An agent-based analysis of main cross-border balancing arrangements for Northern Europe

Transcription

1 1 An agent-based analysis of ain cross-border balancing arrangeents for Northern Europe R. A. C. van der Vee A. Abbasy, and R. A. Hakvoort Abstract The topic of electricity balancing arket integration is first and foreost concerned with the distinction between different cross-border balancing arrangeents. Four ain arrangeents are ACE netting, BSP-TSO trading, an additional voluntary pool, and a coon erit order list. In order to investigate the ipact of different cross-border balancing arrangeents, an agent-based odel has been built, which siulates the operation of the balancing arkets of the, Gerany and the Nordic regio including the surplus and balancing exchange resulting fro integration. The results show that different arrangeents have a different ipact on ibalance costs for the arket. For the case of Northern Europe, the coon erit order list results in the lowest ibalance costs, but ACE netting and the voluntary pool are also beneficial integration steps. Index Ters agent-based odelling, ancillary service arket, balance responsibility, balancing arket, balancing services, cross-border balancing, electricity arket, arket behaviour, arket desig arket integration I. NOMENCLATURE ACE = Area Control Error BRP = Balance Responsible Party BSP = Balancing Service Provider PTU = Progra Tie Unit TSO = Transission Syste Operator II. INTRODUCTION A. Balancing arket LECTRICITY balancing arkets are institutional Earrangeents that safeguard the continuous balancing of supply and deand in liberalized electricity arkets. A balancing arket basically consists of three pillars: balance planning, balancing service provisio and ibalance settleent. Also, it includes three different actors: the Transission Syste Operator (TSO), the Balance Responsible Parties (BRPs), and the Balancing Service This work is supported by Next Generation Infrastructures Foundatio and part of the project Balance Manageent in Multinational Power Markets. R. A. C. van der Veen is with Delft University of Technology, faculty Technology, Policy and Manageent, 2628 BX Delft, the (eail: r.a.c.vanderveen@tudelft.nl). A. Abbasy is with Delft University of Technology, faculty of Technology, Policy and Manageent, Delft, the (e-ail: a.abbasy@tudelft.nl). R. A. Hakvoort is with Delft University of Technology, faculty of TPM, Delft, the (e-ail: r.a.hakvoort@tudelft.nl). Providers (BSPs). Balancing service provision concerns the role of the BSPs, who offer balancing services to the TSO by subitting balancing bids into the balancing arket before real-tie. In real-tie, the TSO activates the cheapest upward and/or downward regulation bids in order to restore the syste balance. Regulation prices are deterined to settle activated bids with. Balance planning and ibalance settleent are related aspects, which concern the role of the BRP. Before real-tie, BRPs ust subit schedules to the TSO, indicating planned electricity production and consuption. After real-tie, deviations fro those schedules (ibalances) are settled with an ibalance price, which is typically based on the regulation price. The basic tie unit in the balancing arket is the Progra Tie Unit (PTU). For each PTU, schedules and balancing bids are subitted, regulation prices and ibalance prices are deterined, and dispatched balancing and BRP ibalances are settled. B. Balancing arket integration In the light of the creation of an Internal Electricity Market for Europe, balancing arket haronization and integration have received the interest of European TSOs, regulators, the European Coissio and the electric power industry. The ost concrete eleent of this is the integration of balancing arkets for secondary and tertiary control, which entails the exchange of balancing between control areas. Such integration is believed to reduce the costs of realtie syste balancing due to enhanced copetition and higher efficiency of allocation of balancing resources. Different arket arrangeents are possible, which realize different degrees and fors of balancing arket integratio and have a different ipact on balancing arket perforance. We ake a fundaental distinction between four ain crossborder balancing arrangeents, naely Area Control Error (ACE) netting, BSP-TSO trading, an additional voluntary pool, and a coon erit order list. C. Ai It is the ai of this work to extend our understanding of the varying ipact of different cross-border balancing arrangeents on balancing arket perforance, by continuing earlier work on an agent-based analysis of the ipact of alternative arrangeents in Northern Europe, i.e. the, Gerany and the. D. Structure In Section III, a literature review on balancing arket integration and cross-border balancing arrangeents is

