronous Nordel grid dropped. The falling frequency activated the instantaneous system disturbance reserves, whereby the fall in frequency was stopped and the frequency was stabilised at about 49.9 Hz, which is an entirely normal consequence of a large loss of production. No abnormal voltages were registered in the East Danish power system. The system is designed to tolerate tripping of a power station by automatically increasing production at other power stations. This must happen within 15 minutes. And it did happen production from other stations was increased (activation of instantaneous system disturbance reserves) in Northern Sweden, Norway, Finland and Denmark. The system thus reacted as it should to the tripping of unit 3 at Oskarshamn At about 1235 hours short circuit Ringhals units disconnected from the grid The next event occurred just five minutes after the first one: at a 400 kv substation (Horred) in the South Swedish transmission grid, a double busbar fault occurred as a result of mechanical failure in an isolator. This meant that four 400 kv transmission lines were disconnected. Two of the disconnected lines were an important connection between Southern and Central Sweden. The other two lines connected units 3 and 4 at Ringhals with the transmission grid, so these units lost their connection to the power system. When these four connections were disconnected, energy transmission rose on the remaining connections. Unit 4 at Ringhals managed to go into house-load operation. The disconnection of the Ringhals units meant that the system lost a further approx. 1,800 MW production. A fifth 400 kv line to the station remained connected but was not connected to other lines at the station and was therefore useless in the continuing course of events. The pressure on the remaining connections between Southern and Central Sweden, which carry energy in the north-south direction, increased. This situation was aggravated when the power stations in Northern Sweden, Norway and Finland increased production still further. The voltage consequently began falling. It fell greatly in Southern Sweden but somewhat less in Eastern Denmark because the East Danish power stations were helping to keep the voltage up locally. The voltage drop was intensified by the fact that the transmission grid had been seriously weakened by the tripping of the 400 kv lines in the damaged substation. 8 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

10 3.4. Events from 1235 to 1237 hours In the period from 1235 to 1237 hours a number of events occurred in the power system. As mentioned, they started at 1235 hours with a double busbar fault in a 400 kv substation (Horred) in the South Swedish transmission system. They ended with the voltage down at zero at 1237 hours, leaving Eastern Denmark and Southern Sweden without electricity. In order to obtain a complete picture of the course of events it is necessary to consider the development in voltage (kv), frequency (Hz), power (MW) and reactive power (Mvar). The course of events can be divided into five phases (see figure 1). These are described in the following. Phases 1 and 2: short voltage drop and frequency fluctuations As will be seen from figure 1a, the frequency before the fault at Horred Substation was about 49.9 Hz and the voltage just under 410 kv in Söderåsen. The fault in Horred and the disconnection of the Ringhals units and the four 400 kv lines in Sweden (phase 1) meant that the voltage dropped briefly. At the same time, the frequency began to fall. During the voltage drop, the frequency fluctuated for about 10 seconds (phase 2) because the East Danish power stations oscillated towards the power stations in Norway, Sweden and Finland. The oscillations were quickly reduced to relatively small frequency fluctuations that did not cause disconnection of the Øresund connection. kv 600 Hz seconds Voltage (kv) Frequency (Hz) Figure 1a. Variations in the voltage (kv) and the frequency (Hz) during the course of events from about 1235 to 1237 hours. The values were measured in Söderåsen on the 400 kv Øresund connection. 9 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

