European solar survives its darkest hour7 October 2015
The solar eclipse of 20 March placed unprecedented stress on the European electricity grid. The pre-event analysis and the lessons learned during the eclipse have provided unique data that suggest how to deal with future events when a far greater proportion of solar generation will be on the grid.
On 20 March the European electricity grid passed with conspicuous success an unprecedented stress test. The event - a near total solar eclipse taking place on a mostly sunny weekday morning with about 90 GW of solar power installed and highly concentrated in some regions - had never before been experienced. This was not just a first for Europe, it was a world first. It pointed the way to the future - dealing with the effects of eclipses when a far greater proportion of solar generation is on the grid.
Solar eclipses occur frequently, but in the traditional generation mix centred on fossil fuels they have been no more than an inconvenience. This has changed with the increasing share of solar generation. For power systems that need to be in balance second by second, the very large and fast variations in solar generation that take place during an eclipse are a severe challenge.
Following the pre-event analysis and its test during the event itself, ENTSO-E and SolarPower Europe launched a joint report, Solar Eclipse: the successful stress test of Europe's Power Grid, from which all the information in this article is taken. The first part of the report describes in detail what happened on that morning for the power sector. In the second part the report takes a looks into the future, the next large scale eclipses in Europe in 2021 and 2026, a time when the PV component is likely to have been significantly expanded, and suggests the measures that will be required to diminish costs and to ensure reliable power supplies across the continent.
No power grid had previously experienced the combination of a large-scale eclipse and a very large PV component, in this case up to 90 GW, and very concentrated in some areas. Therefore it provided for the first time an opportunity to test the grid's response and at the same time provide indications about what measures could be relied on during the eclipses that will occur in the future. In the event, the measures applied were considered a success - Europe stayed switched on throughout the three hours the eclipse lasted.
A study was carried out by the ENTSO-E Subgroup System Protection and Dynamics (SG SPD) to analyse the effects this eclipse could have on the grid. Other regional groups were asked to analyse the impact and make a solid plan to handle the effects. For Continental Europe a subgroup (Co-ordinated System Operations) formed a taskforce that looked at the countermeasures that could be deployed during the eclipse.
Analysis suggested that Continental Europe should expect a reduction of the PV feed-in of up to 34 GW or more, and with a power gradient 2 to 4 times steeper than the norm for daily ramping. This situation could cause serious difficulties in regulating the interconnected power system in terms of available regulation capacity, regulation speed and geographical location of reserves.
Beyond the reduced PV generation the most important anticipated challenge was the reduction of 20 GW within 1 hour, and the increase of almost 40 GW after the maximum impact of the eclipse. Figure 1 shows a comparison of expected feed-in under clear sky conditions, with and without solar eclipse. Figure 2 shows the estimated impact on PV feed-in by country.
Since the SG SPD study suggested that the solar eclipse would not cause a dynamic effect on the grid, the operational task force focused on quasi-static behaviour and the gradient of changes of PV generation. The input of this task force was the basis for all further analysis and defined the raft of countermeasures that could be undertaken by TSOs, with the primary responsibility being that of those with high levels of PV generation - Germany, Italy, and Spain which together contain 78% of the total ENTSO-E solar generation.
Owing to the different sizes of transmission networks and to different national regulations, TSOs were obliged to evaluate and to confirm both on national (with regulator and ministries, DSOs, power plants) and European level (with other TSOs), their individual needs of measures, and the feasibility to activate different measures. In the event careful and co-ordinated planning and good communication before and during the eclipse obviated the need for TSOs to provide assistance to each other. This success does suggest however that careful planning and co-ordination between TSOs in all operational phases will be necessary during similar events in the future.
The weather conditions during the solar eclipse for the western part of Continental Europe were cloudier than expected, which resulted in a less severe impact than predicted. But Germany and Italy both experienced clear skies and the impact of the eclipse was considerable (Table 2).
- All TSOs agreed to impose as few as possible planned outages on their grids during the eclipse.
- Capacity on HVDC interconnectors was reduced by between 18% and 50% for the Nordic, UK and CE regions. This ensured that synchronous areas would be more independent of each other.
- TSOs informed market players, that is, parties responsible for balancing as well as distribution system operators, of the responsibilities they would have during the eclipse.
- An operational teleconference called 'RG CE real-time frequency monitoring Telco' was held among the five TSOs of the frequency- monitoring group - Amprion, REE, RTE, Swissgrid, Terna. This Telco is normally started if frequency deviation exceeds predefined values. It was in place during the entire eclipse. A back-up Telco among managers of system operations was also convened for the duration to co-ordinate actions if needed.
- The setting up of extra training of control room operators and exercises on co-ordination procedures.
- Some TSOs put more operators on duty.
- Some TSOs engaged preventive activation of national crisis policy.
- There was continuous communication between control rooms in CE and the Nordic system in case support power was needed.
Figure 3 shows the aggregated PV feed-in during the solar eclipse from a subset of European TSOs who could provide data on a quarter hourly basis. In the case of clear sky conditions, the PV feed-in would likely be higher than shown in the figure.
