Abstract VALUE OF FLEXIBILITY FOR BALANCING ...

VALUE OF FLEXIBILITY FOR BALANCING WIND POWER GENERATION Özge Özdemir*, Paul Koutstaal, Marit van Hout Policy Studies Unit, Energy Research Centre of the Netherlands, Amsterdam

Abstract In this study, we address how increasing shares of intermittent renewable power production will affect the need for flexibility within the Dutch electricity system. We use a unit commitment model to simulate European power market consisting of 33 countries; taking into account hourly demand and intermittent generation and cross-border transmission. Flexibility capabilities of resources are implemented in the model via constraints for ramping, minimum up and down times, and the lumpiness in generator start-up decisions, a feature not considered in most continent-wide electricity market models. Our analysis show that higher share of wind power generation does not only increase the need for flexibility to accommodate wind variability on the day-ahead market but also it leads to an increased demand for flexibility on intraday and balancing markets to accommodate wind forecast errors. Price volatility is also increased. This results in additional revenues and positive business case for flexible generation and storage providing flexibility for real time balancing. A wellfunctioning market, in which all producers including renewable energy power generators have program responsibility, will be crucial to provide the price signals needed to ensure that the flexible assets are available to accommodate variable production from renewables.

1. INTRODUCTION The future European electricity markets will include higher shares of intermittent renewables generation (I-RES), which is less predictable by nature, resulting in increasing deviations between predicted and realized electricity production. This will increase the need for balancing wind power generation in real time. In addition, a considerable amount of back up capacity is needed for those periods in which renewables production is low. System operators are already calling for flexibility from conventional generation to cope with the uncertainty and variability of increasing wind and solar production. Simultaneously, low operating costs and subsidies for renewables are decreasing electricity prices, reducing the income for conventional power production. There is an ongoing debate on the need of capacity mechanisms in Europe to secure sufficient amount of generation capacity with increasing shares of I-RES in future (ACER (2013)). Increasing shares of I-RES therefore poses challenges for the performance of the current electricity market and for the business model of conventional generation. However, it may also raise new opportunities in competitive markets where all participants are responsible for balancing their programs. This includes, for example, increased demand for ramping up and down to accommodate the variability and forecast errors of intermittent renewable generation. Balancing responsible parties (BRP) will look for flexibility to balance their programs, given differences in their realized wind production compared to their submitted programs. The need to accommodate forecast errors with steeper ramps could create new revenue streams for flexible resources in the form of flexibility markets or bilateral contracts with BRPs. In order to investigate the potential of these new market opportunities and business cases of different flexibility options, it is important to quantify the demand and value of flexibility for balancing wind generation and how markets for flexibility will develop in the future with significant increases in wind generation and forecast errors. In this study, we address how increasing shares of intermittent renewable power production will affect the need for flexibility within the Dutch electricity system and how markets for flexibility will develop in the Netherlands. To this aim, we will analyse the expected future developments on the electricity market in the Netherlands and the rest of Europe for the years 2017 and 2023. According to the Energy Agreement concluded in 2013 in the Netherlands (PBL & ECN (2013)), renewable energy production in the Netherlands will have to reach a share of 14% in 2023, with a capacity of 6000 MW of

*

 

Address: Radarweg 60 1043 NT Amsterdam, Phone: +31-885-15 4001, Email: [email protected]

2

wind on shore and 4450 MW of wind off shore. Assuming that these shares will be realised, the year 2023 will provide a good example of an electricity market with a substantial share of intermittent renewable power production. In the analysis of flexibility, we will make a distinction between variability from wind energy based on the expectations of wind power production on the day-ahead market and the increasing demand for balancing in the intraday and balancing markets because of the forecast error of wind power generation. On the day-ahead market, flexible generation will supply increased ramping up and down in order to accommodate increased variability of renewable production. Furthermore, peak generation will be needed to meet high demand levels at moments of low wind power production. Because of the forecast error, realized wind power production will differ from the forecasted production on the day-ahead. Balancing responsible parties will therefore look for flexibility in the intraday market to balance their programs, given differences in wind production compared to their submitted programs. As far as these differences cannot be met on the intraday market, these will have to be addressed by the TSO, e.g., TenneT, who will contract regulating and reserve power in order to ensure system stability. In our approach, we will not distinguish between the intraday market and the singlebuyer market for regulating and reserve power; instead we will consider the need for flexibility because of forecast errors as one market, and refer to it as “intraday market”. Many studies have addressed the impacts of wind power generation on the performance of system operations. A common approach suggested first by Garver (1962) is utilizing unit commitment (UC) models that account for the ramping capabilities of different technologies and the lumpiness in generators’ start-up decisions. Unit commitment (UC) models are formulated as Mixed Integer Programs (MIP) that are complex by nature and considered to be difficult to solve for real world large scale systems. Hence, most applications are limited to one country or region. To overcome this, we use a practical approach to simulate European power market consisting of 33 countries; taking into account hourly demand and intermittent generation and cross-border transmission. While we formulate the full UC problem for the units in the Netherlands, we use a Tighter Relaxed UC (TRUC) formulation of Kasina et. al (2013) for the other countries. TRUC is a continuous relaxation of UC problem for solving large scale systems within a reasonable time while capturing the most of the characteristics of a unit commitment problem; e.g., prices and price volatility. One of the consequences of the growth in renewable electricity generation is more volatility in the production as a results of large changes in production from one hour to the other. This volatility will have to be accommodated by dispatchable conventional generation or by other flexibility options. Not all conventional power plants will be able to accommodate those changes by ramping up or down their production fast enough because of flexibility constraints. Our analysis show that increased flexibility requirements will increase demand for flexibility to accommodate wind variability and shift production towards more flexible gas fired power plants. Higher share of wind power generation does not only increase the need for flexibility to accommodate wind variability, it also leads to an increased demand for flexibility to accommodate wind forecast errors on the intraday market. This results in significant increase in revenues and positive business case for flexible generation and storage providing flexibility for real time balancing. In addition to the increase in the demand for flexibility, there is also an increase in price volatility on the day-ahead and intraday markets, indicating better business cases for investments in storage and demand side management. The supply of flexibility from both incumbent and new suppliers will not be forthcoming without adequate incentives. A well-functioning market will be crucial to provide the price signals needed to ensure that the flexible assets are available to accommodate variable production from renewables and wind forecast errors. The paper is organized as follows. In Section 2, we give an overview of the methodology and assumptions. Section 3 contains general electricity market results for the Netherlands such as generation-mix, spot prices, and electricity exchange with neighbouring countries. Section 4 illustrates increased demand for flexibility in the Netherlands for both day-ahead and intraday markets and in Section 5 we analyse profitability of a new conventional generation and storage unit in providing flexibility on the intraday market. Conclusions and discussions for future research are presented in Section 6.

