Main findings of MAF 2019
As a preliminary remark, it should be noted that the first target year, i.e., 2021, was chosen due to its temporal proximity as well as its ability to anticipate the system’s adequacy situation in the short term without overlapping with the Seasonal Outlooks. The MAF 2019 updates, and provides a comparison to, the results of MAF 2018, which focused on the target year 2020. The second target year, i.e., 2025, was chosen as a pivotal year for evaluating adequacy due to significant reductions in coal and nuclear capacity expected between 2021 and 2025. Hence, this target year is evaluated again with updated input data.
Generally, the results of MAF 2019 indicate low risks of inadequacy in the system, if input assumptions of the assessment materialize, with the exception of islands and a few continental market zones. To this end, several differences can be observed in these results and those for MAF 2018. Before elaborating on these differences in the following subsections, it is necessary to highlight the two main sources of these differences: (1) the updated input dataset provided by TSOs, which partially showcases the monitoring role of the MAF, and (2) improvements in ENTSO-E databases, i.e., in the thermal generation granularity, demand time series, hydro dataset and hydro modelling assumptions. While a one-to-one comparison is difficult due to the large number of interdependent assumptions and the complexity of the models, the results presented hereafter should be considered to be an updated and improved best estimate of the future adequacy landscape. More information on the updates in the input data and methodology can be found in Appendices 1 and 2, respectively.
How is LOLE calculated?
Simulations of each target year (e.g., 2021) are run multiple times with different Monte Carlo combinations of climate situations and random forced outage events based on available statistics and climate data.
For a large number of simulations, many hours correspond to cases without particularly stressful outages and climate conditions, i.e., situations in which all demands are met. On the other hand, some hours correspond to patterns of forced outages and climate conditions that are particularly stressful, i.e., situations in which demand cannot be met by the available supply or imports via interconnectors. In these situations, the demand that is not served is registered as Energy Not Served (ENS), measured in GWh over the course of a year.
The number of hours during which ENS occurs is recorded as Loss of Load Duration (LLD). The values for ENS and LLD are recorded for each region and for each Monte Carlo year. After performing multiple Monte Carlo simulations, the Expected ENS (EENS) and Loss Of Load Expectation (LOLE) are calculated per region and for the pan-European system. Additionally, the 95th percentile (P95) values are calculated. The P95 values are particularly useful for demonstrating the types of severe events that could happen once every 20 years.
Results of Base-Case scenarios
The development of different scenarios permits a comparison of multiple potential future states of the European power system. In the MAF, bottom-up scenarios collected from TSOs result in Base-Case scenarios for the analyzed time horizons. They will be complemented consecutively by sensitivities (e.g., further reductions in carbon-emitting generating capacities). This section presents the results of the MAF 2019 adequacy assessment while focusing on the Base-Case scenarios for the target years 2021 and 2025. In this executive summary, only LOLEs and the 95th percentiles³ are presented on maps of the simulated area in order to provide the main aspects of the adequacy situation within the MAF perimeter. However, more detailed results, including EENS values, can be found in Appendix 1, along with the results of the sensitivities performed within the framework of this MAF edition.
The results presented in the following paragraphs correspond to the resulting LOLE values per zone for the Base-Case scenario, which are the averages of the results of five different models. The models were built with different software tools and using two different levels of data granularity with respect to the thermal power plants (unit-by-unit and aggregated per technology). A non-zero value of LOLE in this report indicates only a resource inadequacy in the market. Any ENS due to transmission and distribution faults or demand and RES forecast errors is not accounted for. For more information on the methodology and probabilistic indicators, please see Appendix 2. Moreover, there are cases in which the results depend on the specificities of each country or zone. Thus, the reader should also consult Appendix 3, which contains country comments that will help the reader better understand the specific results for some countries and derive more accurate conclusions.
3 The 95th percentile (P95) of the LLD is the value above which lie only 5% of all observations and corresponds to a probability of occurrence of
1 in 20 years. Note that in some cases, observations may include very few but extreme values, resulting in P95 values lower than average (LOLE).
