Wind Power Generation Scheduling Accuracy in Europe: An Overview of ENTSO-E Countries

Abstract

Despite the rapid spread of the use of wind energy to generate electricity, harnessing this energy source remains a great challenge due to its stochastic nature. One way of dealing with this is to prepare accurate wind power forecasts. This paper explored the accuracy of day-ahead and intraday scheduling of energy generation of the onshore and offshore wind farms of the member countries of the European Network of Transmission System Operators (ENTSO-E) in the period from 2013 to 2021. The precision of the scheduling activities showed a varying picture: the onshore wind farms of Germany, Spain, France, and Sweden produced more precise forecasts than others, with annual downward and upward regulatory needs between 0.8% and 14.4%, and from 0.8% to 6.5%, of the yearly energy generation, respectively. In certain countries, however, the forecasts were less accurate, with discrepancies exceeding 41% for downward and 132% for upward regulation. As for offshore wind farms, the annual downward and upward regulatory needs ranged between 0.9% and 61.7%, and from 1.3% to 44.1%, respectively, with Germany and Denmark producing the most accurate schedules. The innovative novelty and practical contributions of this study are that it determines and presents information related to the accuracy of the day-ahead and intraday wind power generation forecasting of the ENTSO-E countries, which is of practical relevance to the transmission system operators (TSOs), the main actors in the energy market and the decision-makers, too. This information may also help investors who invest in onshore and offshore wind farms with the economic aspects, and it may also greatly contribute to the market-related development of the management systems of energy storage solutions related to these technologies.

1. Introduction

1.1. Challenges Related to Integrating Variable Renewable Energy Sources into the Network

The rapidly growing VRE penetration around the world is supported by new regulations and is also a drop in the costs of converter-interfaced technologies. Nevertheless, the increasing amounts of VRE continue to challenge the conventional practices of system planning and system operation [1], and maintaining the balance of demand and supply is also problematic due to the intermittency of VRES and the geographical limitations of their availability [2]. As long as VRE penetration is relatively low, the energy produced is immediately consumed. However, in the case of higher penetration, the issue of imbalance between power generation and consumption becomes increasingly important, as periods with favorable weather are characterized by a surplus of available energy, while those with adverse weather are characterized by greater demand and less available power. Satisfying demand that is not fulfilled by energy generated from VRES at the time of need, and in an appropriate way, is only feasible by the deployment of dispatchable power production and/or other sources of flexibility [3,4].

The handling of the uncertainties inherent in the use of VRES for power generation is unimaginable without forecasting, without which it is unfeasible to operate, manage, or plan energy systems well [5]. As the information on VRE tends to be highly variable, often even capricious by nature, the prediction of renewable energy is still a rather difficult task [6]. The power generation patterns of the different VRES are also very different due to their specific features, which raise different issues in terms of their integration and balancing demand and supply. Electricity from wind energy is practically determined by wind velocity, and does not show the distinctive daily and annual patterns of solar power. Its drawback, however, is that wind speeds may change very abruptly, so the electricity generated by wind farms is difficult to forecast, which makes it challenging to balance supply and demand. As for solar power, its main determining factor is—of course—the quantity of solar irradiation. On the one hand, this is rather calculable due to the patterns related to the cycles of day and night and the changing seasons, while, on the other hand, the actual irradiation tends to be variable and irregular with differing degrees of uncertainty resulting from various weather events, including precipitation and cloud cover [7].

Although a number of measures, including adjustments to the market and the grid, feed-in regulations, and highly developed methods for forecasting demand and supply, have been introduced, triggered by the spread of the use of VRES [8], the majority of these have been reactionary responses aiming at the higher levels of the system, mostly TSOs and distribution system operators (DSOs). Moreover, it seems that they have been sufficient only in the case of relatively small proportions of VRE and a slower pace of development. What the implications of the greater penetration of decentralized power generation from VRES on energy policy, system integration, and the relationships of TSOs, DSOs, and prosumers will be are, however, still hotly debated [9,10]. Inadequate policy coordination, in turn, may cause occasional threats to the appropriate operation of the system, e.g., congestion in the grid and capacity constraints at all levels of the system [7,11]. Although the reformation of policies affecting the meso- and micro levels is still not a priority contrary to the macro level, balancing problems related to the deployment of VRES do arise at the lower levels too. This means that every level of the system deserves the same attention when it comes to policymaking concerning the structural issues of balancing the grid [7].

1.2. A Concise Introduction to the Operation of the Electricity Market

The electricity market is an intricate system with processes taking place at several levels to make sure that the supply of electric energy can satisfy the ever-changing needs of consumption at all times. The place where transactions are carried out first is the so-called daily (spot) market, where sellers and buyers have to make bids for the twenty-four hours of the next day, before gate closure. In the event of any deviation from an agreement reached here, the involved parties are financially liable. Corrective actions can be made in the intraday markets, which are integrated into certain electricity pools. The real-time balance of energy production and consumption is guaranteed by the regulation market, whose management is the responsibility of the system operator. In order to be able to deal with fast variations in load and unanticipated problems of generation capacity, system operators have reserves for deployment if needed [12,13]. The use of energy storage facilities provides the electricity market with further possibilities [14,15]. The broad spectrum of issues, such as the long-term horizon schedule of the daily market, the various sessions in the ID market, the deviation management market, the regulation service market, and real-time load sharing, necessitates a control algorithm with multiple layers suitable for the different timescales of the electricity market [13].

The Electricity Markets of the European Countries

Hu et al. provide a detailed account of the structure of the European electric energy market and its integration in their paper published in 2021 [16]. The liberalization of the electricity market in the European Union started in 1996, marked by the first European Directive concerning the liberalization of the electricity market. While the liberalization process was taking place at the level of the member states, the EU also introduced numerous regulations to create a unified internal electricity market to ensure the secure supply of electric energy at reasonable prices [17]. Based on the ENTSO-E, the market of electric energy comprises multiple separate markets in European countries: the forward market, the day-ahead (DA) market, the intraday (ID) market, the balancing market, and the imbalance settlement [18].

By trading mid-to-long-term products or electricity derivatives, it is in the forward markets where the uncertainties, characteristic of the DA market, are lessened to some degree. The liquidity of the forward markets of the different European nations shows great variation, based on information from the Council of European Energy Regulators (CEER) and the European Union Agency for the Cooperation of Energy Regulators (ACER). For example, the relatively large bidding zones (the geographical areas where the actors in the market can exchange energy without capacity allocation [19]) of Poland and Spain have comparatively low liquidity, while those of the United Kingdom (UK), France (FR), Germany–Austria–Luxembourg (DE-AT-LU, later only DE-LU following 2018) are characterized by the highest liquidity [20].

Between consumers and producers, the settlement of electricity delivery commitments for the following 24 h takes place in the DA market. It is also in this market where market clear prices are generally regarded as the most significant references in terms of electricity prices. The creation of a unified European cross-zonal day-ahead electricity market has been facilitated by the single day-ahead coupling (SDAC) initiative in the European Union [21]. The calculation of the prices of electric energy and the allocation of transmission capacity across frontiers is assisted by the price coupling of regions and the pan-European hybrid electricity market integration algorithm (PCR EUPHEMIA) [22].

Short-term adjustments if needed during the hours when the day-ahead market is closed can be made in the ID markets. These are continuous markets for buying, selling, or auctions around the clock [16].

In the balancing or ancillary services market, service providers, such as power demand response facilities, battery energy storage systems and generators, sell their services to the system operator. The services available here (e.g., black start services, frequency control, voltage support, automatic islanding, etc. [23]) are meant to serve power system security and reliability [24]. The significance of these services is considerably heightened by the fact that, during the process of imbalance settlement, the so-called balance responsible parties (market participants or their representatives) are responsible for any imbalances in their portfolios financially [18].

The present practice is that solar and wind power generation (MW) forecasts are prepared for each bidding zone for each market time unit of the next day, and the data need to be made public before 18.00 (Central European Time) of the day before the delivery. It is also required that this information be systematically updated and published for the duration of intra-day trading. Bidding zones with an annual wind or solar power feed-in of more than 5% are obligated to provide data. The obligation of data provision concerning all bidding zones applies to those ENTSO-E member states where the amount of annual feed-in of wind or solar power generation exceeds 1% [25].

1.3. Challenges of Forecasting Wind Power

As the proportion of electricity from VRES is rising in the world’s energy grids, enhancing the precision of predicting renewable energy is becoming essential to such tasks as planning, operating, and managing power systems [26]. Nevertheless, the forecasting of wind power, regardless of whether it is offshore or onshore, continues to pose a great challenge because of its variable, sometimes even seemingly erratic nature [6]. It is little wonder that both the volume and the depth of research on forecasting wind power are rapidly increasing currently [27].

According to the common categorization, the methods of predicting wind power fall into two basic types: physical and statistical. The techniques used traditionally are mostly based on some time series model, for example, the model of the autoregressive integrated moving average [28], the autoregressive model (AR) [29], or the autoregressive moving average model [30]. However, in the case of systems with linearized simplification, these well-established methods can work well, problems arise when it comes to the provision of accurate forecasts for wind energy, characterized by its innate nonlinearity. A more recent development is the large-scale deployment of methods based on artificial intelligence, which is thought to be able to discover complex nonlinear relationships on the basis of past data [31]. Several other techniques, such as artificial neural networks (ANNs) [32] and support vector machines (SVMs) [33], have also been used to make wind power forecasting more accurate. Overall, it can be stated that statistical methods have proven to be appropriate for short-term forecasting [34].

The method of spatial correlation (SC) is also worthy of mentioning here, as it differs from the more traditional statistical methods, as it can also take the interactions of nearby wind farms into account on the basis of temporal correlations [35] to improve the precision of forecasting [36]. SC techniques utilize information from adjacent wind farms to create a model for the wind farm in question [37]. It is not surprising that the deployment of this novel method for predicting wind speeds and power has become the subject matter of many research studies [38,39].