3

2. METH HODOLOGY Y AND ASS SUMPTION NS mitment Mo odel Unit Comm For our analysis, we extendedd the Europeaan electricity market modeel COMPETE ES1 with a unnit commitmeent European pow wer market. It covers 26 EU U member stattes formulation. COMPETES is a transmisssion-constrainned model of E U countries (i.e., Norway, S Switzerland, annd the Balkan countries) inccluding a repreesentation of the t cross-bordder and 7 non-EU transmission limitations innterconnectingg these Europpean countriees. Every couuntry is repreesented by onne node, exceept Luxembourg is aggregatedd to Germany,, Balkan and Baltic countriies are aggreggated in one nnode, and Dennmark is split in two nodes duue to its particcipation in twoo non-synchroonous networkks (See Figuree 1). The moddel assumes ann integrated E EU market wheree the trade flow ws between coountries are coonstrained by “Net “ Transfer Capacities (N NTC)”. COMPETES involvess a wide-rangge of generatioon technologiies. The geneeration type, ccapacity, and the location of existing geneeration technollogies are reguularly updatedd based on WE EPPS databasee (UDI (2012))). COMPETE ES database is in particular dettailed out withh unit by unitt generation iin the Netherllands. For thee other countriies, the units using the sam me technology annd having sim milar characteriistics (i.e., agee, efficiency, technical consttraints) are agggregated. Figuree 1 Geographiccal coverage a and network rrepresentatio on of COMPETTES UC in 2013 3 according to o ENTSO‐E  TTen‐Year Netw work Developm ment Plan (20 012)  

  The Unit Commitmeent problem (UCP) is thhe problem oof deciding w which powerr generating units must bbe committed/deecommitted ovver a planningg horizon. Ourr unit commitm ment model finnds such a sollution that minnimizes the tottal variable, minnimum-load, aand start-up coosts of generattion and the costs c of load-sshedding withhin a year in aall the countriees. The committted units mustt satisfy the eelectricity dem mand at each hour, h as well as a large sett of technologgical constrainnts summarized as: a  Pow wer balance coonstraints ensuure demand annd supply is baalanced at eachh node at any time.  Generation capaccity constraintts limit the maaximum availaable capacity of o each generaating unit. Theese also include deraating factors too mainly captuure the effect oof planned andd forced outages to the utilizzation of the plants.  Crosss-border trannsmission consstraints limit tthe power flow ws between thee countries forr given NTC vvalues.  Ram mping up and down constraaints limit the maximum inncrease/decreasse in generation of each unnit between tw wo conssecutive hourss.  Miniimum load connstraints set thhe minimum ggeneration leveel of each unitt when it is coommitted. 1

 COMPETES UC is an extennsion of COMPE ETES Model (H Hobbs and Rijkeers (2004)) whicch has been devveloped in cooperation with Hobbs, Professorr in the Whitingg School of Enggineering of Thee Johns Hopkinss University, as a scientific advvisor of ECN. Benjamin F. H COMPETES model m has beenn used in numeroous studies ((O Ozdemir et al. 20009, 2013), Hobbbs et.al. (2008)), Wietse et al. (2010)). (

4



Minimum up and down time constraints (only for the Netherlands) set the minimum number of hours that each unit is on/off after start-up/shut-down. The incorporation of start-up and min-load costs and the constraints for minimum generation and ramping allows for the analysis of flexibility requirements and the resulting price volatility to accommodate the variability and forecast errors of wind. For the Netherlands, every unit is modelled with minimum generation levels and the corresponding start-up and minimum load costs. For other EU countries where generators of similar characteristics are aggregated, the start-up and minimum load costs are approximated by relaxing integer unit commitment decisions (Kasina et al. (2013)). In addition, the model also includes the operation of storage in the Netherlands, maximizing its revenues by charging and discharging electrical energy within a day. The model is utilized in two steps. First, the unit commitment model for the entire EU market is used to simulate the day-ahead market in Europe with predictions of intermittent generation at an hourly resolution for each day of the year. This results in a schedule of imports between countries and the commitment of slow generating units. Second, by fixing the dayahead import/export schedules to/from the Netherlands, we in particular simulate the system balancing within the Netherlands to balance actual wind power generation by dispatching the Netherlands generation against realised net load, subject to the fixed slow generator commitment and net imports from day-ahead market. The generation and price differences between the day-ahead and the intraday markets are used to calculate the demand and value of flexibility for generation technologies and storage to accommodate wind forecast errors.