2 2 presented. The in Section IV, the general structure of the agent-based odel that has been used for the analysis is briefly described. Next, the specific arrangeents that have been odelled and studied in the analysis are given in Section V. This is followed by an overview and discussion of the ain analysis results in Section VI. Last, the conclusions are presented in section VII. III. LITERATURE OVERVIEW Apart fro the technical coordination between the control areas with regard to the reservation of frequency control services [1], balancing arkets in Europe are predoinantly national. The ain exception is the Nordic regio which has a regional regulating power arket in operation since In the light of the general goal of the European Union to create a single Internal Electricity Market in Europe, which anifests itself in the recent and on-going day-ahead and intra-day arket coupling projects, balancing arket integration appears to be a logical follow-up step. Balancing arket integration has been discussed by forer TSO organization ETSO [2], electricity industry organization Eurelectric [3], and by regulator organization ERGEG [4], the last of which presents general integration guidelines, and states that a lack of integration of balancing arkets is a key ipedient to the developent of a single European electricity arket. Last but not least, the European Coission has coissioned a report dedicated to balancing arket integratio which coes up with a practical road ap consisting of three consecutive phases [5]. As stated above, balancing arket integration is in essence about the integration of balancing service arkets, notably the balancing arkets for secondary and tertiary control. ETSO was the first to distinguish between cross-border balancing odels [6, 7], and in other references on this topic such a distinction has also been ade, which indicates its relevance with regard to desig ipleentation and ipact. Based on the ETSO odels, we have defined four ain cross-border balancing arrangeents: ACE netting, BSP-TSO trading, an additional voluntary pool, and a coon erit order list [8]. ACE netting erely involves the prevention of opposing upward and downward regulation in adjacent control areas (balancing arkets). BSP-TSO trading enables Balancing Service Providers (BSPs) to bid into the arket of another control area. The last two arrangeents are fors of TSO-TSO trading, where in an additional voluntary pool TSOs share balancing services with each other on a voluntary basis, whereas the coon erit order list concerns the full integration of the different arkets into a single ultinational (regional) arket for real-tie syste balancing of the entire balancing region [8]. Fro the aforeentioned organizations and corresponding references, ETSO and Eurelectric discuss different arrangeents, but they do not explicitly advocate a specific cross-border balancing arrangeent. The report for the EC calls TSO-TSO trading preferable above BSP-TSO trading. Also, it states that the exchange of excess services (which would reseble what we call an additional voluntary pool) can be ipleented already in an early phase, but that a coon erit order list requires haronization of aong others the reuneration schee. Finally, a short analysis of the ipact of a coon erit order list, based on Belgian and French bid prices, indicates that a costs reduction in the order of 5% is very well possible, leading to the conclusion that the ipleentation of a cross-border balancing arket is a lucrative and achievable goal [5]. The ERGEG report has adopted the view of this report, by calling the coon erit order list the target odel for the exchange of anual balancing services, and by seeing the voluntary pool as a possible first step, and BSP-TSO trading as an option in case the other two are infeasible (on a short notice) [4]. The literature does not provide with an in-depth systeatic analysis of alternative cross-border balancing arrangeents; the iplicit arguent behind the preference for a coon erit order list is that this arrangeent establishes both the ost efficient allocation of available balancing resources and axiizes copetition between BSPs, resulting in the lowest total balancing costs. In earlier work, we have engaged in a systeatic analysis of arrangeents in order to extend the arguent and create ore insights into the ipact of balancing arket integration [8-11]. In [8], a qualitative analysis of the ain cross-border balancing arrangeents confired the superiority of the coon erit order list, but identified ACE netting has a beneficial first step. Also, it was found that the different arrangeents result in varying perforance which depends on the detailed arrangeent desig the balancing arket design and overall power syste and arket conditions as well. Furtherore, quantitative analyses have been carried out to produce ore concrete results and insights, all of which focussed on the case of Northern Europe. First, fro an optiization odel study it was concluded that the ipleentation of a coon erit order list potentially reduces the balancing costs by 50% under the absence of interconnection capacity constraints [9]. Second, an agentbased analysis dedicated to the ipact of BSP-TSO trading for the and Norway shows that the balancing arket price in Norway does not change noticeably thanks to the huge excess of supply and the flat bid ladder of this hydrobased power arket, but that the Dutch downward regulation price is negatively affected during peak hours as a result of the export of downward regulation bids to Norway [10]. Third, an agent-based analysis of the ipact of ACE netting, fully efficient balancing trading (which corresponds with an optially functioning BSP-TSO trading or voluntary pool arrangeent) and a coon erit order list caused a balancing costs reduction of 40-50% due to the iport of the ajority of the fro the [11]. This paper continues on the last entioned analysis by studying and coparing all ain cross-border balancing arrangeents. Finally, related research by colleagues included an analysis of the ipact of balancing arket integration in Northern Europe with a linear optiisation odel, which also included a coon day-ahead arket and both a resource procureent and real-tie syste balancing phase. Here a balancing costs reduction for the coon erit order list of 65% was found [12].