11 The frequency is kept at 50 Hz in normal situations and is the same throughout the Nordic synchronous system. Fluctuations in the frequency are handled in different ways, depending on the magnitude of the fluctuation. In the case of small fluctuations, use is made of frequency-control reserves and system disturbance reserves, which are activated by the plants automatically increasing or reducing electricity production. In the case of major frequency drops, the consumers are automatically disconnected. Power (MW) production at the plants is activated on the basis of the fluctuation in the frequency. The voltage in the main transmission grid is 400 kv or 132 kv. The voltage varies, depending on local consumption, production or transmission factors. There can therefore be differences in measurements of the voltage at different points. In Söderåsen, efforts are made to keep the voltage in the 400 kv grid at approx. 415 kv. The voltage is controlled with reactive power. The reactive power (Mvar) can be generated at primary power stations and, to some extent, in grid components. Production is activated on the basis of fluctuations in the voltage. In an integrated power system the voltage varies and it is therefore necessary for the production of reactive power to be evenly distributed in the system. Increasing reactive power, e.g. in situations with system disturbances, must therefore be done locally and close to the system disturbance in order to have the biggest effect. The fluctuations in frequency were countered by fluctuations in the power (MW) and reactive power (Mvar) that the East Danish power stations were supplying via the Øresund connection (see figure 1b). MW/Mvar seconds Supply of power to Southern Sweden (MW) Supply of reactive power to Southern Sweden (Mvar) Figure 1b. Variations in supply of power (MW) and reactive power (Mvar) to Southern Sweden during the course of events from around 1235 to 1237 hours. The values were measured in Söderåsen on the 400 kv Øresund connection. 10 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

12 Phase 3: Small voltage drop in Eastern Denmark large voltage drop in South-east Sweden About 10 seconds after the fault in Horred the frequency quickly rose to just over 49.7 Hz. At the same time, yet another brief voltage drop occurred in Eastern Denmark (measured at the Øresund connection). The drop in voltage was caused by tripping of a number of 130 kv lines and a 220 kv line in Sweden, probably because of serious overloading as a result of the disconnection of the 400 kv lines at Horred Substation. For Southern Sweden the disconnection of the 130 kv lines and the 220 kv line resulted in the voltage in a large area in South-east Sweden being halved. Shortage of capacity to adjust the voltage in Southern Sweden and the weakening of the transmission grid made it impossible to restore the voltage. As a consequence of the very low voltage in this area, electricity consumption there fell heavily not due to disconnection of consumption but because the consumers themselves stopped using electricity when the voltage was low (this applies mainly to motors and similar). It was because of this fall in electricity consumption that the frequency quickly rose to 49.7 Hz. Figure 2. The grid's connection condition about 10 seconds after the short circuit in Horred. The tripped lines in the Swedish transmission grid are shown in broken line. It will be seen that only two 400 kv lines on the east coast connect Southern and Central Sweden. The voltage drop over these lines was very big because of the length of the lines combined with the large flows in the lines. 11 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

13 The voltage drop caused the East Danish power stations to increase the supply of reactive power (Mvar) in order to maintain the 400 kv voltage. It was particularly unit 5 at Asnæs Power Station and unit 2 at Avedøre Power Station that increased supply because they are directly connected to the 400 kv grid. Figure 1b shows that the quantity of reactive power (Mvar) supplied from Zealand to Sweden briefly increased from about 300 Mvar to over 900 Mvar. Although this was a big increase, it was not big enough to activate the protection systems at the East Danish power stations, nor did the increase cause overloading of the Øresund connection. Phase 4: Slow voltage drop in Eastern Denmark A slow voltage drop then occurred in Eastern Denmark (measured on the Øresund connection) in the space of about 80 seconds. In South-east Sweden, the voltage drop was considerably greater. The reason for the small voltage drop in South-west Sweden at this point of time (measured on the Øresund connection) was that the electrical connection between Southwest and South-east Sweden was so long that the voltage drop could only spread to a limited extent. Besides the long distance, the electrical connection had also been weakened by the tripping of the 400 kv lines to the substation in Horred. Phase 4: The distance relays In Sweden, the combination of falling voltage and greatly increased flow from north to south through the weakened grid meant that the so-called distance relays on the connections between Southern and Central Sweden registered the situation as a far-off short circuit. The relays were activated and, after some time, a number of 400 kv lines between Southern and Central Sweden were disconnected. There was then no longer any electrical connection between the two areas. On the Øresund connection the distance relays were not activated as they were between Southern and Central Sweden because the voltage at the Swedish end of the connection was still relatively high up to the point at which the distance relays on the connections between Southern and Central Sweden were activated. The relay protection on the Øresund connections and in the East Danish transmission grid functioned in accordance with the design principles. When the connections between Southern and Central Sweden had been disconnected, Southern Sweden, Bornholm, Zealand, Møn, Lolland and Falster formed a sub-area without connection to the rest of the Nordic power system. The sub-area had a large production deficit. Phase 5: Disconnection of the East Danish power stations and voltage drop to zero At that time, about 90 seconds after the fault at Horred Substation, the voltage fell to zero in about 1-2 seconds, leaving the entire sub-area without electricity. At the beginning of phase 5, Zealand's power stations were still connected to the transmission grid and were supplying very large quantities of power despite rapid deceleration. This applied particularly to unit 5 at Asnæs Power Station and unit 2 at Avedøre Power station, both of which were heavily overloaded before they were disconnected by the units' protection system. 12 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