The other main category, i.e., physical methods, is primarily built on numerical weather prediction (NWP) and it takes the producer’s power curves into account [40], which can lessen the problem of missing past data. These methods also consider a number of physical features, such as fluid dynamics, thermodynamics, surface roughness, and obstacles, in order to create physical models for wind turbines. The refined wind speed data are then combined with the producer’s wind power curve to forecast the wind power. However, such wind power forecasting is prone to significant error due to the slow speed of NWP updating, resulting in time lags compared to the actual changes. Because they rely on highly complex computing, the use of NWP models is likely to remain limited to the short-term prediction of wind power [34].

Another area that has been the focus of research lately is hybrid forecasting methods combining the practical aspects of different techniques [41]. The combinations of the two basic types of methods, i.e., physical and statistical, can operate better than the separate conventional methods [42] because their limitations are dealt with systematically in the combined model [34].

Overall, it may be concluded based on the relevant literature that a wide variety of methods is used for predicting wind power, including artificial intelligence, time series, SC, and hybrid methods. Although historical or NWP data are often used in various studies, it must be noted that these may be negatively affected by the actual data [43], and enhancing their accuracy is a significant challenge [44]. In spite of the number of studies suggesting ways to deal with the inaccuracy issues of NWP [45], many research studies still choose to disregard them [46]. Although ANNs are extensively used for wind power forecasting, the traditional ANN model also has disadvantages: slow convergence speed and over-fitting. Compared with ANNs, the Gaussian process (GP) model has a number of advantages thanks to its well-founded framework, which allows the identification of the relationship between input and target variables [47]. This model has potential for the simplification of forecasting since it is an effective nonparametric, nonlinear, and probabilistic method of prediction, which has a smaller number of parameters compared to other statistical models [48]. In addition, the GP model is self-adaptive to gain hyperparameters, and its use is flexible too [49]. Nevertheless, the forecasting capability of the traditional GP model, which has only one type of kernel function, may be limited, as wind speed generally has a strong variability. On the other hand, however, by the selection of the optimal kernel function scheme from a range of several kernel function types, forecasting can produce more accurate results [34].

1.4. The Significance of the Present Research concerning Wind Power Generation Forecasting

The rapid global growth of installed wind power capacities is the result of the decreasing costs of such investments, on the one hand, and the international efforts to decarbonize the world’s energy systems, on the other hand. The difficulties inherent in the variable nature of the wind can be dealt with by the accurate forecasting of wind power [34]. Nowadays one can read about numerous research studies on energy generation forecasting algorithms, systems, and services for onshore and offshore wind farms, but these do not reveal much about the accuracy of the scheduling practices prevailing in particular countries. Consequently, the present study is not only timely but also fills a gap. This paper is innovative and novel in the sense that this is the first work that analyzes the accuracy of the day-ahead and intraday onshore and offshore wind farm forecasts of the ENTSO-E countries, as well as its development over the examined years, not only from a theoretical but also from a practical point of view. The research sheds light on the following:

• The start dates of the examined nations’ data provision of the actual and forecasted power of the onshore and offshore wind farms for the ENTSO-E system;
• The extent of the analysis of the data in the particular years;
• The features of the joint distributions of the real and forecasted power data of the onshore and offshore wind farms in the examined nations during the studied period;
• The amounts of the downward and upward regulation needs of the onshore and offshore wind farms compared to their yearly energy generation.

The basic premise of the research is that every commercial energy producer in the examined countries is required to submit day-ahead and intraday schedules to the TSO. In the case of variable renewable energy sources, specifically in the case of onshore and offshore wind farms, however, it is impossible to prepare 100% accurate schedules for the TSOs due to the unpredictability of the environmental factors, which may result in unexpected expenses for all involved parties. Based on the above, the goal of this study is to survey the accuracy of the day-ahead and intraday scheduling of energy generation by the onshore and offshore wind farms of the ENTSO-E countries. The significance of this lies in the fact that the limitations of the countries’ forecasting methods are presented, which may provide important information concerning the economic aspects of onshore and offshore wind farm investments and the modernization of management systems related to energy storage.

2. Materials and Methods

2.1. The Source Databases of the Examined Data of the Offshore and Onshore Wind Farms

The majority of the data used in the study came from the ENTSO-E Transparency Platform [50]. It was thanks to regulation (EU) no. 543/2013 of 14 June 2013 [51] on the submission and publication of data in electricity markets that transparency in this field has witnessed such remarkable improvement in Europe recently. This regulation made it obligatory for data providers and owners in the European member states to provide the necessary data concerning the production, load, transmission, and balancing of electric power for the ENTSO-E Transparency Platform. This has a special significance for the Internal Electricity Market (IEM) as well as for the establishment of European wholesale markets that are expected to be liquid, efficient, and competitive. It is also essential for the fair competition of the players in the market and the prevention of any abuse of power. The electricity market information provided by the platform is useful for the future and helps efficient, integrated, and competitive energy markets to be developed all over Europe. Under the above-mentioned regulation, the ENTSO-E Transparency Platform started its operation, including the collection and publication of data from various providers of information (transmission system operators, power exchanges, etc.) on 5 January 2015 [52]. This, of course, also means that the first year from which ENTSO-E Transparency Platform data were examined in this research was the year 2015.

The European Transmission System Operators (TSOs) are independent of the other actors of the electric energy market. Their responsibilities include high-voltage electricity transmission and the provision of access to the grid to the players in the market according to transparent regulations and without discrimination. They also guarantee the security of supply by ensuring that the system is appropriately maintained and operates safely. In certain ENTSO-E member states, TSOs are also responsible for the necessary progress and improvement of the electricity network. The 34 ENTSO-E member states are the following (in alphabetical order): Albania (AL), Austria (AT), Belgium (BE), Bosnia and Herzegovina (BA), Bulgaria (BG), Croatia (HR), Cyprus (CY), the Czech Republic (CZ), Denmark (DK), Estonia (EE), Finland (FI), France (FR), Germany (DE), Greece (GR), Hungary (HU), Iceland (IS), Ireland (IE), Italy (IT), Latvia (LV), Lithuania (LT), Luxembourg (LU), Montenegro (ME), the Netherlands (NL), Norway (NO), Poland (PL), Portugal (PT), the Republic of North Macedonia (MK), Romania (RO), Serbia (RS), the Slovak Republic (SK), Slovenia (SI), Spain (ES), Sweden (SE), and Switzerland (CH) [53]. The investigations of the present study also covered the United Kingdom, as it provided data for the ENTSO-E Transparency Platform from 2015 until the middle of June 2021 [50].

Regarding Belgium, the source of the data was not the ENTSO-E Transparency Platform, but the TSO of the country, called the Elia Group (EG). The choice fell on this database because the EG has provided data on actual wind power and DA and ID generation forecasts since 2013 [54]. This allowed the analysis of two more years than the ENTSO-E database could have done, as its earliest data were only from 2015.

The research presented herein focuses on wind farms.

Table 1. The yearly changes in the capacities of the installed offshore wind farms in the examined countries, based on [50,54].

	Changes in the Offshore Wind Capacities in the Examined Years (MW)
	2013	2014	2015	2016	2017	2018	2019	2020	2021
BE	490	712	712	712	877	1178	1548	2254	2254
DE	N/A	N/A	993	3283	4131	5051	6393	7504	7774
DK	N/A	N/A	1271	1271	1271	1700	1700	1700	1700
FR	N/A	N/A	0	0	0	0	N/A	14	10
IT	N/A	2	0	0	0	0	0	0	0
NL	N/A	N/A	228	357	638	957	957	1709	2357
PT	N/A	N/A	0	0	0	0	0	0	25
UK	N/A	N/A	5041	5011	5471	6071	9379	10365	12160

and

Table 2. The yearly changes in the capacities of the installed onshore wind farms in the examined countries, based on [50,54].

	Changes in the Onshore Wind Capacities in the Examined Years (MW)
	2013	2014	2015	2016	2017	2018	2019	2020	2021
AT	N/A	N/A	2121	2497	2696	2887	3035	3133	3198
BA	N/A	N/A	N/A	N/A	0	51	87	87	145
BE	997	1123	1249	1249	1745	1979	2248	2416	2629
BG	N/A	N/A	358	701	701	700	700	700	701
CY	N/A	N/A	158	N/A	158	158	158	158	158
CZ	N/A	N/A	270	277	277	308	316	339	339
DE	N/A	N/A	37701	41168	47042	51633	52792	53184	54499
DK	N/A	N/A	3574	3574	3574	4423	4426	4402	4481
EE	N/A	N/A	307	375	384	487	462	329	329
ES	N/A	N/A	22740	22772	22813	22834	22961	24447	26664
FI	N/A	N/A	496	1082	1432	1908	2013	2145	2422
FR	N/A	N/A	10322	11761	13569	12518	13610	16578	17217
GR	N/A	N/A	1613	1875	1875	2228	2355	3153	3755
HR	N/A	N/A	429	489	537	582	616	739	796
HU	N/A	329	329	328	324	325	327	323	323
IE	N/A	N/A	1907	1920	1920	1920	1919	1919	1919
IT	N/A	8455	8457	8739	9024	9261	9617	10224	10302
LT	N/A	N/A	282	366	426	509	525	534	540
LU	N/A	N/A	120	124	124	154	154	154	167
LV	N/A	N/A	51	55	53	53	59	74	84
ME	N/A	N/A	N/A	N/A	72	72	118	118	118
MK	N/A	N/A	35	35	N/A	35	35	N/A	35
NL	N/A	N/A	2646	3284	3479	3675	3669	3973	4500
NO	N/A	N/A	873	873	873	1197	1230	3068	N/A
PL	N/A	N/A	3758	5494	6026	5791	5808	5953	6570
PT	N/A	N/A	4486	4617	5028	5073	5127	5181	5183
RO	N/A	N/A	2896	2938	2985	2987	2968	2972	2957
RS	N/A	N/A	N/A	N/A	17	25	398	397	397
SE	N/A	N/A	3500	5493	5960	6247	7506	9648	10017
SI	N/A	N/A	3	3	3	3	3	3	3
SK	N/A	N/A	3	3	3	3	3	N/A	N/A
UK	N/A	N/A	8420	9138	10800	13139	13633	13830	13932

show the development of the installed offshore and onshore wind farm capacities in the examined countries. It is important to note here that, on the one hand, only those countries whose capacity data were available for the examined period were included in the tables, and, on the other hand, the capacity data here belong to the power plants obligated to provide data for the TSOs of their countries. Therefore, the total capacity of the actually installed wind farms may differ from that of the wind farms registered in the ENTSO-E system.