3 3 IV. MODEL DESCRIPTION Agent-based siulatio or Agent-Based Modelling (ABM), is a odelling paradig that focuses on the odelling of individuals who ake autonoous decisions, which are however directly or indirectly influenced by the decisions of others. Because balancing arket perforance is principally deterined by behaviour of Balance Responsible Parties (BRPs) and Balancing Service Providers (BSPs), ABM is a suitable ethodology for the analysis of balancing arkets. The agent-based odel that is used for the analysis of crossborder balancing arrangeents is built in MATLAB, and has been used as well in the preceding agent-based analysis. Therefore, a ore elaborate version of the below description of the odel structure can be found in [11]. A. Model structure The odel represents the interconnected balancing arkets of the, Gerany and the. The only physical power syste eleents included are the interconnectors between the three arket areas, which for a constraint for the execution of cross-border balancing. The agents that are odelled are the Balance Responsible Parties (BRPs), which eans that analysis takes a BRP perspective, and takes into account the ipact of integration on BRP behaviour. The behaviour of BSPs is not included, except for, to a liited extent, the BSP-TSO trading arrangeent (see Section V). This liits the value of the analysis, but also siplifies it. The conceptual structure of the balancing arket, as applied for all three areas in the odel, is shown in Fig. 1. Up & down regulation bid ladder Unintentional ibalances BRPs BRP nuber and portfolio sizes Activated up & down regulation Syste ibalance Ibalance volue BRPs Intentional ibalances BRPs Decision rules BRPs Up & downregulation price Syste level Agent level Short & long ibalance price Actual Ibalance Costs BRPs Day-ahead arket price Fig. 1. Conceptual structure of a balancing arket within the odel Each balancing arket consists of a fixed aount of BRPs who have a certain portfolio size (in MW), i.e. a total su of production and consuption connections for which a single BRP has balance responsibility. Next to that, each balancing arket has an up- and down-regulation bid ladder, consisting of a specific set of upward and downward regulation bids each with a fixed volue (in MW) and a price (in /MWh). Finally, each balancing arket has a fixed day-ahead arket price. The odel siulation consists of ultiple rounds. Each round equals a Progra Tie Unit (PTU). In each round, all BRPs decide on a specific intentional ibalance volue out of a fixed set of options, using soe decision rules (decisionaking algorith), which ake the choice of each BRP dependent on the individual Actual Ibalance Costs (AIC) that it ended up with for different options in previous rounds. At the start of each round, all BRPs decide on an intentional ibalance option. In additio each BRP is faced with an unintentional ibalance, which arises fro a consuption/production forecast error. The su of the individual intentional and unintentional ibalance is the BRP ibalance. The the syste ibalance is calculated; it is the su of all BRP ibalances in the arket. Next, aounts of required upward and downward regulation are deterined based on the syste ibalance. These aounts are then provided by the activation of up- and down-regulation bids fro the respective bid ladders. Because the bids are ranked in price order, the cheapest bids are activated. The next step is fored by the deterination of the up- and down-regulation price with which dispatched balancing of BSPs is settled, and the short and long ibalance price with which individual ibalances of BRPs are settled. This is subject to soe specific pricing echaniss. Next, the Actual Ibalance Costs of each BRP are calculated, which are the product of the difference between the day-ahead price and the relevant ibalance price and the BRP ibalance volue, see Equation (1). AIC ( P = ( P 0 si, da, P P da, li, ) IV ) IV if IV if IV if IV < 0 > 0 = 0 In this equatio AIC are the Actual Ibalance Costs for BRP n in round, P si, is the short ibalance price in round, P li, is the long ibalance price in round, P da, is the day-ahead arket price in round, and IV is the ibalance volue of BRP n in round. Finally, at the end of each round, for each BRP the expected AIC of the selected intentional ibalance option in that round is updated according to equation (2): E ( AIC) X = ( E( AIC) X * Precency + AIC r) / (Precency + 1) (2) In this equatio E(AIC) X are the expected AIC for BRP n of option X selected in the active round, P recency is the recency paraeter, and AIC r are the Actual Ibalance Costs for BRP n in active round r. Now going forward to the start of the next round, the probabilities of choosing specific intentional ibalance options is inversely proportional to the expected AIC of these options. This way, the agents (BRPs) keep on experienting, while still learning fro the results of past rounds and aking (partly) rational decisions. The recency paraeter has been included to ake the results of recent rounds weigh heavier in the decision-aking. (1)