Scenario Assumptions up to 2023 Future electricity market developments are driven by a number of different factors. The most important factors are (a) electricity demand, (b) generation capacity mix, (c) fuel and CO2 prices, and (d) interconnection capacity. The unit commitment model described in the previous section yields a static market equilibrium. The long-term planning decisions in the form of adequate generation capacity and cross-border import capacity are part of the scenario and thus exogenous to the model. Scenario A of ENTSO-E SO&AF (2013) is chosen to represent the generation and demand developments up to 2020 for all EU countries except for Germany and the Netherlands. Scenario A forecasts low growth of net generation capacity until 2020, since only confirmed generation projects are considered. After 2020, Vision 3 is assumed as the “Green transition” scenario. The trend of installed generation mix in these scenarios shows a decreasing importance of fossil fuels over RES reaching 41% RES share in 2020 and 54 % RES share in 2030. Since only data for 2013, 2015, 2016 and 2020 are provided in SO&AF scenarios, data for 2017 is linearly interpolated between 2016 and 2020. Since the end year of the Energy Agreement (2023) is of interest for the flexibility analysis, data of 2023 for these EU countries are derived by linearly interpolating data between 2020 (Scenario A) and 2030 (Vision 3). The expected demand in 2017 and 2023 in the Netherlands is based on the projections of the Dutch Energy Agreement (PBL & ECN (2013)). Further, the generation capacity mix is similar to the Energy Agreement. Decentralized CHP generation is projected to decrease to about 23-24 TWh in the period 2017-2023, roughly 30% less than 2010 levels.2 Renewable energy production in the Netherlands is expected to reach a share of 14% in 2023, with a capacity of 6000 MW of wind on shore and 4450 MW of wind off shore. Figure 2 presents an overview of Dutch generation capacity over the years. For Germany, the existing capacity has been derived from recent BNetzA figures (BNetzA, 2014). The same source has been used to estimate the amount of planned decommissioning in Germany (nuclear and other thermal power plants) and conventional new build for the coming years. The increase in renewable power has been based on a combination of assumptions in Energy Agreement and SO&AF Scenario A. Germany’s I-RES capacity (Wind+Sun) growth is 5.5 GW/year (7.6% /year) of which 2GW/year is the growth of wind power capacity and 3.5GW/year is the growth of solar-PV capacity. Hourly intermittent renewables production profiles are based on actual data of 2012, obtained from both TSOs and from data available at ECN. The flexibility assumptions for conventional units are assumed to differ with the type and the age of the technology as summarized in Table 10 in the Appendix

2

 Decentralized CHP in the Netherlands is not included in COMPETES and its generation is projected by a bottom-up model of ECN (Daniels and van Drill (2007)).

5

Figure 2 Installed generation capacity Netherlands 45 40 35

GW

30 25 20 15 10 5 0 2012 Nuclear & hydro

2017 Coal

non‐intermittent RES

2023 CHP

Gas

Wind & sun

Prices of fossil fuel input (i.e., coal and gas) and carbon emissions for electricity generation are important determinants of electricity prices. Choosing fossil fuel price scenarios however is difficult, given the many factors that influence fuel markets and the large uncertainties with regard to future developments. For the first couple of years, future markets give some indications about future prices, reflecting current market expectations. Therefore, we use recent forward prices of coal and gas for 2017. For 2023, we use prices based on the World Energy Outlook’s current policies scenario. Table 1 gives the prices used in 2017 and 2023. Table 1 Fuel and CO2 prices Coal Price Natural gas price CO2 price

[€ct/GJ] [€ct/m] [€/ton]

2017 3.3 25.8 9.0

2023 3.5 30.9 13.5

Interconnection capacity developments in Europe are based on 2012 Ten-Year Network Development Plans of ENTSO-E, with some adjustments made to the interconnection capacity of the Netherlands in consultation with TenneT. For the Netherlands, interconnection capacity increases from 4550 MW in 2012 to 8050 in 2017 and 8750 in 2023.

3. GENERAL RESULTS ON WHOLESALE ELECTRICITY MARKET Based on the assumptions described above, generation mix, import/export flows, and wholesale electricity prices for the Netherlands are calculated. Given the increase in intermittent renewables capacity, electricity production from renewables, especially wind, increases significantly over the years in North-west Europe (see Figure 3). Figure 4 shows the production and gross import/export in the Netherlands for 2012, 2017 and 2023. The Netherlands is a net importer in 2012 and 2017 (17 TWh in 2012 and 20 TWh in 2017). It becomes a net exporter in 2023 (26 TWh) with significantly increased exports to UK and Belgium. While net imports from Germany increase in 2017 (15 TWh), exports to Germany increase in 2023 during low wind high demand hours resulting in a significant decrease in net imports from Germany in 2023. The increase in exports from the Netherlands in 2023 is a combined effect of providing flexibility to the countries with high intermittent generation and a relatively lower net generation capacity margin in the North-west European countries; especially in UK, France, and Belgium. Increased flexibility requirements due to more intermittent generation in North-west Europe shifts production towards more flexible gas fired power plants in the Netherlands. Average wholesale prices decline in 2017 compared to 2012 and they are significantly higher in 2023 (see Figure 5). The higher prices in 2023 as compared to 2012 and 2017 can be explained by higher fossil fuel prices and a relatively lower net generation capacity margin in the North-west European countries (e.g., UK, FR, BE). In 2023, there is a limited level of demand curtailment in the Netherlands (about 10 hours, for a total of 13.4 GWh). In these hours, prices reach the level of the given VOLL3. With the rising electricity production from renewable energy sources, large variations in 3 We assumed a low value of VOLL which is related to the marginal costs of an expensive peak unit such as a gas turbine, 200 €/MWh in 2012 and 2017 and 320 €/MWh in 2023. The average price will depend on these assumptions, therefore with a higher VOLL, the average price and price volatility would be higher. 