14 After the voltage fell to zero, Eastern Denmark was automatically disconnected from Southern Sweden via the zero-voltage relays on the Øresund connection. The voltage collapse caused damage to unit 5 at Asnæs Power Station (the generator transformer was damaged) and unit 2 at Avedøre Power Station, which are the two largest plants on Zealand. The damage to unit 5 at Asnæs Power Station meant that the unit was unable to help in restoring the power supply. Unit 2 at Avedøre Power Station was used during the restoration, but tripped five days later, probably as a consequence of the impacts during the voltage collapse. The unit was back in operation one week later. Annex 1 shows the course of the power supplied by unit 2 at Avedøre Power Station and the voltage measured at that unit in the period from about 1235 hours to 1237 hours Black start Voltage disappeared from the grid at 1237 hours. Just a few minutes later, the preparations to restore voltage began. The grid was prepared 2 for restoration of voltage from Kyndby Power Station or via the Øresund connection when the condition in the Swedish grid allowed that. Both possibilities were kept open in order to be able to make use of the fastest possibility of restoring voltage. By around 1320 hours, the grid companies had disconnected - consumers and small-scale production at the 10 kv level - the 50 kv grids - the 400 kv and 132 kv lines that Elkraft System had ordered to be disconnected because of the reactive balance. At 1330 hours, Svenska Kraftnät established a 400 kv connection between the Norwegian, North Swedish and Finnish synchronous system and Söderåsen, where the 400 kv Øresund connections are connected. At that time, Kyndby Power Station was not yet ready to restore voltage. See section 4.1 for further details. At 1341 hours, Svenska Kraftnät gave permission for voltage to be restored to the Zealand grid via a 400 kv Øresund connection. At 1346 hours, the 400 kv Söderåsen-Hovegård connection was connected, whereby voltage was restored to the 400 kv and 132 kv grids on Zealand. The voltages in the Zealand system fluctuated because there was no voltage-control plant connected to the grid on Zealand. At the request of Elkraft System, Svenska Kraftnät gave permission for Zealand to import up to 200 MW. Between hours, the grid companies were given permission to connect consumption corresponding to about 150 MW, the last 50 MW being reserved for margin in order to avoid again disconnecting consumers that had got power back if the connected consumption proved to be higher than expected by the grid companies. It was then not possible to give the grid companies permission to reconnect more consumers before the Zealand power stations were reconnected to the grid. 2 The preparation of the transmission grid implies disconnection of most consumers from the whole of the 400 and 132 kv grid and that the grid has reactive balance that makes it possible to restore voltage to the grid from a power station or from Sweden. 13 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