4 4 B. Model input The ost iportant odel input values are listed in Table I. A ore detailed table and description is given in [11]. TABLE I MODEL INPUT VALUES Areas Paraeters (NL) Gerany (DE) Nordic region (NO) Intentional ibalance options (%) -2.0 / -1.0 / -0.5 / 0.0 / 0.5 / 1.0 / 2.0 PTU (in.) 15 Total portfolio size (MW) 25, ,000 82,500 Day-ahead arket price ( /MWh) Regulation pricing arginal pay-asbid arginal echanis Ibalance pricing echanis dual average single Up- and down-regulation bid ladder fixed set of bids (including bid volue and price) Total capacity up (MW) 600 4,000 12,000 Total capacity down (MW) ,000-10,000 Transfer cap. NL-DE (MW) 2,300 Transfer cap. DE-NO (MW) 1,700 Transfer cap. NO-NL (MW) 700 Cross-border s (MW) fixed series of s for three lines Nuber of rounds 1,000 Recency paraeter 0.9 The BRPs can choose fro a set of seven intentional ibalance options, which are represented by percentages of the portfolio size, naely -2%, -1%, -0.5%, 0%, 0.5%, 1% and 2%. The PTU length used in the odel is 15 inutes, which is actually applied in the and in Gerany (the Nordic one is actually 60 inutes, but has been changed to ake cross-border balancing possible). The total portfolio size of all BRPs in a arket area is in line with the su of the su of average production and consuption (two ties the average syste load) in the three areas, which eans that the Geran and the Nordic arket are about five resp. three ties bigger than the Dutch arket. The day-ahead arket prices are based on the average dayahead power exchange prices in the areas in The included pricing echaniss are based on the actual echaniss used in the different arkets. In the and the Nordic regio the price of the last activated bid becoes the regulation price; in Gerany each activated bid is awarded its own bid price. Because of the latter, the ibalance price in Gerany is the weighted average of activated bid prices. The applies dual pricing, which eans the long ibalance price is set equal to the arginal downregulation price and vice versa, except when single-sided regulation occurs, in which case single pricing is applied. Single pricing eans here setting both the long and short ibalance price equal to the arginal price in the ajor regulation direction. For the Nordic regio this is only applied to consuption balances in reality, but we apply this to the entire arket for the sake of (relative) siplicity [13]. The up- and down-regulation bid ladders of the different areas are based on regulation (bid) data fro 2009 [14-16]. It ust be noted that not a lot of detailed bid data could be obtained, which eans that the accuracy of the bid ladders is not very high. The first parts of the up- and down-regulation bid ladders for the three areas are shown in Fig. 2. It can be noticed that the Dutch ladder is steepest, whereas the Nordic ladder is the flattest. In additio the has a very large over-supply of bids. Thus, the has both the cheapest balancing resources of Northern Europe, and a large abundance of resources, which points to a high econoic value of balancing arket integration in this region. Fig. 2. Fixed up- and down-regulation bid ladders of the three areas. NL =, DE = Gerany, NO =. Furtherore, there are fixed cross-border capacity and crossborder values for the three interconnections between the areas, which are based on ENTSO-E data. The fixed crossborder capacities are derived fro D-1 NTC values, whereas the cross-border data series are directly taken fro a period in 2010 [17]. Finally, one siulation run is 1,000 rounds long, which corresponds with at least ten full days, and the recency paraeter is set to 0.9. V. MODELLED ARRANGEMENTS In order to analyse the ipact of the cross-border balancing arrangeents on balancing arket perforance we have odelled six arrangeents, which are illustrated in Fig. 3. They are presented in order of increasing degree of integration. Included are the four ain cross-border balancing arrangeents introduced in Section III, a odel with axially efficient balancing trading called Balancing trading, and the reference case of separate arkets. A short explanation of the odelled arrangeents is given below. A. Separate arkets In this reference arrangeent, with which the other arrangeents will be copared, the three balancing arkets of the, Gerany and the are operating independently. The interconnection lines are not used. B. ACE netting In this arrangeent, Area Control Error (ACE) netting occurs. For this, the syste ibalances of the three areas are copared. The occurrence of opposite syste ibalances will cause surplus s fro the surplus area to the deficit area, resulting in reduced activation of upward/downward regulation in the deficit/surplus area.