Base-Case results for 2021
Figure 2 shows the estimated levels of resource adequacy for the target year 2021. More specifically, for each zone, Figure 2 plots the LOLE (left-hand side) and the 95th percentile of the LLD that occurred among all simulated Monte Carlo years (right-hand side). The market-modelling results for the year 2021 do not indicate significant adequacy issues in most countries. As was the case in previous adequacy assessments, islands are vulnerable to loss of load. Thus, in Figure 2, Malta has high values for both LOLE and the 95th percentile of the LLD. Adequacy issues are also observed in Sicily, where the LOLE reaches 4 hours, on average, and the 95th percentile of the LLD is 25 hours. In continental Europe, a LOLE of around 4 hours was identified in France. The remaining zones follow, all with LOLE results below 2 hours per year. These results show that continental Europe’s interconnected system is expected to be adequate in 2021.
Even though MAF 2018 focused on the target year 2020, while MAF 2019 focuses on 2021, it can still be insightful to compare the results for these two scenarios due to their temporal proximity. Comparing the results of MAF 2018 with this edition (Figure 3), it is observed that the adequacy situation is quite different in some zones. For the zones consisting of the United Kingdom (UK; both Northern Ireland and Great Britain), Finland, Greece (including the explicitly modelled zone of Crete), Bulgaria and Cyprus, the adequacy issues found in MAF 2018 have disappeared. Regarding the UK and Greece, increased thermal capacity is anticipated for target year 2021 as opposed to the 2020 scenario, explaining the positive results in Figure 2. Furthermore, Finland anticipates a higher penetration of Renewable Energy Sources (RES), i.e., wind and solar, in 2021. In addition, the new hydro database had a considerable impact on the hydro capacity in Bulgaria. Lastly, in the case of Cyprus, the anticipated increase in thermal capacity, along with its updated demand profile, eliminated the expected adequacy risks for this island country.
On the other hand, it is worth noting that the LOLE increased in France by approximately 2 hours between MAF 2019 and MAF 2018. This result is linked to the planned decommissioning of two nuclear units by mid-2020 and to the postponed commissioning of a new nuclear plant to a later date, i.e., 2023. Furthermore, Sweden’s LOLE values increased by 1.7 hours in the 2021 Base-Case scenario compared to 2020.
Figure 2: Loss of load expectation (LOLE) values for the 2021 Base-Case scenario. Circle radii reflect the magnitudes of the LOLE values for the corresponding zones. Zones with missing circles have LOLE values of less than 0.5 h. *In these maps, outliers were removed before averaging the results of all tools for the zones consisting of France, Sicily, Malta and South Norway. Input data for Iceland have not been updated since MAF 2018, thus the outcomes remain the same.
Figure 3: Comparison of LOLE values between MAF 2018 and MAF 2019 for target years 2020 and 2021, respectively. Circle radii reflect the magnitudes of the LOLE values for the corresponding zones. Zones with missing circles have LOLE values of less than 0.5 h. *In the MAF 2019 maps, outliers were removed before averaging the results of all tools for the zones consisting of France, Sicily, Malta and South Norway. Input data for Iceland have not been updated since MAF 2018, thus outcomes remain the same.
Base-Case results for 2025
Figure 4 presents the results for Base-Case target year 2025. As noted above, the magnitude of a LOLE value is indicated by the radius of the circle on the map. Observing the map, one notices that only a few zones have LOLE values greater than 3 hours. More precisely, in the 2025 scenario Malta presents high adequacy risks, followed by Cyprus and Sicily. In continental Europe, very limited adequacy risks are predicted for the 2025 target year, provided that the input assumptions taken for the different countries materialize. With the exception of Turkey and Lithuania4, all zones have LOLE values of less than 3 hours, including France, which had a higher LOLE value in the 2021 scenario. The latter result for France is linked to the planned commissioning of a new nuclear power plant and the development of RES, Demand Side Response (DSR) and interconnectors. Naturally, in terms of the P95 values on the map (Figure 4, right-hand side), the number and radii of the circles is expected to increase but still remain insignificant, with the exception of the islands.