The analyses established that wind farms play an increasingly decisive role in the energy production of the examined nations. This was the reason why the analysis of the yearly changes in the capacities of the onshore and offshore wind farms in the ENTSO-E countries became the focus of this research.

2.2. Using the Information from the Databases in the Study

In the case of a number of nations, 15, 30, or 60-min data are also available in an annual breakdown in the ENTSO-E Transparency Platform [50]. The data provision practices of the examined nations were not uniform, as some nations (e.g., Denmark and Estonia) started to provide data as early as 2015, while some others only did so in 2020 (Croatia) and 2021 (Bosnia and Herzegovina). The nations were analyzed and evaluated one by one in Section 3.1. Section 3.2 summarizes the actual power figures of the onshore and offshore wind farms as well as their DA and ID power generation forecasts, i.e., their schedule data. In the ENTSO-E Transparency Platform four main geographical categories are used:

• Countries;
• Bidding zones (BZ);
• Control areas (CA), which are grid areas with a single system operator;
• Market balance areas (MBA), i.e., areas with a standardized price for balancing energy [55].

This study concentrated on the analysis of national data, according to the notion of the country as used by the ENTSO-E. Nevertheless, the extraction of French ID level data required the use of BZ data, as France has only one BZ (BZN|FR), meaning that the BZ actual power and DA power production forecast data are the same as the national data. Conversely, the accessibility of these data sometimes differs in the case of Italy, where there are 19 BZs. From the point of view of the investigation, analyzing the data at the national level was favorable, since it made it possible to see the actual power and forecasted figures aggregated for each country. This information provides suitable insight into the onshore and offshore wind farm scheduling accuracy, characteristic in the countries in question. The actual power and DA and ID generation forecast data were downloaded and merged in a tabular form in an annual breakdown, if available in the databases.

As for Belgium, the database of the EG [54] contains the monthly 15 min of actual power and DA and ID power generation forecast data, as well as the figures of the monitored onshore and offshore wind farm capacity starting from 2013, which are downloadable in one. Regarding other nations, the EG does not provide such data. Although weekly data were also available, their examination was outside the scope of the present study, as it used monthly data aggregated in databases for the examined years for the purposes of the analyses herein. The reason why is that, instead of data from the ENTSO-E database, the data of the EG were used in the case of Belgium, where the latter contained more detailed information from a longer time period.

When DA, as well as ID forecasts, were also available for a country, both were compared with the actual energy production values, which allowed the examination of the precision of both types of forecasts. In the cases when either 15-min, 30-min, or 60-min data were missing the following rules were observed:

• No actual power data: when 15-min and/or 30-min and/or 60-min forecasts were available but the actual power data were missing, the given period was not included in the analyses.
• No DA schedule data: when real power figures were available, but the DA forecast data were missing, and the given period was not included in the analyses.
• No ID schedule data: when real power figures were available, but the ID forecast data were missing, and the given period was not included in the analyses.
• No DA schedule data, but ID data were available: only actual and ID power data were taken into account in the analyses.
• No ID data, but DA data were available: only actual and DA power data were taken into account in the analyses.

Following the guidelines above, the proportion of the data suitable for the analysis of the DA and ID time series within the periods having DA and ID power generation forecasts was determined for every nation in the study (

Table 3. The completeness of the databases of the DA and ID onshore wind farm forecasts in the examined years.

	Data Resolution (min)	Completeness of DA and ID Databases (%)
		2013		2014		2015		2016		2017		2018		2019		2020		2021
		DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID
AT	15	N/A	N/A	N/A	N/A	100	N/A	99.4	N/A	100	N/A	100	58.2	100	100	100	100	100	100
BA	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	40.7	41.2
BE	15	N/A	97.1	N/A	99.1	38.9	100	100	100	100	100	100	100	99.2	100	100	100	99.7	100
BG	60	N/A	N/A	N/A	N/A	58.6	N/A	98.1	N/A	99.7	N/A	100	N/A	98.4	N/A	99.2	N/A	98.6	N/A
CH	60	N/A	N/A	N/A	N/A	99.2	N/A	99.4	N/A	99.7	N/A	100	N/A	100	N/A	100	N/A	100	N/A
CY	30	N/A	N/A	N/A	N/A	44	N/A	99.6	N/A	93.1	N/A	76	N/A	67.7	N/A	61.1	N/A	N/A	N/A
CZ	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
DE	15	N/A	N/A	N/A	N/A	100	100	100	100	100	100	100	98.9	100	97.5	100	99.4	100	95.6
DK	60	N/A	N/A	N/A	N/A	86	N/A	99.7	N/A	99.7	N/A	100	N/A	100	N/A	99.4	N/A	98.9	N/A
EE	60	N/A	N/A	N/A	N/A	97.5	N/A	99.8	N/A	99.6	N/A	97.6	N/A	99.9	N/A	99.9	N/A	99.7	N/A
ES	60	N/A	N/A	N/A	N/A	99.8	N/A	100	N/A	100	N/A	100	88.2	99.9	99.9	99.7	100	100	100
FI	60	N/A	N/A	N/A	N/A	99.4	N/A	99.4	N/A	98.8	N/A	99.7	N/A	99.5	46.6	100	100	100	100
FR	60	N/A	N/A	N/A	N/A	99.5	N/A	97.6	N/A	99.8	N/A	99.6	48.3	99.4	70.8	98.3	69.8	98.9	70.8
GR	60	N/A	N/A	N/A	N/A	99.9	N/A	99.8	N/A	99.6	N/A	99.7	N/A	89.6	N/A	91.5	N/A	100	42.5
HR	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	56.6	50.3	100	100
HU	15	N/A	N/A	N/A	N/A	92.1	N/A	90.8	N/A	92.2	N/A	96.5	5.2	94.7	94.7	100	100	100	100
IE	60 & 30	N/A	N/A	N/A	N/A	96.9	N/A	97.5	N/A	98.1	N/A	99.7	N/A	99.8	N/A	98.6	N/A	97.1	N/A
IT	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	4.4	N/A	98.9	98.6
LT	60	N/A	N/A	N/A	N/A	96.3	N/A	91	N/A	97.2	N/A	98.6	36.1	99.4	99	100	100	100	100
LV	60	N/A	N/A	N/A	N/A	99.9	N/A	99.5	N/A	99.7	N/A	99.6	56.4	99.6	99.6	99.6	99.6	99.9	99.9
ME	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	49.6	50.1	41.1	98.6	53.5	97.5	100	100
MK	60	N/A	N/A	N/A	N/A	N/A	N/A	51.4	N/A	76.2	N/A	91.8	N/A	85.5	N/A	77.9	N/A	49	N/A
NL	15	N/A	N/A	N/A	N/A	99.4	N/A	100	N/A	100	N/A	99.7	72.3	100	100	100	100	100	100
NO	60	N/A	N/A	N/A	N/A	99.2	N/A	100	N/A	90.7	N/A	94.2	N/A	99.7	N/A	99.9	N/A	100	N/A
PL	60	N/A	N/A	N/A	N/A	94.8	N/A	99.7	N/A	100	N/A	100	33.2	99.7	100	100	100	100	100
PT	60	N/A	N/A	N/A	N/A	97.5	N/A	97.3	N/A	98.6	N/A	79.7	80	92.3	98.1	98.6	99.4	98.3	99.7
RO	60 & 15	N/A	N/A	N/A	N/A	98.7	N/A	99.9	N/A	99.9	N/A	99.3	N/A	100	N/A	100	100	96	96.8
SE	60	N/A	N/A	N/A	N/A	84.1	N/A	99.7	N/A	100	N/A	99.1	69.3	99.7	100	100	99.7	100	99.4
SI	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
SK	60	N/A	N/A	N/A	N/A	95.5	N/A	99.1	N/A	99.2	N/A	98.9	N/A	99.2	80	99.5	99.7	100	100
UK	60 & 30	N/A	N/A	N/A	N/A	89.9	N/A	91.5	N/A	98	N/A	99.6	4.8	98.7	70.1	98.4	69.9	42.1	30.3

and

Table 4. The completeness of the databases of the DA and ID offshore wind farm forecasts in the examined years.

	Data Resolution (min)	Completeness of DA and ID Databases (%)
		2013		2014		2015		2016		2017		2018		2019		2020		2021
		DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID	DA	ID
BE	15	N/A	98.8	N/A	99.4	38.9	100	100	100	100	100	92.9	92.9	90.9	91.7	93.8	93.8	89.1	89.4
DE	15	N/A	N/A	N/A	N/A	99.7	100	100	100	100	99.7	100	100	100	100	100	100	100	100
DK	60	N/A	N/A	N/A	N/A	86	N/A	99.7	N/A	100	N/A	100	N/A	100	N/A	99.4	N/A	99.4	N/A
NL	15	N/A	N/A	N/A	N/A	99.4	N/A	100	N/A	100	N/A	99.7	72.3	100	100	100	100	100	100
PT	60	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	57.7	57.9	98.3	99.7
UK	60 & 30	N/A	N/A	N/A	N/A	99.7	N/A	100	N/A	99.6	N/A	100	4.8	99.7	70.8	100	70.8	44.8	31.9

). It is also displayed in Table 3 and Table 4 which countries submitted data to the EG or the ENTSO-E, together with the starting years of the publication of DA and ID data regarding their onshore and offshore wind farms. The different colors in Table 3 and Table 4 are designed to aid the interpretation of the information:

• Green marks a complete database for the year in question (100%).
• Red indicates a particularly inadequate database for the year in question.
• The darkness of the shade of yellow turning orange is indicative of the inadequacy of the database of the given year, with darker colors marking less complete data.