5 5 Syste ibalances are copletely reoved when possible, otherwise the netting of ACEs takes place proportionally (cf. [11]). The availability of interconnection capacity is checked. Ib. prices NL 700 MW Ib. prices NO 2300 MW 1700 MW Ib. prices DE Gerany Separate arkets Ib. prices NL Ib. prices NO Ib. prices DE Gerany BSP TSO trading Ib. prices NL Ib. prices NO Optial allocation Gerany Balancing trading Balancing Balancing Ib. prices DE Ib. prices NL Ib. prices NL Ib. prices NO Reduced regulation Ib. prices DE Gerany ACE netting Ib. prices NO Regional pool Surplus Balancing Ib. prices DE Gerany Voluntary pool Regional regulation prices Regional ibalance prices Regional bid ladder Balancing Gerany Coon erit order list Fig. 3. The six odelled cross-border balancing arrangeents C. BSP-TSO trading Here, balancing bids ays be offered into the balancing arket of another area instead of the own. Specifically for this arrangeent, the BSPs are represented by agents that are responsible for the subission of one specific bid. Each round each BSP has to choose a arket/area in which to bid, but the bid price and volue reain fixed. The chance of choosing a specific arket is proportional to the average profit that the BSP gained over the previous rounds in which it bid in that arket. On top of that, there is a 10% that the BSP will randoly select a arket. Like in the reaining arrangeents, the availability of cross-border capacity for the exchange of balancing is checked before (a bid) is iported or exported. If there is capacity, the exchange is put through and the capacity value is reduced accordingly; if there is not, the bid is skipped. D. Voluntary pool In the odelled version of the additional voluntary pool each arket keeps a certain national share of bids for national use, which is set equal to 2% of the total portfolio size of the areas. These include the cheapest bids of the area. Reaining bids are offered to the other areas. If the bid price of an offered bid is lower than the bid price of the last (ost expensive) bid of the national share of the area it is offered to, it will be included in the bid ladder of that area. Otherwise, it will be offered to the third area using the sae rules. If it is not included there either, the bid is not included in any bid ladder. The above procedure for the creation of three national bid ladders only takes place at the start of the siulation run. E. Balancing trading In this arrangeent, a regionally optial allocation of balancing services takes place for the real-tie syste balancing of the three areas. This includes ACE netting. The balancing arket rules reain national, and are the sae as in separate arkets. Actually, this arrangeent can be considered to represent a BSP-TSO trading or voluntary pool arrangeent in which an optial allocation of balancing services occurs. However, it can also be considered to be a coon erit order list where national pricing echaniss still apply. F. Coon erit order list The balancing arkets of the three areas are integrated into one regional bid ladder (separately for upward and downward regulation), and a regionally optial allocation of balancing services takes place, like in the last arrangeent. This is odelled in the following way. First, ACE netting is perfored. The bids are considered for activation in order of increasing bid price. If a bid can be utilized to reduce the syste ibalance of the own area, it will be done so. Otherwise it will be utilized in one of the other areas, which is subject to the availability of cross-border capacity. VI. ANALYSIS RESULTS Each odel versio representing a different cross-border balancing arrangeent, has been run ten ties, and the averages of the output variable values have been taken to end up with the final analysis results. The description of the analysis results is divided between general and arrangeentspecific results. A. General results The ost striking ipact of the introduction of different cross-border balancing arrangeents is on the realized surplus exchange for ACE netting and the balancing exchange for the voluntary pool, balancing trading and the coon erit order list. The surplus exchange in the ACE netting arrangeent reduces the dispatched balancing in the three areas by less than 25% (see below), and thereby causes a liited reduction of the total real-tie balancing costs and ibalance costs. In the last four arrangeents however, the vast ajority of the balancing that is dispatched in the and Gerany to restore the syste balance is iported fro the. In Fig. 4 is can be seen that for BSP-TSO trading, balancing trading and the coon erit order list the iport percentages for these two areas are higher than 50% and can becoe as high as 80%. The Nordic regio however, iports nothing in all arrangeents except for BSP-TSO trading (see below). These enorous exchange volues are the result of the cheap and abundant resources in the Nordic regio and the