15 3.6. Restoration Restoration then awaited connection of production plant that could help to stabilise the voltage and maintain a reasonable reactive balance vis-à-vis the Swedish system. Since the small-scale CHP plants do not have these properties, they could not be connected until later, when the conditions for their operation were in place. The same applied to the large offshore wind farms. The later reconnection of these plants was also due to the fact that today the East Danish power system is built up in such a way that there must be voltage on the high voltage grid, and the grid must be stable, before most of the small-scale CHP plants and wind turbines can be connected to the grid at low voltage levels. When voltage was restored at 1347 hours, Energi E2 prepared a start-up plan for their power stations. The plan contained the start-up times for each unit, prepared on the basis of the stations' estimates of the time they needed to get their units onto the grid. This plan was followed for most of the units. There are already start-up times for the individual units, depending on how long the unit has been out. These times are given on the assumption of a controlled shutdown of the unit, not a sudden trip as happened on 23 September The normal start-up times could therefore not be used as estimates of the time needed for the restoration. At 1415 hours about 30 minutes after the restoration of voltage unit 22 at Kyndby Power Station was synchronised to the grid. At 1416 hours, Svenska Kraftnät informed Elkraft System that no more than 100 MW was to be imported from the Swedish system. After reconnection, unit 22 at Kyndby Power Station increased production by about 3 MW per minute. In step with this, the available production was distributed between the grid companies in such a way as to comply with the import limit from Sweden. The Øresund connection's main role in this period was thus to ensure stable frequency. Without the connection with the Nordic system, the frequency would have fluctuated greatly. As the day progressed, the thermal production units were started up and synchronised to the grid as follows: 1457 hours Masnedø Power Station, unit hours Kyndby Power Station, unit hours Kyndby Power Station, unit hours Kyndby Power Station, unit hours Amager Power Station, unit hours Stigsnæs Power Station, unit hours Svanemølle Power Station, unit hours Avedøre Power Station, unit 2, the steam turbine 1825 hours Asnæs Power Station, unit hours Avedøre Power Station, unit 2, gas turbine hours Avedøre Power Station, unit 2, gas turbine hours H.C. Ørsted Power Station, unit hours Stigsnæs Power Station, unit hours Amager Power Station, unit 1. The consumers were reconnected in step with the reconnection of the East Danish power stations to the grid. 14 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

16 MW :00 14:00 15:00 16:00 17:00 18:00 19: MW on Øresund Consumption Figure 3. Electricity consumption and exchange on the Øresund connection during restoration on 23 September 2003 (positive values are import from Sweden to Eastern Denmark) SEAS was given permission to connect Nysted Wind Farm little by little after unit 31 at Masnedø Power Station had been synchronised to the grid and the voltages on Lolland were stable. Similarly, Copenhagen Energy was given permission to connect the offshore wind farm on Middelgrunden little by little after unit 3 at Amager Power Station had been synchronised to the grid and could thus control the voltage in Copenhagen. Stand-alone wind turbines started up when they themselves had measured normal voltage and frequency for 5-10 minutes, and some of the small-scale CHP plants were started up manually when voltage had been restored to the 10 kv connection to which they are connected. Figure 4 shows the production from wind turbines and small-scale CHP plants during the restoration. 15 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

17 MW :00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 Small-scale CHP plants Wind turbines Figure 4. The production from wind turbines and small-scale CHP plants during the restoration on 23 September Bornholm The island of Bornholm, which has a power station in Rønne and a 60 kv cable to Southern Sweden, was also hit by the voltage collapse. Østkraft Power Station was able to black-start within 12 minutes via the diesel plant in Rønne. With that, 40 per cent of the consumption could be covered while Bornholm was in island operation. At about 1600 hours, the island was reconnected to Sweden, whereby the remaining consumers on Bornholm could be reconnected. The other units on Bornholm were not in operation before the power failure and could not get going again before the connection to Sweden was re-established owing to a long start-up time from cold plant The unsupplied energy The power failure meant that about 2.4 million people in Eastern Denmark were without electricity from 1237 hours up to 1900 hours. Around 1,850 MW consumption in Eastern Denmark was instantaneously disconnected at about 1237 hours. By comparing the difference in consumption on 22 and 23 September in the period in which the power failure affected electricity consumption, the expected unsupplied energy can be calculated as approx. 8 GWh. The difference is illustrated in figure Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