Regarding Table 3 and Table 4, Ireland and the UK submitted real power data in a 30-min breakdown, although the forecasted data were on a 60-min basis. Thus, the actual power figures and those in the forecasts had to be compared on an hourly basis. As for Romania, its wind farms had to provide 60-min data until 30 January 2021, which was later modified to 15 min. This fact, however, did not result in any difficulties for the analyses herein.

2.3. The Methods Used in the Analyses of the Study

The accuracy of forecasting is crucial for a number of reasons, including the provision of indispensable information to allow grid operators as well as system designers to create optimal wind farms, and to balance the electric energy supply and demand [56]. According to the literature available on the subject matter of wind power prediction, an array of methods are deployed for this purpose, such as SC methods, time series methods, AI methods, as well as hybrid methods. Most of the research studies on energy generation forecasting algorithms, systems, and services for onshore and offshore wind farms do not delve into the scheduling practices and their accuracy in the particular countries. Forecasts can be verified in various ways, which are also dealt with in numerous research studies [34,56]; however, due to the practical applicability of these, it is hard (or impossible) to compare one to another. This research investigates the accuracy of the day-ahead and intraday energy generation schedules of onshore and offshore wind farms in the ENTSO-E countries, by methods whose results are easy to interpret by the TSOs, the actors of the energy market, as well as the decision-makers. In their study published in 2020, Yang et al. [57] made an important step toward the standardization of the various VRE-based forecasting approaches, and their recommendations were also suitable for the comparison of the countries and the interpretation of the results gained. These suggested methods were also used in this study, with only minor modifications where required by their practical deployment. The analyses in this study followed the example of Yang et al. [57] and researched the forecasting accuracy by applying a method that displayed the joint distributions of the actual power figures and the data in the forecasts graphically. This technique allowed the visualization of all the time-independent forecast characteristics, as shown in the Section 3. Moreover, Figure 1, Figure 2, Figure 3, Figure 4, Figure A1, Figure A2, Figure A3, Figure A4, Figure A5, Figure A6, Figure A7, Figure A8, Figure A9, Figure A10, Figure A11, Figure A12, Figure A13, Figure A14, Figure A15, Figure A16, Figure A17, Figure A18, Figure A19, Figure A20, Figure A21, Figure A22, Figure A23, Figure A24, Figure A25, Figure A26, Figure A27, Figure A28, Figure A29, Figure A30, Figure A31 and Figure A32, in the same section present the information related to time dependence.In practical terms, the failure to generate the forecasted amount of power has important consequences, as the discrepancy equals the quantity of balancing energy, which arises as a need for regulation. This, in turn, can result in increased financial sanctions in the case of VRE technologies in certain countries in Europe today [15]. Provided a particular wind farm generates less power than it was scheduled to do, a requirement for positive regulation ensues. Conversely, if the actual power generation surpasses the quantity predicted, a need for negative regulation occurs [15].

As the current regulation stipulates that the producers of electricity shall be liable for any imbalances in the system, it is in their best interest to minimize the regulatory need by producing the most precise forecasts possible [58]. In this effort, the mean absolute error (MAE) may be of help to them, as it shows the average quantity of balancing power. It is calculated in the following way [59]:

$$\mathrm{MAE}=\frac{1}{N}{\displaystyle \sum }_{i=1}^N\left|P_{fc}-P_{act}\right|$$

(1)

$N$ refers to the total number of available data points, while $P_{fc}$ and $P_{act}$ to the forecasted and actual power.

Even though this index is unable to express the direction of the divergence, it is crucial because negative and positive imbalances result in different problems, which necessitate different solutions too. It is also natural that the prices of the two kinds of balancing energy differ as well, so it is essential that the MAE index be divided into a mean positive and a mean negative error (MPE and MNE), which is done as follows:

$$\mathrm{MNE}=\frac{1}{N}{\displaystyle \sum }_{i=1}^N\min \left(0, P_{fc}-P_{act}\right)$$

(2)

$$\mathrm{MPE}=\frac{1}{N}{\displaystyle \sum }_{i=1}^N\max \left(0, P_{fc}-P_{act}\right)$$

(3)

MPE refers to the mean positive error and MNE to the mean negative error: $\mathrm{MAE}=\left|\mathrm{MNE}\right|+\mathrm{MPE}$.

Table 1 and Table 2 provide information on the onshore and offshore wind farm capacities of the examined nations. Nevertheless, the amounts of energy generated by the monitored wind farms may differ from these to a significant degree [50]. The research was carried out according to the following procedure. First, the 15-, 30-, or 60-min of actual power data and the DA and ID generation forecasts of the wind farms of the examined countries were collected. From this information, the amounts of the downward and upward regulation requirements were determined as percentages of the annual energy production for each country, by using the MNE and MPE indices. The 15-, 30-, and 60-min power data were then converted to electric energy quantity figures, i.e., from MW to MWh. The practical benefits of the method deployed in the analyses are manifold:

• The size of the wind farm capacity does not affect the ability to reliably establish the positive and the negative regulatory needs, which can be deemed significant not only from energy aspects but also the perspective of the directions of network development.
• It becomes possible to assess, during the planning process of wind farms, how big the financial sanctions caused by the positive and negative regulatory needs will be, which may significantly influence the economic decisions related to the investments [15]. In Hungary, for instance, due to the growing surcharge for discrepancies between the 15-min real generation values and the schedules, there is a yearly growing financial pressure on scheduling groups that manage VRE systems and prepare DA and ID schedules for the TSOs. In the event of any deviation from the schedule, the TSO issues an invoice for the surcharge to the scheduling groups, which either pass on a significant proportion of it to the power plant owners or take it over from them, in return for a higher monthly fee for schedule preparation [60,61,62].
• Thanks so the method, it is possible to determine an index, which is easy to interpret by scheduling groups managing VRE systems, TSOs, the decisive actors of the energy market as well as the decision-makers.
• It provides information related to the technical and economic aspects connected to the integration of wind farms into the electric energy system and the building of energy storage systems [15,63].
• If energy storage facilities are used to enhance wind park scheduling accuracy, it may give information about the degree of the improvement of the negative and positive regulatory needs, on the one hand, and the feasibility of any financial benefits related to the improvement of the schedules [15].

3. Results

The visualizations of the joint distributions of the values of the actual power and those of the forecasts prepared for the schedules of the onshore and offshore wind farms of the examined countries, and the needs for downward and upward regulation calculated from the Entso-E data constitute the results of this research. For the convenience of viewing and comparison, the scatter plots of the joint distributions containing the DA and ID schedule discrepancies related to the same period are shown next to each other (Figure A1, Figure A2, Figure A3, Figure A4, Figure A5, Figure A6, Figure A7, Figure A8, Figure A9, Figure A10, Figure A11, Figure A12, Figure A13, Figure A14, Figure A15, Figure A16, Figure A17, Figure A18, Figure A19, Figure A20, Figure A21, Figure A22, Figure A23, Figure A24, Figure A25, Figure A26, Figure A27, Figure A28, Figure A29, Figure A30, Figure A31, Figure A32, Figure A33 and Figure A34). For the same reason, however, the annual downward and upward regulation need data are displayed separately (

Table A1. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Austria.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	15.9	15.8	14.0	15.2	10.3	10.6	9.8
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	11.6	14.2	12.4	12.6	12.0	14.6	14.6
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	10.7	8.0	7.7	6.5
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	12.5	8.1	9.8	15.8
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	4.5	2.2	2.9	3.3
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	0.1	4.0	4.8	−1.3

,

Table A2. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Bosnia and Herzegovina.

	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	6.3
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	44.8
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	6.3
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	44.9
The improvement of the annual downward regulation need when using ID forecasting (%)	0.1
The improvement of the annual upward regulation need when using ID forecasting (%)	−0.1

,

Table A3. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Belgium.

	2013	2014	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	N/A	N/A	5.1	7.3	6.5	5.9	5.7	10.9	12.7
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	N/A	N/A	17.6	15.1	20.6	17.1	16.1	10.2	10.5
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	14.9	7.3	8.3	5.3	5.7	5.0	4.6	7.9	8.0
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	9.0	10.8	12.7	12.6	14.1	10.9	10.3	6.9	6.8
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	−3.2	2.1	0.8	0.9	1.2	3.0	4.7
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	4.9	2.5	6.5	6.2	5.9	3.3	3.7

,

Table A4. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Bulgaria.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	12.6	13.5	13.9	11.3	14.6	15.5	14.6
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	36.6	31.2	29.2	40.3	34.0	19.8	23.6

,

Table A5. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Switzerland.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	18.9	22.8	17.6	15.5	41.5	data error	data error
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	46.1	36.1	29.6	67.9	8.9	data error	data error

,

Table A6. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Cyprus.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	21.6	18.2	19.4	21.9	19.8	17.1	N/A
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	23.4	21.4	20.9	16.7	19.5	20.5	N/A

,

Table A7. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Germany.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	6.2	6.4	5.5	5.7	5.3	4.9	5.0
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	6.5	5.2	4.4	4.5	4.1	4.3	4.7
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	data error	data error	data error	14.4	3.8	3.3	3.1
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	data error	data error	data error	2.4	2.6	2.7	2.9
The improvement of the annual downward regulation need when using ID forecasting (%)	data error	data error	data error	−8.7	1.5	1.6	1.9
The improvement of the annual upward regulation need when using ID forecasting (%)	data error	data error	data error	2.1	1.5	1.5	1.8

,

Table A8. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Denmark.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	5.4	4.6	6.2	5.3	5.3	5.6	7.9
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	12.4	10.3	3.7	4.5	4.4	13.2	12.7

,

Table A9. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Estonia.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	6.2	7.1	13.9	40.5	16.0	16.1	16.1
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	13.9	15.7	4.6	32.4	6.7	2.7	5.2

,

Table A10. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Spain.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	0.9	0.8	0.9	3.8	3.7	4.7	4.7
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	0.8	0.8	1.0	4.2	4.6	4.0	4.1
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	2.8	2.4	2.6	2.7
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	3.0	2.9	2.7	2.8
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.0	1.3	2.1	2.0
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.2	1.7	1.3	1.3

,

Table A11. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Finland.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	6.6	21.3	10.3	7.1	8.8	8.6	10.9
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	7.5	4.3	5.8	7.1	7.3	7.9	8.6
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	6.1	4.4	6.2
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	5.5	6.8	6.7
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	2.7	4.2	4.8
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	1.8	1.0	1.8

,

Table A12. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in France.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	5.7	5.5	6.2	7.5	5.2	5.7	7.1
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	4.6	5.7	4.2	4.5	6.2	5.2	5.7
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	6.1	4.2	4.4	6.3
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	5.4	5.9	4.6	5.5
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.3	1.0	1.3	0.8
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−0.9	0.3	0.5	0.2

,

Table A13. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Greece.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	7.9	3.6	5.1	3.2	5.8	4.7	5.6
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	11.8	17.3	15.9	15.9	10.7	14.4	8.2
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	N/A	N/A	4.2
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	N/A	N/A	7.7
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	N/A	N/A	1.5
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	N/A	N/A	0.4

,

Table A14. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Croatia.