6 6 availability of cross-border capacity, which proved to be large enough to enable this level of balancing exchange. Moreover, the detailed analysis results show only the interconnector between the and the was a large constraint for exchange: in 60% of the rounds this line was congested in the direction fro the to the. Fig. 4. The percentage of dispatched balancing that is iported. NL =, DE = Gerany, NO =. ibalance prices. However, the ost iportant indicator fro the perspective of the electricity arket as a whole is fored by the total Actual Ibalance Costs (AIC), as introduced in Section IV. The AIC reflect the actual costs of real-tie balancing for the arket, because they can be considered the opportunity costs of trading in the day-ahead or intraday arket to prevent ibalances. The total AIC are the su of all actual ibalance costs incurred by all Balance Responsible Parties in an area over the entire siulation run. Fig. 6 gives the total AIC for the three areas for all odelled cross-border balancing arrangeents. It can be observed that the total AIC of the and Gerany generally decrease iensely, by 50-80%, but that the total AIC of the Nordic region reains on the sae level, or even increases significantly (see below). This can be directly explained by the high degree of balancing iports of the and Gerany fro the : These largely reduce regulation and ibalance prices, and thereby the Actual Ibalance Costs, whereas the costs in the cannot reduce uch further as it already has the cheapest balancing resources in case of separate arkets. In Fig. 5, the total iported and exported balancing volues over the entire siulation run are indicated for the three areas, split out for up- and down-regulation. This shows that the volues iported by Gerany fro the are uch higher than the volues iported by the, which is caused by the uch lower deand and the liited available cross-border capacity. Furtherore, Fig. 5 shows that the level of iport of Nordic upward regulation bids is higher than that of downward regulation bids. Fig. 6. Ipact of arrangeents on total Actual Ibalance Costs Fig. 5. Total balancing iport and export volues Next, it can be noted that the exports a significant aount of balancing, which goes to Gerany. This is partly caused by a suboptial activation procedure in the odel versions of the last two arrangeents. Finally, soe peculiar differences between arrangeents stand out; these will be discussed below. In general, balancing exchange results in reduced up- and down-regulation prices, which in turn leads to lower The reduction of Actual Ibalance Costs for the BRPs is also reflected by the average penalty (in /MWh) for BRPs to be long (BRP surplus) and to be short (BRP shortage), which are shown in Fig. 7 and Fig. 8. These penalties are just the day-ahead price inus the long ibalance price resp. the short ibalance price inus the day-ahead price, and are the basis for the calculation for the AIC (see equation (1)). A positive penalty eans average costs (which the BRP pays); a negative penalty eans an average profit (which the BRP receives). The first thing that can be noticed is that the penalties generally reduce as well, which is in line with the total AIC reductions. However, it is rearkable that the penalties differ quite unsysteatically between arrangeents, and that the sign of the penalties often changes. This is caused by changes in the relative shapes of up- and down-regulation bid ladders effectuated by the integratio as a result of which it can suddenly be ore favourable for BRPs to opt for a BRP ibalance in the other direction. An iportant exaple can be provided by the coparison of separate arkets with the coon erit order list. In separate arkets, there is a general incentive for Geran BRPs to be long and for Nordic