18 MW :00 09:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 22 Sep. 23 Sep. Figure 5. Electricity consumption in Eastern Denmark from 0800 to 2300 hours on 22 and 23 September The difference between the bars in the diagram and the fully drawn line is the expected unsupplied energy The state of the system at the end of September 2003 Unit 5 at Asnæs Power Station was damaged in connection with the voltage collapse and is still out of operation here at the beginning of November. During the night between 23 and 24 September, consumption fell to a minimum of about 1,200 MW, and with capacity of about 2,600 MW ready to go into operation, there was at that time ample power to cover consumption and operating reserves. On Sunday, 28 September, unit 2 at Avedøre Power Station tripped, probably as a result of the impacts during the voltage collapse. This caused a reduction in production capacity, and the power balance became strained. However, as it was a Sunday, with relatively low consumption, the power balance remained positive despite the reduction. The unit was back in operation one week later. On Tuesday, 30 September, the 250 MW unit 4 at Asnæs Power Station was taken out of mothballs, which improved the power balance. At that time, there was thus a slight surplus of production capacity and a stable system with the possibility of both importing and exporting. 4. Special issues 4.1. Restoration of voltage to the transmission grid Voltage can be restored to the East Danish transmission grid in three ways by means of the Øresund connection, from a power station designed to start up without voltage from outside or from a power station unit in house-load operation. On 23 September, the Øresund connection was used. 17 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

19 Preparation of the grid When the grid is to be prepared for voltage restoration, all consumption and production units must be disconnected from it. This is to ensure that, at the moment voltage restoration commences either from a power station on Zealand or via the Øresund connection there is always balance between consumption and production. Some of the preparation is done automatically, while other preparation is done manually on orders from Elkraft System's control room. Restoration of voltage from Sweden Voltage can be restored to the Zealand grid from Sweden, via a 400 kv connection or a 132 kv connection. It is preferable to use a 400 kv connection because these connections are stronger than the 132 kv connections. If voltage is to be restored to the Zealand grid from Sweden, Svenska Kraftnät must agree to that, the voltage in Sweden must be stable and there must be surplus production capacity in Sweden. Restoration of voltage from Sweden is normally the possibility that is given highest priority because it is regarded as the fastest and most reliable solution. For example, the East Danish power system has the advantage of synchronisation with the rest of the Nordic system, which maintains the frequency. Black start from Kyndby Kyndby Power Station is actually 3 the only power station on Zealand designed to blackstart. There are two possibilities of black-starting at the station: 1. Starting up unit 21 (KYV21) by means of the diesel plant (KYV41) and possibly KYV51 or KYV52 2. Starting up unit 22 (KYV22) by means of two gas turbines (GT0 and KYV51 or KYV52). When the voltage disappeared from the Zealand grid, the diesel plant was started up manually so that it could be used to start up one of the units (KYV21). However, during the start-up of KYV21 the regulating equipment on the diesel plant was found to be defective 4 and it was therefore not possible to keep the frequency sufficiently stable to start up KYV21. The conditions for restoring voltage to the Zealand grid by means of the diesel plant were thus not in place. By the time the preparations for trying to black-start by means of the two gas turbines, GT0 and KYV51, were completed, voltage had already been restored to the Zealand grid from Sweden. An attempt was then made to start up GT0, but this was found to be damaged (a printed circuit board in the control system had burnt out). Since remote monitoring of such faults had not been installed, this was not registered. Such remote monitoring has since been installed. 3 Under very special/favourable conditions, the diesel generator at H.C. Ørsted Power Station can also be used as a black-start option. 4 During the first frequency drop at 1230 hours, when unit 3 at Oskarshamn tripped, Kyndby Power Station's diesel generator (KYV41) started up automatically as it should. As the unit was thus connected to the grid during the voltage collapse, it was also subjected to the big impacts caused by the collapse. 18 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

