CONGESTION IN THE ISO-NE ELECTRICITY MARKETS

Transcription

1 CONGESTION IN THE ISO-NE ELECTRICITY MARKETS BY ANNA BARBARA IHRIG THESIS Advisor: Prof. George Gross Submitted in partial fulfillment of the requirements for the degree of Diplom-Wirtschaftsingenieur, Fachrichtung Elektrotechnik of Darmstadt University of Technology, Germany University of Illinois at Urbana-Champaign, 2002 i

2 FOREWORD This technical report is a reprint of the thesis written by Anna Barbara Ihrig as the technische Studienarbeit, a partial fulfillment of the requirements for the degree of Diplom-Wirtschaftsingenieur, technische Fachrichtung Elektrotechnik at Darmstadt University of Technology, Germany. G. Gross Thesis Advisor June 2002 ii

3 ACKNOWLEDGEMENTS I would like to first thank my thesis advisor, Professor George Gross, for his knowledge and helpful guidance during my studies. Doing research with him was an invaluable experience. I would also like to thank all my friends at the University of Illinois for their support, especially Mary and my fellow students in the Power Group. Finally, I would like to thank my family and Niels for constantly encouraging me and being there when I needed them. iii

4 ABSTRACT The restructuring of the electricity industry in the U.S. has made the problem of transmission congestion increasingly significant. It frustrates the smooth functioning of competitive markets and typically high costs are associated with it, which have to be eventually borne by the consumers. This thesis aims to investigate the impacts of congestion on the electricity markets in one of the major power systems of the northeast, the New England system. We give a brief review of the structure of the New England system and the entity in charge of operation and control, the ISO-NE. We describe how the ISO-NE administers the energy market and manages congestion. A general definition of congestion is provided and the congestion metrics and market power mitigation procedures used in New England are discussed. The main effort lies in the analysis of congestion data over a 26-month period. Congestion quantities and costs are analyzed in detail and examined for discernible monthly and hourly patterns. A high volatility is found in both metrics and no distinct seasonal or hourly patterns can be detected. The geographic characteristics of the New England transmission area are discussed with respect to the distribution of congestion across the various transmission areas in the region. Three major congested areas are identified and examined. The role of generation availability as a factor in causing congestion is analyzed; no significant correlation was found. In addition, the impacts of the ISO-NE s market power mitigation on the congestion costs are also reviewed. The thesis concludes with a set of recommendations for future work. iv

5 TABLE OF CONTENTS 1. INTRODUCTION The problem of congestion Brief review of the previous work Scope and contribution of this thesis THE NEW ENGLAND POWER SYSTEM A brief history of the New England markets and the ISO-NE The transmission-unconstrained market CONGESTION A workable definition The Congestion Metrics used by the ISO-NE The ISO-NE Market Power Mitigation Procedures CONGESTION DATA ANALYSIS Data Description Analysis of monthly and hourly patterns The geographic distribution v

6 4.4 Selected transmission regions Impacts of market power mitigation Impacts of generation availability CONCLUDING REMARKS REFERENCES APPENDIX A APPENDIX B APPENDIX C vi

7 LIST OF FIGURES Figure 2.1: The geographic scope of the New England system... 7 Figure 2.2: Time line of ISO-NE energy market operations Figure 2.3: Information flow in the ISO-NE energy market Figure 2.4: The supply curve constructed from the offers Figure 2.5: Determination of the ECP for Example Figure 2.6: Determination of the ECP for Example 2, Set Figure 2.7: Determination of the ECP for Example 2, Set Figure 3.1: Unconstrained 2-bus system of Example 3a Figure 3.2: Constrained 2-bus system of Example 3b Figure 3.3: Unconstrained 5-bus system of Example 4a Figure 3.4: Constrained 5-bus system of Example 4b Figure 3.5: The supply curve with out-of-merit-order dispatch incorporating offer E Figure 3.6: The supply curve resulting in a lower ECP with transmission constraints considered Figure 3.7: Example of ISO-NE calculation of congestion costs Figure 3.8: ISO-NE market power mitigation procedure Figure 4.1: Monthly congestion for May 1999 June Figure 4.2: The overall upward trend of the monthly congestion for May April Figure 4.3: The overall downward trend of the monthly congestion for May 2000 June Figure 4.4: Seasonal behavior of monthly congestion for May 1999 June Figure 4.5: Daily and monthly average congestion for May 1999 June Figure 4.6: Congestion duration curve for May 1999 June Figure 4.7: Congestion duration curve for March Figure 4.8: Congestion duration curve for April Figure 4.9: Congestion duration curve for May Figure 4.10: Monthly mitigated congestion costs from May 1999 June vii

