1. Introduction
The European power (electricity) system is currently subject to profound changes and so it will be during the next few years. On 30th November 2016, the European Commission published the new legislative package “Clean Energy For All Europeans”, designed with the aim of further completing the single internal market for electricity and implementing the Energy Union [
1]. One of the key elements of this legislative proposal is the deepening of integration of the European power markets. It is important to develop new operational policies supporting power system evolution towards a more intermittent, non-synchronous generation fleet, while considering challenges to system stability and uncertainty from wind and solar generation. Thus, there is the need to study operational flexibility for managing large quantities of variable renewables when performing future electrical system analysis. Flexibility in this paper refers to the ability of a power system to ensure power balance (i.e., power generation equals demand at any time). This is challenging when deployment of generators from non-dispatchable Renewable Energy Sources (RES) is high.
When analyzing high RES scenarios, it is important to assess the flexibility requirements and availability from the generation fleet to design specific measures to cope with the uncovered needs. Some new elements are considered essential in the evolution of the system: a greater responsiveness of consumers (i.e., demand response), the deployment of energy storage facilities, and an efficient use of existing and planned cross-border power exchange capacities.
There is extensive work on power system flexibility. Recent reviews can be found in [
2,
3,
4]. Some work focuses on qualitative analysis or theoretical developments [
4]. Setting up a quantitative flexibility evaluation method is essential to identify flexibility needs and sources; however, there is no consensus on which metrics are best suited [
5], and how the indicators relate with RES curtailment or loss of load. Moreover, there is the need to find a compromise between computational complexity and meaningfulness of the indicators [
6].
From the methodological point of view, there are four sequential steps to assess flexibility according to [
7]:
Flexibility requirements;
Availability of flexibility resources;
Power system adequacy;
Power transmission grid adequacy.
Each step covers different aspects of the flexibility assessment, completes collected information from the previous steps and is characterized by an increasing level of complexity and data requirements
Under Step 1—flexibility requirements—the dynamics of the net load, magnitude of ramps, frequency of occurrence, and timing are perceived over a variety of time horizons. Net or residual load is defined as the difference between electricity demand and the amount supplied by non-dispatchable renewables. This is a deterministic assessment aiming at evaluating the flexibility needs of the electricity system under the assumption that no renewable curtailment is permitted.
Under Step 2—available flexible resources—the existing flexibility resources in the power system are identified. The goal of this level is to understand the potential flexible capabilities in a given system. Like Step 1, this is a deterministic analysis. The technical specificities of the flexibility resources are not examined in full detail, neither the impact of other factors such as market and operational arrangements.
The system flexibility analysis, step 3, includes the assessment of the power system adequacy once flexibility resources have been dispatched, in comparison with the flexibility required by the system at different times. Here, a chronological power system modeling for one year with a 1-h time step is applied. This analysis can have various degrees of complexity and detail: from a deterministic analysis in which only the specific technical constraints of the generating units are considered, to a probabilistic one where RES and load variability are taken into account based on historical data and Monte-Carlo simulations. Moreover, the full spectrum of flexibility resources such as demand response or RES curtailment, can be systematically included, as well as market and operational arrangements (e.g., reserve capacity constraints of the system).
Finally, step 4 is built on step 3 analysis to include the constraints of the power transmission network such as thermal limits of the transmission lines when distributing power flows. The objective is to determine the extent to which the transmission network impacts the ability of the system operator to balance the system using the available flexibility resources. This final step is out of the scope of the present work given that would require detailed modelling of the transmission network in the examined area (Europe).
Summarizing, in step 1, a mapping of the future needs in power system flexibility is performed. In step 2, a systematic recording of the flexibility resources is conducted. The combination of these two steps can provide useful insights to both power system operators and policy makers. However, the ability of the system to cope with the flexibility challenges is investigated in steps 3 and 4. Thus, the value of flexibility resources can be fully analyzed, while assessing the impact of policy, market, technical and operational arrangements, such as reserve dimensioning, planned and unplanned outages, transmission congestions and active participation of demand response and RES into markets.
The main element to consider in flexibility analysis is the more frequent and extensive need of ramping capabilities. There is already extensive work performed on ramp analysis: two regional transmission system operators in the United States, California Independent System Operator (CAISO) and Midcontinent Independent System Operator (MISO), have proposed market-based flexible ramping products to avoid balance violations caused mainly by the variability and uncertainty in generation from RES [
8]. The first step needed when designing new market products is the assessment of the requirements of the system. Ramp events should be characterized in terms of ramping start and end time, ramping duration, ramping rate, and ramping magnitude [
9]. Not only deeper ramp events will be a constant new feature in the future system, but also inertia will be challenged. The phase-out of coal and nuclear power plants, at least in some European countries, and the significant increase of wind and solar generation will transform the system towards an inverter-dominated grid [
10].
The objective of this work is to propose a methodological approach to analyze power system flexibility in a systematic way. The focus is in the evolution of the European power system from 2020 to 2025, where the new legislative proposal will enter into force. There are several previous works on specific countries, as Germany [
11], France [
5] and Greece [
12], but less on assessing the European system as a whole [
13]. Although there are several previous works focusing on the analysis of 1-h ramps, we consider essential to extend this analysis up to 3 or 4 hour ramps, at least in countries with significant solar photovoltaic (PV) installed capacity, as these longer-lasting ramps impact the operation of the grid and the sizing of the operational reserve needs. A more complete analysis considering the evolution of system inertia is also included.