7 7 BRPs to be short as the corresponding average penalties are actually a profit to the. However, in the coon erit order list, these incentives are the other way around. Such shifts in incentives bring the BRPs in the odel to choose intentional ibalance options in the other directio which leads to shifts in the doinant syste ibalance direction. This could ean a negative effect of integration fro a security of supply perspective, because syste surpluses pose a lower security threat than syste shortages. Fig. 7. Ipact of arrangeents on the BRP penalty to be long Fig. 8. Ipact of arrangeents on the BRP penalty to be short Also, it should be noted that single ibalance pricing is applied in all cases, except for the in the first five arrangeents. This eans for the Dutch BRPs relative high ibalance costs, which are largely reduced in the coon erit order list due to introduction of regional single pricing. Single pricing generally leads to an evening out of ibalance costs for BRPs, but on the syste level high penalties will still introduce high financial risks, as the total AIC values show. B. Arrangeent-specific results A discussion of the ipact of individual cross-border balancing arrangeents on balancing arket perforance in Northern Europe is given below. 1) Separate arkets: Most striking is that the total Actual Ibalance Costs (AIC) of the are ore than 50% lower than those of the, whereas the syste is three ties bigger. This is caused by the presence of a huge oversupply of very cheap hydro-based balancing resources in the Nordic regio and by its use of single pricing. 2) ACE netting: ACE netting reduces the activated up- and down-regulation volues in all areas, which results in ore favourable regulation prices, and therefore lower total AICs for the different areas. However, the AIC reduction is largest for the, about 25%, because of its saller syste size and the use of dual pricing [cf. 11]. 3) BSP-TSO trading: This arrangeent leads to a uch lower total AIC for the and Gerany, but a uch higher total AIC for the. This is caused by the assive exodus of bids fro the to Gerany and the, leaving the behind with a relatively sall aount of bids. As a result, the iports 12% of the dispatched downward regulatio and the ibalance penalties are really high, with BRP shortages being uch ore favourable then BRP surpluses. The latter has led to a very high syste shortage occurrence of 70%. 4) Voluntary pool: Copared with BSP-TSO trading the total AIC reductions fro the ipleentation of the voluntary pool are a lot saller for Gerany, and even ore so for the. On the other hand, the total AIC of the Nordic region reains about the sae. This can be explained by the national share of bids that stays within the own area, which results in a uch lower iport of balancing fro the. The Dutch iport percentage is only 5%, because the cheapest Dutch bids have a ore favourable bid price than the cheapest Nordic bids that are shared with the other arkets in this arrangeent. This situation holds to a uch lower extent for Gerany. 5) Balancing trading: The Dutch and Geran total AIC reductions are clearly lower than for the forer arrangeents, but here too the Nordic total AIC reain the sae (as the has the cheapest bids of the North- European region). The first is caused by the iport of 60-80% of the balancing fro the Nordic regio and the execution of ACE netting. 6) Coon erit order list: The total AIC values are even a bit lower for the and Gerany copared to balancing trading, which is probably caused by the change to single ibalance pricing. The Nordic total AIC is uch higher than in the balancing trading arrangeent, however, which is the effect of the regional arginal pricing by which ibalance prices are based on the arginal bid prices of the activated bids in the entire region. As a reark, the results show that the unifor regulation and ibalance prices lie in the iddle of the range of prices in separate arkets, which eans that prices within the region have converged. This will, like holds for arket integration in general, benefit soe areas ore than others, and ay also be unfavourable to soe, even though the net benefit for the region is clearly positive. C. Sensitivity analysis A sensitivity analysis has also been conducted. It turns out that the odel results are not very sensitive to the reduction of available cross-border capacity for cross-border balancing, even though an increase in congestions has been observed. This is probably caused by the fact that balancing can always be exchanged in the direction opposite to the oentary cross-border, in cobination with a siultaneous deand for both up- and down-regulation within the sae areas for ost PTUs. Furtherore, the ipact of sall bid ladder changes do not significantly affect the results, but obviously, large changes do, which is a ain reason why the ipact of integration is case-dependent.