8 Figure 4.11: Monthly congestion and mitigated congestion costs for May 1999 June Figure 4.12: Daily mitigated congestion costs and the monthly averages for May 1999 June Figure 4.13: Mitigated cost duration curve for May 1999 June Figure 4.14:Mitigated cost duration curve for March Figure 4.15: Mitigated cost duration curve for April Figure 4.16: Mitigated cost duration curve for May Figure 4.17: Monthly per MWh mitigated costs for May 1999 June Figure 4.18: The general upward trend of the monthly per MWh mitigated costs for May 1999 April Figure 4.19: The lack of a definite trend in the monthly mitigated costs per MWh for May 2000 June Figure 4.20: Seasonal behavior of the monthly per MWh mitigated costs for May 1999 June Figure 4.21: Daily per MWh mitigated costs and monthly averages for May 1999 June Figure 4.22: Per MWh mitigated cost duration curve for May 1999 June Figure 4.23: Per MWh mitigated cost duration curve for May 1999 April Figure 4.24: Per MWh mitigated cost duration curve for May 2000 June Figure 4.25: Average hourly system load for May Figure 4.26: Average hourly congestion for March, April and May Figure 4.27: Average hourly congestion for August and November Figure 4.28: Average hourly system ECP and load for May Figure 4.29: Average hourly per MWh mitigated costs and ECP for August Figure 4.30a: Average hourly per MWh mitigated costs and congestion for March Figure 4.30b: Average hourly normalized mitigated congestion costs for March Figure 4.31a: Average hourly per MWh mitigated costs and congestion for April Figure 4.31b: Average hourly normalized mitigated congestion costs for April Figure 4.32a: Average hourly per MWh mitigated costs and congestion for August Figure 4.32b: Average hourly normalized mitigated congestion costs for August viii

9 Figure 4.33: Distribution of congestion in MW within transmission areas Figure 4.34: The contribution of transmission areas to the monthly congestion for May 1999 June Figure 4.35: Distribution of mitigated costs within transmission areas Figure 4.36: Per MWh mitigated costs and total mitigated costs for each area for May 1999 June Figure 4.37: The contribution of the transmission areas to the total monthly mitigated congestion costs for May 1999 June Figure 4.38: Unmitigated and mitigated monthly congestion costs for May 1999 June Figure 4.39: Percentage of reduction of monthly unmitigated costs by mitigation Figure 4.40: Daily outaged capacity for September 1999 June Figure 4.41: Daily congestion versus daily outaged capacity for September 1999 June ix

10 LIST OF TABLES Table 2.1: The New England transmission areas... 8 Table 2.2: Offer set for Example Table 2.3: The two offer sets in Example Table 3.1: Offers and demand quantities for Example Table 4.1: Frequency information for congestion over the May 1999 June 2001 period Table 4.2: Frequency information for daily mitigated congestion costs over the May 1999 June 2001 period Table 4.3: Frequency information for daily per MWh mitigated congestion costs over the May 1999 June 2001 period Table 4.4: Number of days with per MWh mitigated costs in certain range over the May 1999 June 2001 period Table 4.5: Months with significant correlation between congestion and outages x

11 1. INTRODUCTION In this chapter we review the nature of congestion and its impacts on electricity markets. We discuss the state of the art of the literature in congestion and present a brief summary of the previous work that was conducted in this field. In the last section, we outline the objectives, the scope and the contributions of this thesis. 1.1 The problem of congestion The restructuring of the electricity industry in the US has created a new regime with many new players. The myriad changes brought huge shifts in the planning, operations and management of power systems. The introduction of competitive markets did not only bring benefits, but also made the industry face unprecedented problems. Unlike other markets, the electricity market has salient characteristics, which make the operation of competitive markets a major challenge. The lack of major storage capability, the just-intime-manufacturing nature of electricity and the central role the transmission and distribution network plays are some of the principal complexities in electricity. A problem that is becoming more and more significant nowadays is transmission congestion. With the increasing number of market participants, the number of desired transactions between the various players is growing. Each transaction requires energy to be transferred from a sending point to a point of receipt. The sellers and buyers of energy rely on the transmission network for the transportation of their goods as sellers and buyers in other markets rely on trucks or trains. In the days before restructuring, the 1