6

generation from renewablees will be accompanied by increased pricce volatility. This T becomess apparent especially in 20223. Price volatilitty will be an iimportant drivver for flexibility options suuch as storagee and demand side responsee. These optioons are able to prrofit from the pprice differencces caused by the variabilityy of wind pow wer production.. Figure 3 Inteermittent Ren newable Geneeration in North-West EU U until 2023

Figure 4 Geeneration mixx Netherlands

Figure 5 Volatility of an nnual averagee prices in thee Netherlandss

€ 120 € 110 € 100 € 90 € 80 € 70 € 60 € 50 € 40 € 30 € 20 2012

2017

2023

7

4. DEMAND FOR FLEXIBILITY Flexibility on Day-Ahead Market Table 2 shows the total domestic demand for flexibility in the Netherlands on the day-ahead market.4 While the total domestic demand for flexibility in 2012 is fully supplied by domestic generation in the Netherlands, interconnection capacity is becoming more important in future years since change in imports/exports are contributing to the supply of flexibility. Developments that are likely to increase the importance of interconnection capacity for flexibility are the increasing scarcity of generation capacity, stronger links between the Netherlands and surrounding countries, and the increasing variability due to wind, not only in the Netherlands, but in the whole of Europe. The potential for new entrants is calculated based on the level of curtailment (i.e. unmet demand). In 2012 and 2017 there is no curtailment of demand, and the total demand for flexibility only slightly increases (+6%). In 2023, there is an increase in total demand for flexibility of 46% compared to 2012. The increase in demand for flexibility has strongly been affected by the increase in wind power production. The share of demand for flexibility due to the variability in wind power generation is limited to only 1% in 2012 and increases to 20% in 2023.5 Table 2 Total demand for flexibility in the Netherlands, Day-Ahead Market Total demand for flexibility (GWh)

2012

2017

2023 3397

Total demand:

2332

2480

of which potential for new entrants

0

0

4

Demand for flexibility ramp up:

2332

2480

3393 2521

supplied by generation

2339

2313

supplied by net imports/exports

-6

167

872

Demand for flexibility ramp down:

-2332

-2480

-3395

supplied by generation

-2338

-2313

-2521

supplied by net imports/exports

7

-167

-874

Figure 6 shows the supply of domestic flexibility per technology. From this figure it becomes clear that especially Gas CCGT units are becoming important when providing domestic flexibility in the future in the day-ahead market. Furthermore, even though around 2.7 GW of gas GT capacity is decommissioned between 2012 and 2023, the remaining units in 2023 are utilized more to provide flexibility. On the contrary, base load units such as coal and biomass standalone that provide a certain share in the supply of flexibility in 2012 are contributing much less to the demand for flexibility in especially 2023. An important reason for the diminution of the role of coal-fired power plants in providing flexibility is that they are used mainly to provide base load in 2023. Hence these units are less available to provide flexibility. Another reason is that the older coal-fired power plants are not flexible enough. In order to better understand the role of different types of power plants in the Netherlands to accommodate the intermittent renewables in North-West Europe, we have compared the model outcomes when unit commitment and flexibility constraints are implemented and when these constraints are not implemented, i.e., a simple economic dispatch. Table 3 shows the differences per technology type, both absolute and as a percentage of the production levels compared to the case without flexibility constraints. The increase in generation from flexible units such as CCGT and GT in the Netherlands is larger than the reduction in less flexible generation. This is due to the fact that flexible generation in the Netherlands will also provide flexibility to neighbouring countries, where generation from coal fired power plants is also decreased. The turnover from this increased production is 101 M€ for CCGT plants and € 3.4 M€ for gas turbines in the Netherlands.

4

 Demand for flexibility in day-ahead is defined as the sum of the change in generation from one hour to the next over a year plus the total curtailment of demand in a year. 5  The share of flexibility demand due to wind power has been calculated by comparing a case in which it is assumed that wind power generation is flat on a monthly basis resulting in the same annual wind power production.