	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	9.9	9.7
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	9.0	7.9
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	9.4	9.3
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	8.1	7.4
The improvement of the annual downward regulation need when using ID forecasting (%)	0.5	0.3
The improvement of the annual upward regulation need when using ID forecasting (%)	0.9	0.5

,

Table A15. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Hungary.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	13.4	15.8	13.1	15.4	14.9	10.3	11.1
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	22.5	19.6	23.3	26.7	22.7	20.0	11.3
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	7.2	14.9	9.7	9.5
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	27.0	22.7	18.8	8.1
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	8.2	0.0	0.6	1.7
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−0.3	0.0	1.3	3.2

,

Table A16. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Ireland.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	9.6	8.9	8.4	11.5	14.5	10.8	10.0
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	7.2	7.9	7.3	7.3	11.2	16.2	16.6

,

Table A17. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Italy.

	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	10.4	11.8
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	2.7	4.8
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	11.6
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	4.7
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	0.2
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	0.1

,

Table A18. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Lithuania.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	14.2	11.4	6.1	14.0	12.1	9.7	9.2
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	10.4	18.1	24.3	9.2	9.2	10.1	15.2
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	11.3	10.9	7.2	6.8
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	7.9	9.1	10.3	10.3
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	2.7	1.2	2.5	2.4
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.3	0.1	−0.2	4.8

,

Table A19. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Latvia.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	13.9	19.0	12.3	24.2	12.7	10.3	21.4
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	29.6	18.2	26.4	14.3	19.0	20.4	18.1
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	22.6	11.4	9.0	20.1
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	13.0	18.1	19.9	17.3
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.6	1.3	1.3	1.3
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	1.3	0.9	0.5	0.8

,

Table A20. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Montenegro.

	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	24.6	10.7	10.6	14.8
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	20.5	36.2	22.1	18.4
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	24.7	13.1	14.3	14.8
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	20.4	32.3	18.7	18.4
The improvement of the annual downward regulation need when using ID forecasting (%)	0.0	−2.4	−3.7	0.0
The improvement of the annual upward regulation need when using ID forecasting (%)	0.1	3.8	3.4	0.0

,

Table A21. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in the Republic of North Macedonia.

	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	33.4	32.0	34.0	49.1	38.0	26.5
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	115.3	132.4	44.6	42.3	25.0	22.9

,

Table A22. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Poland.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	10.4	10.3	12.7	11.2	8.4	8.4	9.0
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	2.0	1.4	0.2	1.1	4.0	3.7	3.9
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	9.1	7.6	7.0	7.4
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	1.4	2.4	2.6	2.8
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	2.2	0.9	1.4	1.6
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−0.3	1.7	1.2	1.1

,

Table A23. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Portugal.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	11.2	7.5	4.9	8.1	8.8	8.5	7.1
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	4.9	7.1	2.7	5.2	5.2	7.7	9.2
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	4.9	4.9	5.8	5.1
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	6.1	3.9	5.5	7.1
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	3.2	3.9	2.7	2.0
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−0.9	1.3	2.2	2.1

,

Table A24. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Romania.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	8.8	6.8	6.3	9.1	23.6	20.6	12.5
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	26.6	24.6	21.4	18.8	5.1	5.2	11.6
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	N/A	21.0	11.1
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	N/A	N/A	4.8	10.7
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	N/A	−0.4	1.4
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	N/A	N/A	0.4	0.9

,

Table A25. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in Sweden.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	5.5	5.0	2.5	5.0	5.2	1.5	1.9
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	2.5	3.6	2.5	2.6	2.6	3.4	3.9
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	2.9	1.7	0.6	0.9
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	5.4	2.3	2.6	2.9
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	2.1	3.5	0.9	1.0
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−2.8	0.4	0.9	1.1

,

Table A26. The proportions of the downward and upward regulation needs to the annual energy generation in the case of onshore wind farms in the United Kingdom.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	3.0	2.7	9.2	0.3	2.7	1.8	2.7
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	19.9	15.6	20.8	29.8	19.4	22.5	15.9
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	1.6	5.1	3.6	5.3
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	11.0	11.0	12.3	8.5
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−1.3	−2.4	−1.7	−2.6
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	18.8	8.3	10.2	7.4

,

Table A27. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in Belgium.

	2013	2014	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	N/A	N/A	10.0	12.5	9.6	9.9	9.3	5.5	3.1
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	N/A	N/A	9.3	10.4	11.9	11.2	14.1	20.0	30.4
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	13.0	9.1	10.2	10.0	7.7	7.2	6.5	2.7	1.1
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	10.9	16.6	10.7	7.7	8.1	7.8	11.2	15.5	27.6
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	−0.2	2.5	1.9	2.7	2.8	2.7	2.0
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	−1.3	2.7	3.8	3.3	3.0	4.6	2.9

,

Table A28. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in Germany.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	33.3	7.3	8.4	8.8	6.7	6.9	8.0
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	7.4	7.7	7.6	9.5	8.4	8.2	7.9
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	10.2	12.4	11.6	5.2	4.8	5.6	5.9
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	7.1	2.8	1.3	17.9	6.4	5.2	5.1
The improvement of the annual downward regulation need when using ID forecasting (%)	23.1	−5.1	−3.2	3.7	1.9	1.4	2.0
The improvement of the annual upward regulation need when using ID forecasting (%)	0.4	4.9	6.4	−8.4	2.0	2.9	2.8

,

Table A29. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in Denmark.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	10.2	9.2	7.0	7.0	5.2	8.1	26.0
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	5.2	3.9	2.9	3.5	8.2	8.6	2.9

,

Table A30. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in the Netherlands.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	39.6	59.0	61.7	36.5	12.9	25.8	50.4
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	12.8	2.8	1.5	5.4	15.2	8.0	1.5
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	26.7	12.9	25.9	50.4
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	7.7	15.3	7.9	1.5
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	9.7	0.0	−0.1	0.0
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	−2.3	−0.1	0.1	0.0

,

Table A31. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in Portugal.

	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	50.4	38.7
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	18.8	35.4
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	49.2	25.4
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	20.9	32.3
The improvement of the annual downward regulation need when using ID forecasting (%)	1.2	13.3
The improvement of the annual upward regulation need when using ID forecasting (%)	−2.0	3.1

and

Table A32. The proportions of the downward and upward regulation needs to the annual energy generation in the case of offshore wind farms in the United Kingdom.

	2015	2016	2017	2018	2019	2020	2021
The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling (%)	6.5	4.5	2.4	1.7	1.6	1.3	1.4
The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling (%)	10.8	19.3	25.9	33.4	41.7	44.1	40.8
The proportion of the downward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	0.9	1.4	1.2	1.2
The proportion of the upward regulation need to the annual energy generation in the case of ID scheduling (%)	N/A	N/A	N/A	29.4	40.2	43.9	40.4
The improvement of the annual downward regulation need when using ID forecasting (%)	N/A	N/A	N/A	0.8	0.2	0.0	0.2
The improvement of the annual upward regulation need when using ID forecasting (%)	N/A	N/A	N/A	3.9	1.5	0.2	0.4

). Due to the different data provision practices witnessed in connection with the examined nations, the results are presented in several subsections:

• Section 3.1.1 includes the countries without data provision.
• Section 3.1.2 contains those with severely faulty or incomplete data provision.
• Section 3.1.3 presents those nations whose wind farm scheduling data could be deemed adequate.

3.1. The Scheduling Features of the Onshore and Offshore Wind Farms of the Examined Nations

3.1.1. The Countries with No Data Provision

A careful study of the ENTSO-E Transparency Platform database revealed that certain nations did not submit either real or forecasted wind power generation data, or neither of the two, which prevented their analysis. These countries were:

• Onshore wind farms: Albania, the Czech Republic, Iceland, Luxembourg, Serbia, and Slovenia;
• Offshore wind farms: Albania, Austria, Bosnia and Herzegovina, Bulgaria, Switzerland, Cyprus, the Czech Republic, Estonia, Spain, Finland, France, Greece, Croatia, Hungary, Ireland, Iceland, Italy, Lithuania, Luxembourg, Latvia, Montenegro, the Republic of North Macedonia, Norway, Poland, Romania, Serbia, Sweden, Slovenia, and the Slovak Republic.

3.1.2. The Countries That Provided Significantly Inadequate or Incomplete Data

Scheduling can be regarded to be perfect when the dots in the scatter chart showing the joint distribution of the real and the forecasted wind power data lie on the diagonal. Ideal situations such as that, however, never take place, so dots lying on the diagonal signal some problem in the provision of data.