12 power grid used to be operated by vertically integrated utilities, who had control over both generation and transmission facilities. Since the unbundling of generation and transmission and the advent of more decentralized decision-making, it has become a challenge to coordinate and operate the system. The current transmission networks were not originally planned for trading in a competitive market. In addition, one of the key characteristics of the electricity market is that the good that is traded, energy, will not necessarily take the direct route from the sender to the receiver, but will travel across the transmission system according to the laws of physics and is -- especially in a highly interconnected network -- likely to result in loop flows and affect various parts of the system. If market participants intend to undertake a high number of transactions to transfer energy between various points in the network, the realization of all schedules might lead to violations of one or more limits of the transmission system. This situation is called transmission congestion. Whenever this is the case, not all of the desired transactions can be realized. The market players value the transmission of energy differently and the fact of not being able to realize certain transactions can have severe impacts and cause high additional costs. Energy that cannot be purchased from the supplier who offers it at the lowest price because the current state of the transmission system does not allow the transfer, has to be purchased from an alternative resource at a higher price. The situation is especially severe if an area with high demand does not posses sufficient generation and relies on the import of energy from neighboring systems to serve the network load. In this case, congestion on the tie lines between the two regions can significantly endanger the ability of the system to meet its demand. 2

13 Since the opening of the transmission grid to independent market players, different power market systems and congestion management schemes have been developed in the U.S. to administer competitive market operations and at the same time maintain system reliability at the desired level. After some years of restructuring, operating rules and procedures are still constantly changing. The main effort lies in providing an effective market design for the restructured environment. One of the key requirements for the implementation of competitive markets is an effective management of congestion. 1.2 Brief review of the previous work Several studies have explored the impacts of congestion and different transmission and congestion management approaches. The study in [1] discusses three transmission management models that are implemented today and their impacts on the economics of the energy market: the optimal power flow model found in the United Kingdom and parts of the United States, the price area congestion control model used in Norway and Sweden and the U.S. transaction-based model. The work in [2] reviews different congestion management schemes and the associated pricing mechanisms in several countries and in some states of the U.S., and provides a unified framework for the study of various congestion management schemes implemented in different jurisdictions. Metrics are developed to assess the efficiency of the various schemes and the effectiveness of the market signals provided to the market participants. The paper in [3] provides background on the market structure in California and discusses the role of the California Independent System Operator as the facilitator of its congestion management process. In [4] the author 3

14 conducts an analysis of transmission congestion in the PJM market (Pennsylvania, New Jersey, Maryland, Delaware, D.C., Virginia) and discusses the method of flowgates as a market-based model to manage congestion. A key impediment to the functioning of competitive markets is the existence of market power. The occurrence of congestion can create the potential for certain market players to exercise market power for their own profit and thereby lower the benefits to consumers. This is asserted in the study of [5], which finds that transmission capacity limits have significant effect on outcomes of market gaming and that relieving transmission congestion can discourage the exercise of market power. The work in [6] critically discusses the existence of market power in electricity markets and presents a method to test for the exercise of market power examining optimal generator behavior by taking price as a given exogenous input. The study in [7] assesses the New England Electricity Market in comparison with California and PJM in terms of competitiveness. The study measures market performance by estimating the difference between actual prices and prices that would result if no market power were exercised. Congestion is considered as a situation that is conducive to the exercise of market power. The problem of congestion is recognized in federal circles and has prompted a number of studies. The Federal Energy Regulatory Commission (FERC) conducted a study on the various effects of transmission congestion in the U.S. [8]. Major transmission bottlenecks throughout the country were identified and the costs arising from congestion assessed. A study conducted for the Department of Energy recently investigated the transmission system in the U.S. and identified measures to eliminate transmission bottlenecks [9]. A number of recommendations were provided on the federal 4

15 role of assessing constraints and encouraging new transmission investment. Supportive papers discuss, for example, transmission planning and siting, transmission technologies and alternative business models for transmission investment. 1.3 Scope and contribution of this thesis In this thesis we study the nature and impacts of congestion in one of the major northeastern power systems -- the New England system. This system was one of the early ones in the US to implement restructuring by opening its transmission grid and harnessing the competition in wholesale electricity markets. Our objectives are to obtain a better understanding of the congestion in the New England system, to explore the impacts and ramifications of congestion on the New England markets and to quantify the effects of mitigation. The scope of our work focuses on the 26-month period of May 1999 to June We discuss the two metrics for congestion: total energy quantities and the associated costs. Congestion data are analyzed over the 26-month period and the identifiable characteristics, including temporal patterns, are explored. We study the possible causes and explanations for the observed congestion situations. We discuss regional characteristics of congestion in the New England power system. We analyze the impacts of market power mitigation on congestion costs in New England. In addition, we investigate the role of generation availability as a factor in causing congestion. Our study gives a detailed assessment of the occurrences of congestion in the New England market and the resulting costs during the 26-month period of study. This can 5