8

Figure 6 Sup pply of domeestic flexibilityy per technoloogy, day-aheaad market

Table 3 Imp pact of flexibiility constrain nts Generation technoology Coall Biom mass standalonne CHP P CCG GT GT

Change in prodduction Wh) (TW -0.3 -0.044 0.1 2.5 0.6

Perccentage changge -1% % -1% % 0% 6% 21% %

Incrrease in turnovver (M €€}

101 3.4

Flexibility on Intraday y Market The increase in w wind power geeneration alsoo increases deemand for flexxibility on thee intraday maarket in order to accommodatee wind forecasst errors. Winnd power prodducers or the BRPs B which arre assumed too have responssibility for winnd production neeed to compennsate these foorecast errors. Or, alternativvely, the TSO has to contraact reserve pow wer to meet thhe imbalances. T This generatess additional deemand for flexxibility on the intraday markket. We assum me the same foorecast errors in 2023 as obseerved in 20122, utilizing thee forecasted aand realised hhourly wind prrofiles6 of 2012. This results in 5% meaan absolute erroor normalized by the installeed capacity (N NMAE). This vvalue is withinn the range of 3%-20% NM MAE given in thhe literature for different counntries. Dem mand for flexibbility has beenn determined by b taking the difference bettween two moodel runs, one with forecasted wind producttion and anothher with realiised hourly wiind productionn. It has beenn assumed thaat net imports/exports remaain fixed at the level based onn the day-aheaad schedules w with forecastedd wind powerr generation annd only the geenerators withhin t accommoddate wind foreecast errors. T This gives an estimate of thhe the Netherlannds is allowedd to adjust itss production to increased demand for fleexibility as a result of wiind forecast errors within the Netherlaands and the most efficieent accommodatiion from the inncumbent genneration. Tablle 4 shows thhe demand forr flexibility inn GWh on thee intraday maarket due to w wind forecast errors. Demannd increases signnificantly withh the increasinng level of winnd generation. Figure 7 shoows that the suupply of flexibbility to balannce wind forecast errors is to a large extennt from gas unnits (i.e., CCG GT and gas tuurbines), whicch are the mosst flexible uniits available in oour scenarios. Some flexibillity is supplied by coal fired power plantts, especially bby new units which are moore flexible than the units in pplace in 2012.. Given our aassumption of a national Duutch balancingg scheme wheere cross-bordder capacity do nnot contribute in real time baalancing, not aall demand for flexibility caan be met in 22023 from incuumbent sourcees. If balancing prices p during hours with unnmet demand aare sufficientlly high, this will w provide ann incentive forr additional neew flexibility souurces such as flexible generration, storagee, and demandd response to eenter the markket or a shift of o capacity froom 6 Hourly foreccasted and actuaal realised wind data for the Neetherlands were acquired from tthe Wind energgy unit at ECN.

9

the day-aheadd market to thhe intraday maarket. In Sectioon 5, we invesstigate the proffitability of enntrants providiing flexibility in the intraday m market. T Table 4: Demaand for flexibillity to accomm modate forecasst errors in thee Netherlands GW Wh Dem mand for flexibilitty ramp up

Non-committed incumbents Additional suuppliers

Dem mand for flexibilitty ramp down

20112

2017

5995

1143

-3668

-734

2023 2340 824 -2041

Figurre 7 Supply of upward and d downward fflexibility on the intraday market

Balancing g Prices and d Value of F Flexibility o of Incumbe ent Units on n the Intrad day Market Withh increasing level of renew wables in 20233, there is alsoo an increase in price volaatility on the intraday markeet. Figure 8 shoows the monthhly prices calculated in 2023 for the sppot market dayy-ahead and ffor the intradaay market. It is expected thatt in those hourrs where theree is an upwardd demand for flexibility (geeneration incrrease) on the iintraday markeet, prices will be higher com mpared to the day-ahead maarket, while iin those hourss where theree is a demandd for downwaard flexibility (geeneration reduuction) prices will be loweer. On the intrraday market,, for upward adjustments ccapacity will bbe offered that hhas not been ssold on the daay-ahead markket and thereffore prices willl be higher thhan on the dayy-ahead markeet. For downwarrd adjustmentss, suppliers wiill be preparedd to pay the baalancing party or the TSO, bbecause they w will already havve sold their eneergy on the dayy-ahead markeet (Abassy et aal. (2011)). Figure 8 Monthlyy spot and intrraday prices in 2023

10

Given the volumes and prices on the intraday market, we can determine the value of the flexibility from incumbent generation provided on the intraday market. We have assumed that upward flexibility is recompensed at the actual market price as realised in the specific hour in which the flexibility is supplied. This price is equal to the variable costs of the marginal production unit at that hour. For downward adjustment, the value equals the difference between the day-ahead prices for the hour under consideration minus the price on the intraday market. Net revenue equals price times volume supplied minus the variable production costs in the case of demand for upward flexibility. Based on these assumptions, Table 5 shows the net revenue realised by different types of technologies in the Netherlands. Table 5 Net revenue on intraday market by technology M€ 

2017 

2023 

Coal  Gas CCGT 

2  1 

5  16 

Gas CHP  Gas GT 

6  0 

6  55 

The large increase in wind power generation from 2017 to 2023 raises the net revenues on the intraday market, especially for Gas CCGT and gas turbines. Besides the incumbent generators, additional upward flexibility could be provided by new generators or by an increase in net imports in the hours in which demand for flexibility cannot be met in 2023. Given the assumed low VOLL of €320/MWh, the potential gross revenue for unmet demand is 264 M€.

5. BUSINESS CASES FOR FLEXIBILITY SUPPLY IN 2023 Given demand and prices on the intraday market, we analyse the profitability of a new conventional generation or storage unit in providing the flexibility demanded in this market. In our analysis, we concentrate on the intraday market. It should be realized that in practice, investments and generation decisions will take into account all markets on which these assets can be used, from longer term forward markets and day-ahead markets to intraday markets and the provision of ancillary services. Concentrating on the intraday market allows us to analyse the additional opportunities which demand for flexibility can provide. However, we will take into account that a plant will also produce on the day-ahead market and therefore investment costs do not have to be covered solely on the intraday market. Our analysis in this chapter is concerned with business cases for the investment in a single new plant or facility and investigating their profitability in the intraday market given expected future developments. It is not the purpose of this analysis to derive the optimal capacity mix in the day-ahead or intraday market.