Among the countries involved in the investigation, there were some which submitted information related to wind power generation and/or forecasted schedule data to the ENTSO-E Transparency Platform, but their study indicated seriously flawed or incomplete provision of data. These were the countries below:

• Onshore wind farms: the Netherlands, Norway, and the Slovak Republic;
• Offshore wind farms: no country in this category.

Onshore wind farms.

The Netherlands.

In the case of the Netherlands, Figure A1 shows that until the end of the period examined, the DA and ID schedule energy generation forecasts contained values exceeding the actual data significantly, i.e., overscheduling occurred. This suggests that the actual data and the forecasted values belonged to onshore wind farms of different capacities, whose order of magnitude is illustrated by data from 2015, 2018, and 2021:

• Period: 24 August 2015 21:15–21:30 (CET), real power: 383 MW, DA forecasted power: 2196 MW;
• Period: 21 June 2018 18:30–18:45 (CET), real power: 972 MW, DA forecasted power: 4042 MW;
• Period: 1 November 2021 12:15–1 November 2021 12:30 (CET), real power: 1532 MW, ID forecasted power: 3941 MW.

Because of the scheduling issues above, the quantities of the downward and upward regulation requirements were not determined from the differences between the actual energy generation and the forecasts.

Norway.

Although there were data provisions for the ENTSO-E system on the part of Norway, it can be seen that the actual power data and the DA schedule energy generation forecast data were mostly identical during the period 2015–2021 (Figure A2). Therefore, this country was classified as a ‘country with severely inadequate or incomplete data provision.’ Due to the above scheduling problems, the amounts of the downward and upward regulation requirements were not established from the differences between the actual energy generation and the forecasts for this nation.

The Slovak Republic.

Slovakia was also included in the category of the ‘countries with severely inadequate or incomplete data provision,’ since there was data provision to the ENTSO-E Transparency Platform, albeit inadequate. During most of the period 2015–2021, there is a value of 0 in the database of the actual power, while 1 and 2 MW can be seen for the DA and ID values. Due to these scheduling anomalies, the quantities of the downward and upward regulation requirements were not determined from the differences between the real energy generation and the forecasts, and the distributions were not displayed either.

3.1.3. The Nations with Suitable Data for the Comparative Analysis of the Scheduling Characteristics of Their Wind Farms

Among the examined countries, there were 26 (Austria, Bosnia and Herzegovina, Belgium, Bulgaria, Switzerland, Cyprus, Germany, Denmark, Estonia, Spain, Finland, France, Greece, Croatia, Hungary, Ireland, Italy, Lithuania, Latvia, Montenegro, the Republic of North Macedonia, Poland, Portugal, Romania, Sweden, and the United Kingdom) whose onshore wind farm scheduling parameters proved to be suitable for analysis, while in the case of onshore wind farms, this was only true for 6 nations (Belgium, Germany, Denmark, The Netherlands, Portugal, and the United Kingdom). It was highlighted by the results of the investigation that under- or overscheduling was characteristic (to a lesser or greater degree) to the countries studied (clusters of dots below or above the diagonal in the visualizations of the joint distributions). The following section presents the characteristics of the onshore and offshore wind farm scheduling and the amounts of the downward and upward regulation requirements in the particular countries. The case of every examined nation is presented in a similar logical arrangement to enhance comparability.

Onshore wind farms.

Austria.

In the case of Austria, it can be observed that both under- and overscheduling characterized the DA and ID forecasts made for the onshore wind farms (Figure A3). In the case of ID forecasts, there were scheduling periods between 27 and 28 October 2021 when the value of the ID forecast was 10000 MW, but the actual data differed significantly from that. This is illustrated below:

• Period: 27 October 2021 15:30–15:45 (CET), real power: 4 MW, ID forecasted power: 10000 MW;
• Period: 28 October 2021 19:00–19:15 (CET), real power: 2024 MW, ID forecasted power: 10000 MW.

The situation outlined above suggests operational disorders at the system level, however, in the schedules submitted to the TSO, these forecasted values were taken into consideration. Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A1):

• In the case of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 9.8% to 15.9%, which could be improved by a few percentage points by the deployment of ID scheduling.
• As for the proportion of the upward regulation need for the annual energy generation in the case of DA scheduling, it varied between 11.6% and 14.6%, which could also be improved by some percentage points when using ID scheduling, in the majority of the cases.

Bosnia and Herzegovina.

In the case of Bosnia and Herzegovina, it can be seen that overscheduling was characteristic of the DA and ID forecasts made for the onshore wind farms; a significant part of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A4). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following observations can be made (Table A2):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling was 6.3%. Compared to the DA forecasts, the deployment of ID scheduling only resulted in minimal improvement.
• The need for upward regulation compared to the annual energy generation in the case of DA scheduling was 44.8%. The application of ID scheduling made the accuracy even worse, compared to the DA one.

Belgium.

Of the examined nations, Belgium was the country with the most detailed data available on onshore wind farms. Based on the results, (Figure A5), it can be seen that under- and overscheduling occurred in the case of both DA and ID forecasting for onshore wind farms. The proportions of the downward and upward regulation needs per year for energy production are shown in Table A3:

• In the case of DA scheduling, the proportion of the downward regulation need to the annual energy generation varied between 5.1% and 12.7%, which could be improved by some percentage points by using ID scheduling in the overwhelming majority of the cases.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 10.2% and 20.6%. Compared to the DA forecasts, the deployment of ID scheduling caused a maximum improvement of 6.5% of the annual upward regulation need.

Bulgaria.

In the case of Bulgaria, it can be observed that overscheduling was characteristic of the DA forecasts made for the onshore wind farms; the majority of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A6). Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A4):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling ranged between 11.3% and 15.5%.
• Compared to the downward regulation needs, the annual upward regulation requirement posed a greater challenge in terms of the management of scheduling, which might have been caused by the fact that those preparing the forecasts were not in possession of the weather data related to meteorological situations which proved to be more changeable than expected during the scheduling period. The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling varied between 19.8% and 40.3%.

Switzerland.

As for Switzerland, both under- and overscheduling could be observed in the case of the DA forecasts made for the onshore wind farms (Figure A7). During the examined period, it was only in 2020 and 2021 when the data provided was not adequate (the forecasted power figures had zero value), and this was the reason why this nation was not classified as one of the ‘countries with significantly inadequate or incomplete data provision’. It was due to this that the amounts of the downward and upward regulation requirements were not determined from the differences between the real energy generation and the forecasts for these two years. The proportions of the annual downward and upward regulation requirements for energy production were as follows (Table A5):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 15.5% and 41.5%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 8.9% and 67.9%.

Cyprus.

In the case of Cyprus, it can be observed that both under- and overscheduling characterized the DA forecasts made for the onshore wind farms (Figure A8). Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A6):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 17.1% and 21.9%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling ranged between 16.7% and 23.4%.

Germany.

One of the most accurate DA and ID forecasting mechanisms connected to onshore wind farms is seen in the case of Germany. It can be observed that, in the case of the DA forecasts, the dots are nearer to the diagonal, which indicates accurate forecasting (Figure A9).

For this country, appropriate ID forecast data are available starting from 2018, so the amounts of the downward and upward regulation requirements related to them were established for the years 2018–2021. The ID forecasts dated earlier than 2018 appear to be connected to onshore wind farms of different capacities, which resulted in underscheduling. This is illustrated below:

• Period: 1 June 2015 00:00–00:15 (CET), real power: 15052 MW, ID forecasted power: 6820 MW;
• Period: 7 February 2016 06:30–06:45 (CET), real power: 20098 MW, ID forecasted power: 10967 MW;
• Period: 23 February 2017 01:30–23 February 2017 01:45 (CET), real power: 26327 MW, ID forecasted power: 10535 MW.

Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A7):

• In the case of DA scheduling, the proportion of the downward regulation need to the annual energy generation varied between 4.9% and 6.4%, which could be improved to some extent by using ID scheduling in the overwhelming majority of the cases.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation ranged from 4.1% to 6.5%, which could be improved by the deployment of ID scheduling.

There may be several reasons for Germany’s forecasting accuracy. On the one hand, among the examined ENTSO-E countries, it has the most onshore wind farms over a large geographical area, and thus they offset one another’s inaccuracies to some degree. On the other hand, developments related to onshore wind farm power generation forecasting have a continuous history of more than 10 years there [64].

Denmark.

In Denmark, both under and overscheduling can be observed in the case of the DA forecasts made for the onshore wind farms. However, it is also to be seen that most of the dots are located near the diagonal, which suggests a relatively high level of forecasting precision (Figure A10). The proportions of the annual downward and upward regulation requirements for energy production were as follows (Table A8):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 4.6% and 7.9%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling ranged between 3.7% and 13.2%.

Estonia.

In the case of Estonia too, it can be seen that both under- and overscheduling were characteristic of the DA forecasts made for the onshore wind farms (Figure A11). This observation is also supported by the proportions of the annual downward and upward regulation requirements for energy generation (Table A9):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling was between 6.2% and 40.5%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling ranged between 2.7% and 32.4%.

Spain.

Similarly to the case of Germany, the results of Spain also indicate that this country can boast one of the most accurate DA and ID forecasting mechanisms related to onshore wind farms. It is to be seen that, in the case of the DA forecasts, the clusters of dots are near the diagonal, which signals accurate forecasting (Figure A12).

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 0.8% to 4.7%, which could be improved by a few percentage points by the deployment of ID scheduling.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation varied between 0.8% and 4.6%, which could be improved further by the deployment of ID scheduling (Table A10).

Similar to the case of Germany, there may be several reasons behind Spain’s accuracy in forecasting. On the one hand, it also belongs to those examined ENTSO-E countries which possess a large amount of onshore wind farm capacity (26.6 GW in 2021), spread over a vast geographical area, and thus the wind farms compensate for one another’s inaccuracies to a certain degree. On the other hand, developments related to onshore wind farm power generation forecasting have been going on dynamically for more than 10 years [65,66].