The insights gained applying the proposed methodology can then be used as input to guide the development of new flexibility resources and to design new market products.
The following in this article is organized as follows:
Section 2 proposes the metrics for the first three power system flexibility assessment steps conducted in this work;
Section 3 presents the European power dispatch model.
Section 4 shows the main results of the flexibility analysis. Finally,
Section 5 highlights the main conclusions.
3. Model and Data
Step 3—power system adequacy—was assessed with a Europe-wide power system model. The model is developed for 2020 and 2025 scenarios following the European Network of Transmission System Operators (ENTSO-E) Mid-term Adequacy Forecast (MAF) 2016 [
21]. The model [
22] comprises of (i) 32 European countries (Austria, Belgium, Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Denmark, Estonia, Finland, France, Former Yugoslav Republic of Macedonia, Germany, United Kingdom (which is modelled as two different regions: Great Britain and Northern Ireland), Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Montenegro, Netherlands, Norway, Poland, Portugal, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, and Switzerland, modelled as 33 nodes), and (ii) the cross-border transmission connections between these nodes. The modelled nodes and their cross-border connections are shown in
Figure 1.
The model was built on the PLEXOS
® power market simulation software [
23]. A time series modelling approach is applied: each scenario is a one-year period with an hourly time step, using deterministic programming techniques that aim to minimize an objective function (the total cost), subject to some constraints representing the physical characteristics of the system: power balance, power plant availability, power reserves, power transmission constraints, and fuel/emissions prices, up and down ramping capabilities of the generation fleet, etc. Thorough description of the model is available in a previous study [
14], which includes objective function of the optimal dispatch of power generators, power system constraints, and details on input data.
Generation and cross-border capacities, hourly demand profiles, and CO
2 prices for 2020 and 2025 modelling scenarios were set following the modelling dataset of the MAF 2016 published by ENTSO-E [
21].
Wind and solar hourly generation profiles are acquired from an open database—
http://renewables.ninja/ version 1.0 [
24,
25] based on historical records covering 25 years—1990–2014. Every future scenario year (e.g., 2020/25) was modelled under 25 different historical weather conditions. Monthly water inflow profiles for the hydro power plants was extracted from the historical generation records of Eurostat [
26].
The main properties and constraints of the modelled generation technologies followed the Joint Research Centre report on projected energy technology indicators [
27]. Outage rates for power generators were obtained from the World Energy Council [
28]. The unplanned outages were distributed randomly through a year; but the planned outages were modelled to occur mostly during times of low electricity demand. Outage pattern was different from one historical year to another.
5. Conclusions
This article presents a systematic approach for flexibility assessment in a power system based on a set of indicators. This allows comparison between countries and consideration of different scenarios. The first two steps of the approach, (i) flexibility requirements and (ii) existing flexibility resources, can be assessed without power system modelling.
The proposed flexibility analysis approach has been applied to a European power system to assess its adequacy in 2020 and 2025. We present an overview of different flexibility requirements and flexibility resources for 32 European countries (modelled as 33 nodes) and how their flexibility requirements and capabilities can complement each other.
Three main ramp characteristics have been assessed for wind and residual load: most frequent, maximum, and maximum average rate ramps. It was shown that the most significant wind ramps occur within one to two hours, although the same does not hold for the residual load ramps. Up and down ramps of residual load present distinct behaviors, reflecting different timings of up and down PV ramps, load variations, and wind energy cycles. It has been shown that ramps up to three hours must be analyzed, and that up and down ramps are not identical, giving rise to differentiated, specific ramp-up and ramp-down needs. This is important if considering the introduction of new power balancing products in a market, and for dimensioning the operating reserve needs.
Inertia will be also a challenge in the future. It is shown that its evolution depends largely on synchronous area-specific resources, although the probability distribution function of the non-synchronous penetration ratio shows a long right-hand tail for all the simulated nodes. This means that extreme low inertia values are expected at some particular periods of the year, so new requirements for inverter-based generators and/or new specific products in the market are foreseen to deal with these low inertia periods of time.
Different power balancing resources contributing in power system flexibility (i.e., interconnection capacity, hydro power plants, demand response, etc.) are present in the system to cover the flexibility requirements in the analyzed countries. We have shown the different available resources in the system, highlighting that it is important to have complementary means in neighboring countries to address future needs. Efficient utilizations of these resources depend on power market developments. This is assessed in Step 3 of the analysis approach (generation adequacy assessment) through a pan-European dispatch model. The potential risk of renewable energy generation curtailment disappears for all modelled nodes thanks to the export capacity and the pumping capability, except for Ireland and Northern Ireland, where it is significantly reduced. The minimum reduction varies from 91% in 2020 to 73% in 2025 in the case of Ireland, and from 87% in 2020 and 60% in 2025 in the case of Northern Ireland.
For 2025, different adequacy indicators are presented for those nodes with positive values (Bulgaria, Finland, Greece, Ireland, and Poland). The indicators, loss of load duration, loss of load occurrence and energy not served, complement each other.
Future work will be to extend the step 3 assessment, developing a methodology for differentiating between ENS due to inadequate capacity, and ENS due to inadequate flexibility. Regarding inertia, a more detailed analysis shall be performed to estimate the impact of low system inertia periods on the expected Rate of Change of Frequency (RoCoF) under a major disturbance in the system.