16 provide a basis and encouragement for further efforts to improve the efficiency of congestion management schemes. The thesis has three more chapters. Chapter 2 gives an overview over the background of the New England markets, the ISO-NE and the functioning of the transmissionunconstrained market. In Chapter 3, we provide a generic definition of congestion and the measures for congestion used in New England. We also describe the market power mitigation procedures applied by the ISO-NE. In Chapter 4, we study the congestion data in detail to analyze the principal characteristics including the presence of monthly and daily patterns. We present the findings on the geographic distribution of congestion in the distinct transmission areas of the New England system and discuss the importance of three specific areas in detail. We show the impacts of market power mitigation on congestion costs and investigate the implications of generation availability on the amount of congestion. The last chapter summarizes the key findings and points out directions for future work. 6

17 2. THE NEW ENGLAND POWER SYSTEM In this chapter we provide an overview of the New England power system. We briefly review the characteristics of the region, the organizational structure and the different markets. In the second section, we explain the functioning of the transmissionunconstrained energy market and give an example on strategic bidding behavior in which no transmission constraints are considered. 2.1 A brief history of the New England markets and the ISO-NE In 1971, the electric utilities in the six New England states Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island and Vermont formed an association called NEPOOL or the New England Power Pool, covering an area of 66,672 square miles. The Vermont New Hampshire Maine Massachusetts Connecticut Rhode Island Figure 2.1: The geographic scope of the New England system 7

18 objective was to establish a single regional network and centrally dispatch the generating units of the Pool [10]. NEPOOL operates as a tight power pool, in which Pool members coordinate their planning and the network monitors and directs the operations of virtually all the generating facilities. Today the Pool generation system comprises 350 generating units with a total capacity of 28,500 MW. To meet load requirements, the New England system must import power from its neighboring systems New York, New Brunswick and Quebec [7]. The 6.5 million customers are served using more than 8,000 miles of transmission lines. On August 9, 2001, New England experienced its record peak load of 25,038 MW. A map of the region is shown in Fig The New England transmission region is split into twelve transmission areas. These are shown in Table 2.1. Table 2.1: The New England transmission areas name of area description EAST WEST NORTH SOUTH VT W MASS CONN SW CONN SEMA/RI MAINE NEMASS/BOST ME/NH 345 & 115 KV East West North South Vermont West Massachusetts Connecticut Southwestern Connecticut Southeast Maine / Rhode Island Maine Northeast Massachusetts / Boston Maine / New Hampshire 8

19 On July 1, 1997 the Independent System Operator New England (ISO-NE) was established as a private, not-for-profit organization by transferring staff and equipment from NEPOOL. This was done in response to Orders 888 and 889 issued by the Federal Energy Regulatory Commission (FERC) in 1996 requiring open access to transmission systems in order to allow a competitive wholesale electricity market. For further information, a history of the restructuring of US electricity markets is given in Appendix A. Since May 1, 1999, the ISO administers, as required by FERC s Order 888, the NEPOOL Open Access Transmission Tariff [11] according to the Interim Independent System Operator Agreement with NEPOOL [12]. ISO-NE operates and controls the region s power grid, provides operational planning services and administers the wholesale electricity markets. The fact that ISO-NE both controls the transmission grid and decides about transactions in the market makes it a very powerful entity. ISO-NE operates the following six markets [13]: real-time energy market reserve markets: o ten-minute spinning reserve o ten-minute non-spinning reserve o thirty-minute operating reserve automatic generation control installed capability The market for operable capability was eliminated in March,

20 The ISO-NE is at the moment developing a new market design and planning to join in the formation of a regional transmission organization (RTO) in the northeast, that would encompass the New York ISO, the ISO-NE and PJM. The market for energy in New England is a residual market. Market participants, who generate power in excess of the demand of their customers, may sell it in the wholesale market to other participants. Participants, who do not generate sufficient power to serve the demand of their customers, may buy it in the market. In this way, only the residual amounts are traded in the market. To give a feeling of the scope of this market, the percentage of energy traded in this market rose from 16% of the total energy usage in New England in to 24% in [14]. 2.2 The transmission-unconstrained market The New England energy market is a single real-time market operated typically every five minutes. Other regions as, for instance, the PJM (Pennsylvania, Jersey, Maryland and Delaware) region and California trade energy in more than one single market. PJM administers two distinct day-ahead and real-time markets. The time line in Fig. 2.2 is used to explain the sequence of operations in the energy market performed by the ISO- NE. Fig. 2.3 shows the information flow in the market. In this section we discuss the transmission-unconstrained market. 10