Finland.

In the case of Finland, the results show that the DA and ID forecasts made for onshore wind farms were characterized by both under- and overscheduling (Figure A13). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following observations can be made:

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 6.6% and 21.3%. Compared to the DA forecasts, the deployment of ID scheduling resulted in a maximum improvement of 4.8% of the annual downward regulation need.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling ranged between 4.3% and 8.6%, which could be improved even further by 1.8% when ID scheduling was applied, as shown in Table A11.

France.

Similarly to Germany and Spain, the French results also suggest that this nation possesses one of the most accurate DA and ID forecasting mechanisms connected to onshore wind farms. It can be seen that, in the cases of both the DA and the ID forecasts, the clusters of dots are located near the diagonal, which indicates accurate forecasting (Figure A14):

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 5.2% to 7.5%, which could be improved by a maximum of 1.3% by the deployment of ID scheduling.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation varied between 4.2% and 6.2%, which could be improved further by the deployment of ID scheduling in most of the cases (Table A12).

The reasons leading to France’s accuracy in forecasting must be similar to those in the cases of Germany and Spain. On the one hand, France is also one of the examined ENTSO-E countries that possess a large volume of onshore wind farm capacity (17.2 GW in 2021), scattered over a vast geographical area, with the wind farms offsetting one another’s inaccuracies to some extent. On the other hand, developments related to onshore wind farm power generation forecasting have been going on dynamically in France for more than 10 years too [65,66].

Greece.

In the case of Greece, it can be observed that overscheduling was characteristic of the DA and ID forecasts made for the onshore wind farms; a significant part of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A15). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following conclusions can be drawn (Table A13):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 3.2% and 7.9%. Compared to the DA forecasts, the deployment of ID scheduling could achieve a maximum improvement of 1.5% of the annual downward regulation need.
• Regarding scheduling, the volume of the upward regulation need was more significant than that of the downward regulation requirement, which may have been caused by the fact that those preparing the forecasts could not precisely calculate with the changeable weather conditions. In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation ranged between 8.2% and 17.3%, which could be improved further by the deployment of ID scheduling, albeit only to a negligible degree.

Croatia.

Both under- and overscheduling could be observed in the case of the DA and ID forecasts made for the onshore wind farms in Croatia (Figure A16). The proportions of the downward and upward regulation needs per year for energy production are displayed in Table A14:

• In the case of DA scheduling, the proportion of the downward regulation need to the annual energy generation varied between 9.7% and 9.9%, which could be improved to a negligible extent by using ID scheduling.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 7.9% and 9.0%. Compared to the DA forecasts, the deployment of ID scheduling could lead to a maximum improvement of 0.9% of the annual upward regulation need.

Hungary.

Overscheduling was characteristic of most of the DA and ID forecasts made for the onshore wind farms in a Hungary; a significant part of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A17). The proportions of the annual downward and upward regulation requirements for energy production were as follows (Table A15):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 10.3% and 15.8%. Compared to the DA forecasts, the deployment of ID scheduling could improve the annual downward regulation need even by a maximum of 8.2%.
• Regarding scheduling, the volume of the upward regulation need was more significant than that of the downward regulation requirement, which might have been caused by the fact that those preparing the forecasts were not in possession of the weather data related to meteorological situations, which proved to be more changeable than expected during the scheduling period. The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 11.3% and 26.7%, which could be improved by 3.2% maximum by the use of ID scheduling.

Ireland.

Based on the results, (Figure A18), it can be seen that under- and overscheduling both characterized DA forecasting for onshore wind farms in Ireland. The proportions of the annual downward and upward regulation requirements for energy production were as follows (Table A16):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 8.4% and 14.5%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 7.2% and 16.6%.

Italy.

In the case of Italy, the results indicate that underscheduling was characteristic of the DA and ID forecasts made for the onshore wind farms; a significant part of the dots are below the diagonal in the scatter plots of the joint distributions (Figure A19). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following observations can be made (Table A17):

• Regarding scheduling, the volume of the downward regulation need was more significant than that of the upward regulation requirement, which might have been caused by the fact that those preparing the forecasts were not in possession of the weather data related to meteorological situations that proved to be more changeable than expected during the scheduling period. The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 10.4% and 11.8%. Compared to the DA forecasts, the deployment of ID scheduling could only lead to a minimal improvement of the annual downward regulation need.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation ranged between 2.7% and 4.8%, which could be improved only to a negligible degree by the deployment of ID scheduling.

Lithuania.

In the case of Lithuania, the results show that the DA and ID forecasts made for onshore wind farms were characterized by both under- and overscheduling (Figure A20). The proportions of the annual downward and upward regulation needs for energy production also illustrate this observation (Table A18):

• In the case of DA scheduling, the proportion of the downward regulation need to the annual energy generation varied between 6.1% and 14.2%, which could be improved by some percentage points by using ID scheduling.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 9.2% and 24.3%, which could be improved by 4.8% maximum by the use of ID scheduling.

Latvia.

Both under- and overscheduling could be observed in the case of the DA and ID forecasts made for the onshore wind farms in Latvia (Figure A21). The proportions of the annual downward and upward regulation requirements for energy production were as follows (Table A19):

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 10.3% to 24.2%, which could be improved by a maximum of 1.6% by the deployment of ID scheduling
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 14.3% and 29.6%, which could be improved by 1.3% maximum by the use of ID scheduling.

Montenegro.

Both under- and overscheduling could be observed in the case of the DA and ID forecasts made for the onshore wind farms in Latvia (Figure A22). This is also verified by the proportions of the annual downward and upward regulation requirements for energy generation (Table A20).

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 10.6% and 24.6%. The application of ID scheduling either did not make any difference or even made the accuracy worse by up to 3.7% compared to the DA one.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 18.4% and 36.2%. Compared to the DA forecasts, the deployment of ID scheduling could improve the accuracy by up to 3.8%.

The Republic of North Macedonia.

One of the least accurate forecasting mechanisms is observable in the case of the Republic of North Macedonia (Figure A23). The analysis of the actual and the forecasted data could not reveal clearly whether they were related to differing onshore wind farm capacities. This was the reason why this country was not classified as one of the ‘countries with significantly inadequate or incomplete data provision’ (Table A21).

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 26.5% and 49.1%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 22.9% and 132.4%.

Poland.

In the case of Poland, it can be observed that underscheduling characterized the DA and ID forecasts made for the onshore wind farms. It can be seen in Figure A24 that a significant part of the dots are below the diagonal in the scatter plots of the joint distributions. The proportions of the downward and upward regulation needs per year for energy production are displayed in Table A22:

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 8.4% and 12.7%. Compared to the DA forecasts, the deployment of ID scheduling could achieve an improvement of only a few percentage points of the annual downward regulation need.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation ranged between 0.2% and 4.0%, which could be improved by the deployment of ID scheduling in the majority of the cases.

Portugal.

In the case of Portugal, both under- and overscheduling can be observed in the case of the DA and ID forecasts made for the onshore wind farms. However, it is also to be seen that most of the dots are located near the diagonal, which suggests a relatively high level of forecasting precision (Figure A25). Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A23):

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 4.9% to 11.2%, which could be improved by a few percentage points by the deployment of ID scheduling.
• In the case of DA scheduling, the proportion of the upward regulation need to the annual energy generation varied between 2.7% and 9.2%, which could be improved to some extent by using ID scheduling in the overwhelming majority of the cases.

Romania.

Both under- and overscheduling were characteristic of the DA and ID forecasts made for the onshore wind farms in Romania (Figure A26), which is also verified by the proportions of the annual downward and upward regulation requirements for energy generation (Table A24).

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 6.3% and 23.6%. The deployment of ID scheduling decreased the forecasting accuracy to a slight degree in 2020, while it improved the precision in 2021.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 5.1% and 26.6%, which could be improved by the use of ID scheduling, albeit only to a slight degree.

Sweden.

Similarly to the situation of scheduling in Germany, Spain, and France, Sweden also seems to possess one of the most accurate DA and ID forecasting mechanisms connected to onshore wind farms. It can be seen that, in the cases of both the DA and the ID forecasts, the clusters of dots are located near the diagonal, which indicates accurate forecasting (Figure A27):

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 1.5% to 5.5%, which could be improved by a maximum of 3.5% by the deployment of ID scheduling.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation varied between 2.5% and 3.9%, which could be improved further by the deployment of ID scheduling in most of the cases (Table A25).

Similar to the case of Germany, Spain, and France, there may be similar reasons behind Sweden’s accuracy in forecasting. On the one hand, this country also has a large amount of onshore wind farm capacity (10.0 GW in 2021), spread over a vast geographical area, and thus the wind farms compensate for one another’s inaccuracies to a certain degree. On the other hand, developments related to onshore wind farm power generation forecasting have been going on dynamically in recent years [67,68].

The United Kingdom.

In the case of the United Kingdom, the results indicate that overscheduling was characteristic of the DA and ID forecasts made for the onshore wind farms; a significant part of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A28). The proportions of the annual downward and upward regulation needs for energy production also verify this observation (Table A26):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 0.3% and 9.2%. The use of ID scheduling resulted in a slight decrease of the forecasting accuracy compared to DA scheduling in all the cases.
• Regarding scheduling, the volume of the upward regulation need was more significant than that of the downward regulation requirement, which may have been caused by the fact that those preparing the forecasts could not precisely calculate with the changeable weather conditions. In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation varied between 15.6% and 29.8%, which could be improved further by the deployment of ID scheduling by up to 18.8%.

Offshore wind farms.

Belgium.

Of the examined nations, Belgium was the country with the most detailed data available related to offshore wind farms. The results (Figure A29) show that under- and overscheduling characterized both DA and ID forecasting in the case of offshore wind farms. The proportions of the annual downward and upward regulation needs for energy production also verify this observation (Table A27):

• In the case of DA scheduling, the proportion of the downward regulation need to the annual energy generation varied between 3.1% and 12.5%, which could be improved by some percentage points by using ID scheduling in the overwhelming majority of the cases.
• Compared to the downward regulation need, the annual upward regulation requirement posed a greater challenge in terms of the management of scheduling in certain cases (e.g., 2020, 2021), which might have been caused by the fact that those preparing the forecasts were not in possession of the weather data related to meteorological situations which proved to be more changeable than expected during the scheduling period. In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation ranged from 9.3% to 30.4%, which could be improved by the deployment of ID scheduling.