Plotting the results of the two investigated scenarios, 2021 and 2025, side-by-side (Figure 5) allows a better exploration of the evolution of adequacy, as anticipated by the MAF models. For a few zones, it is observed that changes between 2021 and 2025 have impacts on ensuring adequacy. This is the case for Lithuania, Turkey and Cyprus, where LOLE values show increases of over 6 hours. Other examples with lower LOLE value increases are Ireland, Belgium and Italy. The reader is invited to read Appendix 1 as well as the country comments in Appendix 3 to better understand the specific results.
As mentioned in the introduction, the target year 2025 is currently of specific interest for adequacy assessments due to the number of changes that are anticipated with respect to thermal decommissioning in numerous countries. For this reason, 2025 was evaluated in both MAF 2018 and 2019. The results for both MAF versions are presented in the maps in Figure 6.
4: Due to a data issue, net generating capacity for Lithuania includes an additional 400 MW of hydro turbine capacity, which should have been removed and reserved for providing Frequency Restoration Reserves (FRR). In MAF 2019, reserved capacity should not be considered as being available for assessing adequacy.
Figure 4: Loss of load expectation (LOLE) values for the 2025 Base-Case scenario. Circle radii reflect the magnitudes of the LOLE values for the corresponding zones. Zones with missing circles have LOLE values of less than 0.5 h. *In these maps, outliers were removed before averaging the results of all tools for the zones consisting of Cyprus, Sicily, Lithuania and Turkey. Input data for Iceland have not been updated since MAF 2018, thus outcomes remain the same.
Figure 5: Comparison of LOLE values between the two target years, i.e., the 2021 and 2025 Base-Case scenarios. Circle radii reflect the magnitudes of LOLE values for the corresponding zones. Zones with missing circles have LOLE values of less than 0.5 h. *In these maps, outliers were removed before averaging the results of all tools for the zones consisting of Cyprus, Sicily, Lithuania and Turkey. Input data for Iceland have not been updated since MAF 2018, thus outcomes remain the same.
One of the first observations one can make is that the results look mixed. A number of countries are in better situations, according to the latest findings in MAF 2019, including Poland, Ireland, the UK, Cyprus and the island of Crete. On the other hand, for some zones, e.g., Latvia, continental Italy and Lithuania, there are increases in the expected LOLE hours. The reader can consult the country comments in Appendix 3 for detailed national comments on the results for some of the zones.
The most significant improvements concern Ireland, which falls below the 3 h threshold, as well as Cyprus and Crete, where improved demand modelling and increased interconnection capacities had clear impacts on results.
Figure 6: Comparison of LOLE values between MAF 2018 and MAF 2019 for the target year 2025 Base-Case scenario. Circle radii reflect the magnitudes of the LOLE values for the corresponding zones. Zones with missing circles have LOLE values of less than 0.5 h. *In these maps, outlying results were removed before averaging the results of all tools for the zones consisting of Cyprus, Sicily, Lithuania and Turkey. Input data for Iceland has not been updated since MAF 2018, thus outcomes remain the same.
ENTSO-E and the participating TSOs have followed accepted industry practice in the collection and analysis of available data. While all reasonable care has been taken in the preparation of this data, ENTSO-E and the TSOs are not responsible for any loss that may be attributed to the use of this information. Prior to taking business decisions, interested parties are advised to seek separate and independent opinions with respect to topics covered by this report and should not rely solely upon data and information contained herein. Information in this document does not amount to a recommendation with respect to any possible investment. This document is not intended to contain all the information that a prospective investor or market participant may need.
ENTSO-E emphasises that ENTSO-E and the TSOs involved in this study are not responsible in the event that the hypotheses presented in this report or the estimations based upon these hypotheses are not realised in the future.