8 8 VII. CONCLUSIONS Based on the agent-based analysis of six different crossborder balancing arrangeents for the case of Northern Europe, it is concluded that the choice of a cross-border balancing arrangeent as the key decision in the balancing arket integration process has a very large ipact on balancing arket perforance. This is clearly shown by the varying ipact on the total Actual Ibalance costs (AIC), i.e. the costs of real-tie balancing for the arket. The total AIC of the and Gerany are largely reduced by ost arrangeents, whereas the Nordic total AIC reain the sae or even increase, which is the result of the iport of the vast ajority of the balancing in the and Gerany fro the Nordic regio which has a huge oversupply of balancing bids that are also cheapest in the North-European region. Ibalance penalties are also reduced, but the signs are often shifting, resulting in shifts in doinant syste ibalance directions. The analysis shows that, for Northern Europe, the ipleentation of ACE netting and a voluntary pool have a sall positive effect on the total AIC for all three areas, but that balancing trading causes a very large AIC costs reduction for all three areas. The coon erit order list leads to siilar AIC reductions as balancing trading for the and Gerany, but the costs for the Nordic region are higher due to the regional arginal pricing. BSP- TSO trading leads to a huge increase of total Actual Ibalance Costs for the Nordic regio because any of the cheapest Nordic bids subit their bids in another arket. This does not occur in the voluntary pool due to the national share that is not exchanged, which however liits the AIC reductions for the and Gerany as well. Viewing the ipact of arrangeents on total ibalance costs in Northern Europe, we can recoend ACE netting and the additional voluntary pool as beneficial balancing arket integration steps, and the coon erit order list as the ost beneficial and final step. This is in line with the recoendations in [5]. The results for the odelled balancing trading arrangeent indicate that an optially functioning voluntary pool, or a erit order list in which national pricing echaniss are still used, ight however be preferable in order to prevent an ibalance costs increase in the. on the ipact of cross-border balancing arrangeents on balancing arket perforance ipossible. Moreover, the arrangeents can take any different fors, whereas the odelled arrangeents involve specific assuptions and operational rules, which prevent fir conclusions regarding the ipact of different arrangeents for Northern Europe. REFERENCES [1] UCTE, P1 Policy 1: Load-Frequency Control and Perforance [C], part of UCTE Operation Handbook, March [2] ETSO, Balance Manageent Haronisation and Integration - 4th Report, Jan [3] Eurelectric, Towards Market Integration of Reserves & Balancing Markets, July [4] ERGEG, "Revised ERGEG Guidelines of Good Practice for Electricity Balancing Markets Integration (GGP-EBMI)," Ref: E09-ENM-14-04, Sep [5] KU Leuven en Tractebel Engineering, Study of the interactions and dependencies of balancing arkets, intraday trade and autoatically activated reserves, final report, TREN/C2/84/2007, Feb [6] ETSO, Current State of Trading Tertiary Reserves Across Borders in Europe, Nov [7] ETSO, Key Issues in Facilitating Cross-Border Trading of Tertiary Reserves and Energy Balancing, May [8] R. A. C. van der Vee A. Abbasy, and R. A. Hakvoort, A qualitative analysis of ain cross-border balancing arrangeents, presented at the 7 th Int. Conf. on the European Energy Market, Madrid, Spai [9] A. Abbasy, R. A. C. van der Vee and R. A. Hakvoort, Effect of Integrating Regulating Power Markets of Northern Europe on Total Balancing Costs, presented at the PowerTech Conference, Bucharest, Roania, [10] A. Abbasy, R. A. C. van der Vee and R. A. Hakvoort, Possible Effects of Balancing Market Integration on Perforance of the Individual Markets, paper for the 8 th Int. Conf. on the European Energy Market, Zagreb, Croatia, 2011 (forthcoing). [11] R. A. C. van der Vee A. Abbasy, and R. A. Hakvoort, Analysis of the Ipact of Cross-Border Balancing Arrangeents for Northern Europe, paper for the 8 th Int. Conf. on the European Energy Market, Zagreb, Croatia, 2011 (forthcoing). [12] S. Jaehnert and G. L. Doora Modelling An Integrated Northern European Regulating Power Market Based On a Coon Day-Ahead Market, conference paper, IAAE s Rio 2010 International Conference, Rio de Janeiro, Brazil, [13] O. S. Grande, G. L. Doora and B. H. Bakke Exchange of balancing resources between the Nordic synchronous syste and the / Gerany / Poland, SINTEF Energy Research, technical report, Feb [14] TenneT website, [15] Geran digital platfor for reserve tenders, [16] NordPool website, [17] ENTSO-E data website, Reflecting on the analysis, it ust be ephasized that this analysis has only taken a BRP perspective, that is, only the ipact of arrangeents on BRP behaviour has been taken into account. In reality, BSPs will probably adapt their bidding strategies as a result of integratio as new arket opportunities and risks eerge. If, as literature iplicitly argues, balancing arket integration indeed enhances copetition in balancing arkets, balancing costs and ibalance costs can be expected to further reduce. This will only hold if gaing opportunities or new fors of arket power do not eerge, which is soething we expect will not happen for Northern Europe given the huge over-supply of cheap Nordic balancing bids [cf. 8, 10]. A final reservation regarding the results is that the analysis is based on the power syste and (balancing) arket conditions of Northern Europe, which akes a generalization