Germany.

Similarly to Belgium, under- and overscheduling occurred in the case of the DA and ID forecasts made for the offshore wind farms in Germany too. However, it is also to be seen that most of the dots are located near the diagonal, which suggests a high level of forecasting precision (Figure A30). Taking the annual energy generation as a basis, the proportions of the downward and upward regulation developed as follows (Table A28):

• In the course of DA scheduling, the proportion of the downward regulation requirement to the annual energy generation ranged from 6.7% to 33.3%, which could be improved by a few percentage points by the deployment of ID scheduling in the majority of the cases.
• In the case of DA scheduling, the proportion of the upward regulation requirement to the annual energy generation varied between 7.4% and 9.5%, which could be improved further by the deployment of ID scheduling in most of the cases.

Denmark.

In the case of Denmark, it can be observed that both under- and overscheduling characterized the DA and ID forecasts made for the offshore wind farms (Figure A31). The proportions of the annual downward and upward regulation requirements for energy production are shown in Table A29:

• Compared to the upward regulation need, the annual downward regulation requirement posed a greater challenge in terms of the management of scheduling in certain cases (e.g., 2015, 2021), which might have been caused by the fact that those preparing the forecasts were not in possession of the weather data related to meteorological situations which proved to be more changeable than expected during the scheduling period. The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 5.2% and 26.0%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 2.9% and 8.6%.

The Netherlands.

In the case of the Netherlands, it can be seen that underscheduling was characteristic of most of the DA and ID forecasts made for the offshore wind farms; the majority of the dots are below the diagonal in the scatter plots of the joint distributions (Figure A32). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following observations can be made (Table A30):

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 12.9% and 61.7%. It was only in 2018 that some slight improvement could be achieved by the use of ID forecasting compared to DA forecasting. In the subsequent years, the DA and ID schedules were practically identical.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 1.5% and 15.2%. Although the use of ID scheduling decreased the forecasting accuracy to some slight degree compared to DA scheduling in 2018, the DA and ID schedules were practically identical in the subsequent years.

Portugal.

Both under- and overscheduling were characteristic of the DA and ID forecasts made for the offshore wind farms in Portugal (Figure A33), which is also verified by the proportions of the annual downward and upward regulation requirements for energy generation (Table A31).

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 38.7% and 50.4%. The use of ID scheduling improved the forecasting accuracy compared to DA scheduling, in 2021 by even 13.3%.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling lay between 18.8% and 35.4%. The application of ID scheduling made the accuracy worse compared to DA scheduling in 2020, while it resulted in some improvement in 2021.

The United Kingdom.

In the case of the United Kingdom, it can be seen that overscheduling was characteristic of a significant part of the DA and ID forecasts made for the offshore wind farms; the majority of the dots are above the diagonal in the scatter plots of the joint distributions (Figure A34). Based on the proportions of the annual downward and upward regulation requirements for energy generation, the following observations can be made:

• The proportion of the downward regulation need to the annual energy generation in the case of DA scheduling varied between 1.3% and 6.5%. Compared to the DA forecasts, the deployment of ID scheduling could achieve a maximum improvement of 0.8% of the annual downward regulation need.
• The proportion of the upward regulation need to the annual energy generation in the case of DA scheduling ranged between 10.8% and 44.1%, which could be improved even further by 3.9% when ID scheduling was applied, as shown in Table A32.

3.2. The Comparison of the Examined Countries Based on Their Annual Downward and Upward Regulation Requirement Data

The accuracy of the scheduling of the onshore and offshore wind farms in the examined nations did not present a uniform picture. Figure 1, Figure 2, Figure 3 and Figure 4 show the annual downward and upward regulation requirements of the studied countries. The color coding in Figure 1, Figure 2, Figure 3 and Figure 4 represents the following:

• The darker the green cell of a given figure is, the more accurate the forecast is. A deviation of 0% is marked dark green.
• The closer the color of a cell of a given figure to red is, the more inaccurate the forecast is. A deviation of 150% is marked red.
• The farther the color of a value from green is and the more it fades into yellow, orange, and then red, the greater the magnitude of the downward or upward regulation need was in the case of the DA and ID energy production forecasts.

Based on the results, it was established that, during the examined period, Germany, Spain, France, and Sweden were able to produce more accurate forecasts in terms of the annual DA downward and upward regulation requirements than the other nations, which may be attributable to their geographical location, size, installed onshore wind capacities and their locations spread over vast areas, as well as the fact that the development of energy generation forecasting of onshore wind farms in them looked back on a significant past. Although Denmark, Finland, Poland, and Portugal have smaller territories than Germany, Spain, France, and Sweden and their schedules were less precise, they still produced relatively accurate forecasts. Less accurate scheduling was carried out by Bosnia and Herzegovina, Bulgaria, Switzerland, Estonia, Montenegro, and the Republic of North Macedonia. This lower level of precision may have been caused by the deployment of less time-tested forecasting algorithms, operational errors at the system level, scheduling activities preferring either over or underscheduling, or the occurrence of weather conditions more difficult to predict.

There were six countries whose downward and upward regulation requirement characteristics related to their offshore wind farms were analyzed. Compared to the other countries, Germany and Denmark prepared more accurate forecasts. Concerning Belgium, the United Kingdom, the Netherlands, and Portugal, there were years when there were higher levels of inaccuracy in the forecasts. This lower level of accuracy may have been the result of the application of less time-tested forecasting algorithms, operational errors at the system level, or the occurrence of weather conditions more difficult to forecast.

Regarding onshore wind farms, the analysis of the ID downward and upward regulation requirements was possible in the case of 19 countries, while concerning offshore wind farms only in the case of 5. The application of ID scheduling had a favorable effect on the scheduling accuracy in most of the countries, i.e., it could enhance the precision compared to DA scheduling in the great majority of the cases.

Figure 1. The annual DA downward and upward regulation needs in the countries whose data allowed the comparative analysis of their onshore wind farm scheduling.

Figure 1. The annual DA downward and upward regulation needs in the countries whose data allowed the comparative analysis of their onshore wind farm scheduling.

Figure 2. The annual ID downward and upward regulation needs in the countries whose data allowed the comparative analysis of their onshore wind farm scheduling.

Figure 2. The annual ID downward and upward regulation needs in the countries whose data allowed the comparative analysis of their onshore wind farm scheduling.

Figure 3. The annual DA downward and upward regulation needs in the countries whose data allowed the comparative analysis of their offshore wind farm scheduling.

Figure 3. The annual DA downward and upward regulation needs in the countries whose data allowed the comparative analysis of their offshore wind farm scheduling.

Figure 4. The annual ID downward and upward regulation needs in the countries whose data allowed the comparative analysis of their offshore wind farm scheduling.

Figure 4. The annual ID downward and upward regulation needs in the countries whose data allowed the comparative analysis of their offshore wind farm scheduling.

4. Discussion

The high variability and limited predictability inherent in wind power pose considerable challenges to planners and operators alike, regardless of the time scale. It is obvious that this uncertainty can be reduced by forecasting.

Nowadays, numerous research studies are investigating energy generation forecasting algorithms, systems, or services for onshore or offshore wind farms, but they do not reveal anything (or much) about the accuracy of the scheduling practices prevailing in the particular countries. Thus, this study fills a gap in the current knowledge in this field. The innovative, practical significance of the research presented in this paper is that it determines those day-ahead and intraday onshore and offshore wind farm energy generation forecasting accuracy characteristics of the ENTSO-E countries that are relevant from a practical perspective to the TSOs, the main actors of the energy market as well as the decision-makers, in an easily comprehensible form. The research has proven that it is not possible to prepare 100% accurate schedules for energy generation by onshore and offshore wind farms because of various natural effects, which leads to unpredictable expenses in the electric energy system. The analyses of the day-ahead and the intraday scheduling practices of the onshore and offshore wind farms in the ENTSO-E member states allow the exploration of the limitations of the precision of the forecasting methods of all the ENTSO-E countries suitable for analysis. This knowledge can provide important information for investments in terms of the economic features of wind farms, as well as they may also contribute to the development of the management systems of energy storage solutions, keeping the requirements of the market in mind.

5. Conclusions

In order to be able to operate reliably and efficiently, power systems need precise wind power production forecasting. The accuracy of the scheduling of the onshore and offshore wind farms in the examined nations showed a rather varied picture.

Compared to the other countries, Germany, Spain, France, and Sweden produced more accurate forecasts related to their onshore wind farms during the examined period. The proportions of the annual downward regulation requirements to the annual energy generation in the case of DA and ID forecasting in these countries varied between 0.8% and 14.4%. Furthermore, the proportions of their yearly upward regulation requirements for energy generation, in the case of the DA and ID forecasts, ranged between 0.8% and 6.5%.

Contrary to that, there were also less precise scheduling practices in connection with other countries. It even occurred that the proportion of the annual downward regulation requirement to the annual energy generation in the case of the DA forecasting reached 49.1% (Republic of North Macedonia). In the case of ID scheduling, the highest value was 24.7% (Montenegro). The highest proportion of the annual upward regulation need for energy generation in the case of the DA forecast reached 132.4% (the Republic of North Macedonia). When deploying ID scheduling, the highest value was 44.9% (Bosnia and Herzegovina).

The annual upward and downward regulatory needs of the offshore wind farms, expressed as percentages of the yearly energy generation, varied between 0.9% and 61.7% for the day ahead, and from 1.3% to 44.1% for the intraday forecasts. In this category, Germany and Denmark produced the most accurate schedules.

The objective of future investigations is to determine the information presented in this paper in connection with solar energy too.