Conventional Generation We have analysed the profitability of a single CCGT and a GT power plant on the intraday market to accommodate wind foreast errors, taking into account the flexibility constraints for these units. Whether an investment will be profitable or not depends on the costs and benefits of the project. Table 6 gives an overview of the main costs and benefits which are incurred by an investor. In evaluating business cases, the costs and benefits over the whole lifetime should be taken into account. As we focus on a specific year, 2023, we use the annuity of fixed and investment costs over the whole lifetime to evaluate the business case for the year under consideration. Table 6 Indication of Costs and Benefits Costs 

Benefits 

Investment costs  Fixed operation & maintenance costs 

Revenues  Residual value investments  

Variable costs 

 

Table 7 provides an overview of our main assumptions with regard to the fixed costs for the generation technologies considered. These assumptions are based on a recent review of the costs of power plants by Brouwer et. al (2014). Brouwer et. al (2014) derive the values of different cost components based on a range of studies and provide data for different future years, which include average cost reductions in investment costs for different technologies. We have

11

used these cost estimates for 2020. The investment costs include capital costs and residual value. Parameters such as lifetimes and discount rates are similar to those used in the 2010 edition of the IEA’s projected costs of generating electricity study (IEA (2010)). Typical capacities of a single plant are based on those constructed recently and on data from the literature. Table 7 Input parameters for conventional power plants  

CCGT 

GT 

Investment costs 

€661/KWe 

€355/KWe 

Lifetime  Capacity 

30 years  435 MWe 

30 years  150 MWe 

Fixed O&M  Efficiency 

€14/KWe  60% 

€9/KWe  38% 

10%  37 M€ 

10%  €7 M€ 

Discount rate  Annual fixed costs 

Based on the input data in Table 7 and on demand and prices on the intraday market, annual profits for a new CCGT and a new GT unit on the intraday market are calculated, taking into account flexibility constraints and marginal production costs for these technologies. Given our focus on the intraday market, we do not include revenues from generation on the day-ahead market. Instead, we allocate investment and fixed costs to both the day-ahead and the intraday market based on the production volumes and use the share of annual fixed costs attributed to the intraday market. Our calculations indicate that gas-fired power plants, in particular combined cycle gas turbines, will be able to make a profit on the intraday market. They not only replace the old and less flexible incumbent generation but also provide additional flexibility at the hours with unmet demand. Since the efficiency of a GT plant is lower than that of a CCGT plant, its fuel costs are substantially higher. While a gas turbine plant is only just profitable (approx. zero profits), profits for a CCGT plant are about €25 million in 2023. The positive profit for a CCGT plant indicates that there is room for additional sources of flexibility in those hours in which CCGTs provide flexibility and the hours with unmet demand. Assuming constant net revenues over the whole lifetime of a plant, the internal rate of return for the CCGT plant is high, 47% while for the GT it is 1%. This difference reflects the operating hours of both plants, where GTs have a much lower load factor.

Storage Another option to provide flexibility is storage. The storage technologies which can provide volumes useful on the intraday market are mainly pumped hydro storage and Compressed Air Electricity Storage (CAES). Given the focus on balancing in the Netherlands, a business case for a 300 MW adiabatic CAES has been analysed. In contrast to diabatic CAES, this variant does not need natural gas to expand compressed air, because heat generated in compressing air is stored and used to provide the heat needed in decompressing. Although the investment costs are higher, energy efficiency is also higher compared to diabatic CAES. Table 8 shows the characteristics and data with regard to this storage option based on DNV KEMA (2013). Table 8 Input parameters for storage   Investment costs  Lifetime  Capacity 

Adiabatic CAES   600 – 1200 €/kWe  30 years  300 MW 

Maximum storage capacity  Discount rate 

2700 MWh  10% 

Efficiency  Annual fixed costs 

70%  19 ‐ 38 M € 

We assume that storage can be used within a day, charging when electricity prices are low and discharging when demand and therefore prices are high. The optimal use and the revenues of storage are calculated through modelling of the CAES unit in the intraday market analysed with the COMPETES model. Table 9 presents the results of this simulation.

12

Depending onn the investment costs, for which currentt estimates proovide a range of 600 - 12000 €/kWe, the bbusiness case is positive, yieldding a net profit in 2023 of 17 - 36 M€. T The increased price volatilityy is an importtant driver for the profitabiliity of a CAES sttorage facilityy, in particularr during scarcity hours which drive up prrices to the loow VOLL of €320 € per MW Wh. While CAES S mainly provvides upward flexibility durring scarcity hhours reducinng unmet dem mand, it replacces some of thhe downward fleexibility from CCGT and cooal fired poweer plants. Furthhermore, windd curtailment iis slightly reduuced. T Table 9 Busin ness Case for CAES C Adiabatic  

 

Yearly discharge  Yearly ch harge  Revenuees  Chargingg costs  Yearly fixxed costs  Profit 