The estimated PV injection at the start of the eclipse (about 09:30 Central Eiropean Time) was approximately 22 GW. During the maximum of the solar eclipse at 10:00 CET the feed-in decreased to approx. 14 GW. The estimated PV injection at 12:00 CET was 35 GW. Hence the change in injection between 10:30 and 12:00 was around +21 GW. Table 2 provides an overview of the influence of the solar eclipse on the PV feed-in of European TSOs. For some TSOs this is an estimate, and for others it is not known owing to the low amount of installed PV on their grids.
The costs to the TSOs of purchasing reserve power are shown in Table 3. It should be noted that the disconnection of PV during the eclipse resulted in higher power prices during that period.
During normal operation, frequency has to be managed within a range of ± 50 mHz, but there are other limits set to avoid wide area incidents caused by frequency disturbances. Some generators start disconnecting at 50.5 Hz, but in some zones 50.2 Hz is the frequency leading to automatic disconnection of PV generation. 49 Hz is the frequency at which automatic load shedding starts. The evolution of system frequency between 08:00 and 12:00 on the day is shown in Figures 4 and 5.
The frequency quality during the eclipse timeframe was very good. The maximum absolute frequency deviation from the set point observed was 48 mHz (between 49.968 Hz and 50.048 Hz), therefore at no point was the acceptable range of ± 50 mHz exceeded.
The much smaller GB system routinely experiences larger frequency fluctuations. Statutory frequency limits are ± 0.5 Hz, and operational standards allow 1500 excursions per year outside the operational targets of ± 0.2 Hz. GB frequency was well controlled throughout the event, only briefly exceeding its targets at 0800 CET.
The Nordic region would not be directly affected but its TSOs established in November 2014 a working group to ensure that Nordic measures and communication would be capable of providing support, if needed, for Continental Europe. This WG determined that CE would probably not need direct support but that in the event of blackouts steps should be taken to ensure that the Nordic system was not put in jeopardy, and would then be in a position to help re-energise CE. Based on dynamic power system analyses, the available transfer capacities given to the market were 800 MW import to and 3500 MW export from NSS towards CE. All automatic supporting functions on the interconnections were in operation to be able to supply a possible imbalance in Continental Europe.
Forecasts for Great Britain assumed demand suppression owing to people being outside watching the event. During the partial eclipse in August 1999 there was a 3 GW drop in demand. Forecasts for 2015 assumed 40% of the 1999 effect, based on the level of media interest and weather forecasts for the day. In the event, flattening of the demand curve at peak suggested a small demand suppression of around 10% of the 1999 effect coupled with a 1300 MW demand increase, predominantly owing to the increased lighting load.
The loss of embedded PV generation caused around 1 GW of demand increase, around 200 MW higher than forecast and was dealt with by the control room in real time using pumped storage. Plans for the event had ensured that maximum possible pumped storage was available - at one point all six Dinorwig machines were generating at the same time, which is unusual.
Wind generation during the eclipse dropped by around 10% (500 MW) as forecast, due to the fall in wind speeds associated with eclipses.
Controllability of PV
It was shown to be possible to disconnect in advance part of the installed PV production. This showed promising results,, but to use such a practice on a large scale certain factors need to be considered - the exact amount of PV feed-in that will be switched off, the timing of the switching off and on to the grid, the steps that need to be taken during the switching, and the retroactive effect of shutdowns on the system.
Visibility of PV generation
It became clear that an accurate description of the installed PV capacity and their capabilities is needed for accurate forecast studies. Real time measurement of the dispersed PV generation is the key to adapting the operational strategy in real-time.
The German TSOs had instructed power plants to maintain continuous operation, which had a positive impact on the available control power at quarter hour intervals.
The hourly day-ahead market was mainly unaffected by the eclipse. German TSOs successfully marketed PV output in a first step at the hourly market and in a second step at the quarter-hour market.
In case of high demand or supply, there is a de facto quarter hour market (OTC and power exchange) in Germany, Austria and Switzerland that can provide significant contributions for intra-quarter- hourly compensation. This solution is a fine-tune balancing by the TSO.
The quarter-hour market showed large spreads. A coupling of quarter-hour markets should contribute to increased liquidity of the market and reduce these spreads. At the same time the quarter-hour trading should be combined with the hourly market.
Thanks to the large reserve, all European TSOs managed to balance their systems and to keep their individual ACEs close to zero in real time (ie, faster than 5 minutes). Nevertheless, these additional reserves have a significant cost, and this solution must be kept for exceptional situations.
In the future, fast gradient changes of PV feed-in are expected owing to the increasing PV capacity on the grids. Solar generation, which now covers 3.5% of the EU demand for electricity, could cover 7% in 2021 and 10% in 2026 going up to 15% by 2030, driven by energy policies and lower prices - in less than ten years, PV system prices have already decreased by around 75%. By 2025, large scale power system are expected to produce at 4 to 6 ct€/kWh, falling to 2 to 4 ct€/kWh by 2050.
Managing a solar eclipse of a significant scale with 90 GW took a year's meticulous preparation and co-ordination. Managing similar eclipses in 2021 and 2026 with twice and three times as much PV in Europe calls for change and rapid action in key areas.
SolarPower Europe and ENTSO-E see five ingredients for a successful management of eclipses in the immediate future:
- The speedy adoption and implementation of all Network Codes.
- The adoption of regional co-ordination initiatives.
- An active and empowered customer participation in all markets.
- New system services provided, notably by PV.
- Enhanced TSO/DSO co-operation.