20 In Energi E2 s experience 5, voltage can realistically be restored to a dead grid from Kyndby Power Station in about 90 minutes by means of the diesel plant and in about two hours by means of the two gas turbines, always assuming that everything is functioning as planned and that no damage occurs. Energi E2 believes that if the diesel plant had been working on 23 September, unit 21 at Kyndby Power Station could have restored voltage to the Zealand grid at around 1430 hours. Voltage restoration from power station unit in house-load operation The Zealand power stations are set to disconnect from the grid in the event of big frequency fluctuations. The power stations are not similarly set to disconnect themselves when the voltage is low. That is because it is not advantageous in all situations for the stations to disconnect themselves in the event of low voltages. Registration by a power station unit of low voltage in the grid can, for example, be due to a short circuit in the grid nearby that lasts just a few milliseconds. In such situations, security of supply would be adversely affected if a power station unit disconnected from the grid. The way the frequency and voltage in the Zealand grid developed on 23 September, the power stations did not disc0nnect from the grid until late in the course of events, which made the switch to house-load operation more difficult than usual. Four out of ten of Zealand's primary power station units succeeded in going into houseload operation, although only briefly, after which they tripped completely. The possibility of restoring voltage to the Zealand grid from one of the power station units in house-load operation has never been tried out in practice. A review must therefore be carried out to determine whether this possibility is realistic and, if so, under what assumptions Connecting consumers The consumers are not connected directly to the transmission grid, and it is therefore the grid companies that can in practice disconnect and rec0nnect consumers. In connection with the restoration of normal supply, Elkraft System gives the grid companies permission to reconnect consumption. This is based on fixed routines, the consumption being divided into 10 steps, each consisting of 10 per cent of total consumption. The grid companies prioritise reconnection themselves within their respective supply areas. With respect to time, the consumers were reconnected as follows on 23 September: 1354 hours Copenhagen Energy receives permission to reconnect 50 MW 1413 hours NESA receives permission to reconnect 50 MW 1414 hours SEAS receives permission to reconnect 25 MW 1414 hours NVE receives permission to reconnect 25 MW. The remaining reconnection from step 10 to step 1 is carried out in the order shown in the following table. 5 This has been tried on a small section of the grid. 19 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,

21 Step 10 Step 9 Step 8 Step 7 Step 6 Step 5 Step 4 Step 3 Step 2 Step 1 NESA 1) SEAS ) KE ) NVE 2) ) Frederiksberg Elnet is included in NESA's reconnections. 2) Due to the reactive balance in the transmission grid, NVE was allocated 25 MW early on, corresponding to steps 10 to 8. The reconnection in step 7 also covers steps 6 to 5. 3) Time not registered. Both disconnection and the subsequent reconnection were carried out prioritising customer groups of special social importance. The general picture is that the reconnection functioned as planned, but it has also been found that the present routines should be reviewed to see whether they are appropriate Public communication Elkraft System has an information preparedness plan that is implemented in the event of major power failures like the one that occurred on 23 September. One of the features of the plan is that instructions in the control room ensure that, in the case of major power failures or in anticipation of a major power failure, the authorities are informed of the situation as soon as possible. After that the main media are informed with a view to providing information to the general public on the extent, duration and cause of the power failure. To evaluate and revise the information preparedness after the power failure on 23 September, Elkraft System has asked the authorities, the press, large customers and the grid and production companies about their experience with the information provided by Elkraft System during and after the power failure. The analysis will provide a basis for improving communication in similar circumstances. Information effort and main conclusions On the whole, the communication lived up to the primary aim of informing the general public via the media as soon as possible about the extent, duration and cause of the power failure, although with the reservation that it took longer to restore the power supply than initially expected and that Elkraft System did not know the real cause of the power failure until early evening. That is why, at first, it was not possible to provide precise information about the duration and cause. In compliance with the information preparedness plan, Elkraft System s control room informed the Copenhagen Police a few minutes after the power failure had been registered, and the police then informed DR Radio News and DR Radio s regional stations on Zealand. Information to DR Radio News and Ritzau's Bureau was given priority in the first hectic hours because these media have the widest coverage. Incidentally, DR Radio News had a journalist stationed with Elkraft System during the power failure. Secondarily, other media were serviced, giving priority to radio, TV and other electronic media rather than the written media. Also based on the degree of coverage, DR TV News and TV2 News were given priority, for example via direct interviews from Elkraft System s offices. 20 Power failure in Eastern Denmark and Southern Sweden on Final report on the course of events,