536 GWh 788 GWh 117 M€  45 M€  119 ‐ 38 M€  117 ‐ 36 M€ 

Minimum Prices P for P Profitability y of New Fllexible Sup pply Our results show that demand oon the intradayy market for rramping up is high, resulting in periods w with high pricees, which will inncite additionaal flexibility providers p to ennter the markeet. This will bbid down the price of flexibbility up till thhe point where eentrants will no n longer be aable to recoupp their fixed ccosts. In equillibrium, pricess on the intradday market w will therefore include a scarcitty rent in peaak demand hoours up and aabove the marrginal costs of o generation; otherwise neew generators woould not enterr the market. As A an indicatioon of this scarrcity rent, we hhave calculateed the averagee monthly pricces on the intradaay market at w which the business case for ggas-fired unitss and storage is i just positivee, or, in other words, at which price these teechnologies caan just recoup ttheir investmeent costs. m and miinimum upwaard Figuure 9 shows thhe monthly avverage spot prrices, balancinng prices on thhe intraday market, prices requireed for CCGT, GT, and storaage to break evven. The miniimum price foor CCGT is beelow the day-aahead spot pricce, illustrating thhe profitabilityy of a CCGT plant. For GT T, the minimuum price is moore or less equual to the upw ward adjustmeent price, illustraating its zero profit p at that prrice. The miniimum price foor storage is baased on the uppper limit of innvestment cossts of €1200 per kWe. Storagee requires a higgher monthly average price than a CCGT plant, but subbstantially low wer than those of gas turbine. Figu ure 9 Comparrison of 2023 market m pricess and break eeven prices foor flexible gen neration and sstorage

13

6. CONCLUSIONS In this study, the market for flexibility in the Netherlands has been quantified, focussing on the day-ahead, intraday, and balancing markets. Future electricity market developments are derived using the COMPETES European electricity market model, which is adapted for this study to include flexibility constraints and unit commitment. The analysis is based on assumptions regarding future developments of fuel prices, interconnections, demand and capacity mix in the individual European countries and renewable energy generation profiles derived from various sources such as the IEA’s World Energy Outlook 2013, ENTSO-E scenario’s, network development plans, and the renewable energy data from TSO’s and other sources. Our analysis has shown that higher share of wind power generation does not only increase the need for flexibility to accommodate wind variability, it also leads to an increased demand for flexibility to accommodate wind forecast errors. In an efficient intraday market, this results in additional revenues and positive business case for flexible generation and storage providing flexibility for real time balancing. Efficient intraday markets can contribute to accommodating increasing levels of variable and uncertain renewables in the electricity market. These markets will give a price incentive for both flexible generation and other sources of flexibility such as storage to provide the increased need for flexibility. On this market, the most cost-effective options will be selected to compensate higher or lower than forecasted power production from renewable energy sources, thereby reducing overall costs of integrating renewables in the electricity system. An important requirement for an efficient intraday market is program responsibility for all producers, including renewable energy power generators. Otherwise, renewable power generators will not have an incentive to trade on the intraday market. Without program responsibility for renewable generators, the intraday market will not develop and there will be no incentives to develop new sources of flexibility. Furthermore, there should be an incentive for balancing responsible parties to be active on the intraday market instead of leaving it to the TSO to balance the market. This requires that they bear the full costs of balancing incurred by the TSO. If this would not be the case, for example because part of these costs are socialized or because the price paid for imbalance is based on average costs instead of marginal costs, it would be less costly to leave balancing to the TSO and the intraday market would not develop. In essence, both - the responsibility and the incentives are part of the current balancing regime in the Netherlands. In our analysis, we have focussed on a single year based on a scenario for the future development of the electricity market. In actual investment decisions, the business case analysis also includes an analysis of uncertainty and potential risks, taking into account different assumptions for fuel prices, demand and generation mix developments. We therefore have not only focussed on the specific results for the business cases but also looked at the minimum prices needed for a positive business case. These calculations provide a kind of sensitivity analysis, indicating the range of market conditions in terms of prices that allow profitable investments for the supply of flexibility on the intraday market. While our assumptions for future developments have some impact on the results, they are likely to be robust regarding the increased demand for flexibility resulting from increased intermittent renewables generation. Probably the major factor which affects the results is the available capacity relative to net demand on both the day-ahead and intraday markets. Obviously, a situation of overcapacity will not allow all market participants to operate at a profit. However, in such a case it is to be expected that there will be a response of the market, which reduces overcapacity and allows generators to cover all their costs. Furthermore, we compute optimal dispatch for realized wind production on the intraday market, keeping import and export schedules from the day-ahead market fixed. While intraday markets can be expected to have a higher price due to the higher scarcity of capacity, a very large difference will incite generators to bid more on the intraday market and less on the day-ahead market; thereby increasing prices on the day-ahead market and driving down prices on the intraday market. Moreover, entrants (such as flexible conventional generation, storage and demand side response) are likely to have incentives to provide additional capacity. Although we have not calculated the final equilibrium on the intraday market, the minimum prices determined in the business cases provide some indication of equilibrium prices for individual technologies. A further step in the analysis of the impact of variable and uncertain renewables on day-ahead and intraday markets would be to model these markets explicitly in a dynamic setting in which not only generation but also investments are optimized. While developments and interactions with other European countries are taken into account for the day-ahead market, the system balancing against wind forecast errors is only analysed for the Netherlands, reflecting current practice in which system balancing takes place within countries. However, with further integration of electricity markets, balancing over a

14

larger geographical area can be expected to reduce overall balancing costs by improving the exchange of flexibility. This is also illustrated by the important role of imports and exports in providing flexibility on the day-ahead market as shown in our analysis. It would therefore be valuable to look at the effects of integrating both intraday and balancing markets across country borders. Furthermore, our analysis has been based on hourly data. However, volatility of renewable power generation is continuous. An analysis based on shorter time periods, such as 15 minutes, will probably show an increased demand and a higher value for flexibility. Finally, as an alternative flexibility option, demand side response (DSR) can potentially provide a cost-effective means to provide some of the flexibility needed to accommodate intermittent power generation from renewables. DSR encompasses a wide range of different models and options, from switch-off contracts between large customers and TSOs to flexibility provided by household appliances within a smart grid environment. To acquire a better understanding of the future role of DSR in accommodating intermittent renewables, it would first of all be necessary to gather more information on the potential and characteristics of DSR in the Netherlands. There is currently limited information in the literature regarding the characteristics of DSR such as potential MW, the incurred costs, the number of hours in which DSR would be available and the required warning time.

ACKNOWLEDGEMENTS This paper is based on the insights from research performed in several projects. We explicitly like to acknowledge the support of EDGaR and Tennet. In preparing this paper the authors have benefited from comments of Jeroen de Joode at ECN and the partners in the EDGaR A1 UGSIIMI. We would especially like to thank Professor Benjamin Hobbs and Robin Broder Hytowitz from Johns Hopkins University for their contributions in modelling and methodology used in this study.

APPENDIX Table 10 Flexibility Assumptions for conventional technologies in COMPETES Technology Time of Minimum Ramp rate Start-up cost being load (% of max (€/MW installed commissio (% of max capacity/hour) per start) ned capacity) 2010 50 20 46 ±14 2010 30 50 46 ±14 2010 35 40 46 ±14 2010 30 80 39 ±20 2010 10 100 16 ±8 2010 10 90 16 ±8 Source: Brouwer et. al. (2014)

Min up time

Min down time

8 8 8 8 8 8 8 8 8 1 1 1 1 1 1 1 1 1

4 4 4 4 4 4 4 4 4 3 3 3 1 1 1 1 1 1

15

REFERENCES Abbasy, A., van der Veen, R., & Hakvoort, R. 2011 (May). Possible effects of balancing market integration on performance of the individual markets. Pp: 608-613, Proc. 8th International Conference on European Electricity Markets, 2011. ACER (2013). Capacity Remuneration Mechanisms and the Internal Market for Electricity, Report, 30 July 2013. Brouwer, A., M. van den Broek, A. Seebregts, and A. Faaij (2014): “Impacts of large-scale intermittent energy sources on electricity systems, and how these can be modelled”, Renewable and Sustainable Energy Reviews (forthcoming), 2014. Daniels B.W., van Drill A.W.N. (2007), “Save production: A bottom-up energy model for Dutch industry and agriculture”, Energy Economics, vol. 29, no. 4, pp. 847-867, 2007. DNV KEMA (2013). Systems Analyses Power to Gas: A Technology Review, Report, 3 June 2013, Groningen ENTSO-E SO&AF (2013): Scenario Outlook and Adequacy Forecast 2013-2030, Report, 2013, Available at https://www.entsoe.eu/about-entso-e/system-development/system-adequacy-and-market-modeling/soaf-2013-2030/ Garver L. (1962), “Power generation scheduling by integer programming development of theory,” Power Apparatus and Systems, Part III. Transactions of the American Institute of Electrical Engineers, vol. 81, no. 3, pp. 730–734, 1962. Hobbs, B. and F. Rijkers (2004), "Strategic generation with conjectured transmission price responses in a mixed transmission system I: Formulation," IEEE Transactions on Power Systems, vol. 19, no. 2, pp. 707-717, 2004. Hobbs, B., G. Drayton, E. Fisher and W. Lise (2008), "Improved transmission representations in oligopolistic market models: quadratic losses, phase shifters, and DC lines," IEEE Transactions on Power Systems, vol. 23, no. 3, pp. 1018-1029, 2008. IEA (2010), Projected Cost of Generating Electricity, Report, 2010. Kasina S., Wogrin S., Hobbs B.F. (2014), A comparison of unit commitment approximations for generation production costing., Working paper, 2014 Available at: https://jshare.johnshopkins.edu/bhobbs1/site/docs/papers/KasinaWorginHobbsUnitCommitment.pdf Lise, W., J. Sijm and B. Hobbs (2010), "The impact of the EU ETS on prices, profits and emissions in the power sector: Simulation results with the COMPETES EU20 model," Environmental and Resource Economics, vol. 47, no. 1, pp. 23-44, 2010. Özdemir, Ö., Joode, J. De, Koutstaal, P., Hout, M. van. (2013): Generation capacity investments and high levels of renewables: The impact of a German capacity market on Northwest Europe., Discussion Paper, Energy Research Centre of the Netherlands (ECN), ECN-E—13-030. Petten/Amsterdam, 2013. Özdemir Ö., J. Hers, E. Fisher, G. Brunekreeft and B. Hobbs (2009), "A nodal pricing analysis of the future German electricity market," in Proc. 6th International Conference on European Electricity Markets, 2009. PBL & ECN (2013): Uitgangspunten voor het Referentiepad bij de Evaluatie van het SER Energieakkoord, PBL Report no: 1214, 2013, Bilthoven and Petten/Amsterdam. UDI (2012): Word Electric Power Plants Database, Utility Data Institute, 2012, Washington.