21 hour 12:00 publish forecast schedule 00:01 24:00 month end generator offers and bilateral contracts day-ahead scheduling update forecast schedule real-time dispatch monthly settlement statement day ahead schedule day up to a month Figure 2.2: Time line of ISO-NE energy market operations bilateral transactions generator offers updated system status ISO-NE dispatching real-time dispatch, energy clearing prices Figure 2.3: Information flow in the ISO-NE energy market 11

22 According to ISO-NE Market Rules and Procedures [13], all generator offers 1 and bilateral transactions information for the next day have to be submitted by the trading deadline, which is 12 noon. An offer is defined as all the information related to price, quantity, technical parameters and timing of offers to provide special services. An offer must contain prices for the range of output from 0 MW to a certain upper limit 2. The output may be provided in more than one segment (block), each with its own distinct price. Since NEPOOL is a mandatory pool, each generator that has an available capacity of more than 5 MW is required to submit an offer. After the day-ahead trading deadline, the ISO determines and publishes the forecast schedule based on its load forecast, reserve requirements, offer and bilateral contracts information and generator and transmission constraints. Generators must submit redeclarations of their offers any time an offer variable or parameter changes due to changing generator status. A redeclaration may contain changes related to the physical capabilities of a unit, but may not change any prices past the trading deadline. The ISO- NE performs at least five forecast schedule updates throughout the day based on actual system conditions and generator redeclarations. The schedule day is the 24-hour dispatch period starting at midnight. The ISO reviews the forecast schedule and updated system information hourly during the day. Dispatch points for generators are typically calculated once every five minutes, taking 1 Note that the ISO-NE in its publications and its web site [10] is incorrectly using the term bid for what is in fact a generator s offer to supply power at a specified price. In this thesis, wherever the term bid is used, it is done in compliance with ISO-NE terminology; the reader should keep in mind that generators are on the supply-side of the market and they submit offers to sell their power. In general, we try to use the term offer. 2 The upper limit is the higher of the high operating limit or the claimed capability; the definition of these variables are given in NEPOOL Market Rules and Procedures [13]. 12

23 into account the following system conditions: the NEPOOL Area Control Error (ACE) 3, forecast demand, reserve requirements, offer parameter redeclarations, up-to-the-minute generator outputs and transmission constraints. Generators must obey the ISO s dispatch instructions and are required to immediately notify the ISO if certain requests cannot be implemented. Hourly energy prices are determined as the average of the real time marginal prices of the five-minute dispatch intervals. In the following we give a basic example of how energy prices are determined in the transmission-unconstrained market. In Example 1 we consider one hour of dispatch. The offered quantities are given in terms of energy per hour, denoted by MWh/h or, for short, MW. The set of offers is given in Table 2.2. To minimize total production costs, we consider lowest price offers first. Dispatching units in an ascending offer order is called merit order dispatch. In our example, the lowest offer is A, offering to serve 30 MW at a price of 5 $/MWh. The next higher offer is B over 50 MW at 10 $/MWh, then C over 25 MW at 15 $/MWh, D over 20 MW at 20 $/MWh and finally E over 55 MW at 30 MWh. Fig. 2.4 shows the aggregated supply curve of offers A to E. Table 2.2: Offer set for Example 1 offer quantity (MW) price ($/MWh) A 30 5 B C D E The area control error is the instantaneous difference between net actual and scheduled interchange, taking into account the effects of frequency bias including a correction for meter error [13]. 13

24 $/MWh B A 10 5 E 30 D C MWh/h Figure 2.4: The supply curve constructed from the offers Consider a fixed system load of 100 MW. The least-price dispatch for this load consists of the dispatch offers A and B and 20 MW of offer C to meet the demand. As seen from the supply curve in Fig. 2.4, this dispatch corresponds to the merit order dispatch constructed from the offers. We say that offers A, B and C are dispatched in merit order. The market clearing price is determined by the equilibrium point at which supply equals demand. Therefore suppliers willing to sell at or below this price are able to sell all of their offered quantity to the bidders who are willing to buy at or above this price. All quantities bought and sold are priced at the uniform market clearing price. The corresponding quantity sold and bought is called the market clearing quantity. The interpretation of the market clearing price is the marginal price of energy in the market. The ISO-NE defines the energy clearing price (ECP) to be the market clearing price established with all offers in the merit order. Therefore the ECP is defined as the price of 14

25 one additional MW 4, which may be dispatched in merit order, should the load be increased by one MW. In our example offer C is dispatched at 20 MW, but offered to serve 25 MW. Therefore the next MW of generation would be supplied by offer C, determining the ECP to be C s offer price of 15 $/MWh. We illustrate this in Fig $/MWh ECP B A D C E 15 MWh/h Figure 2.5: Determination of the ECP for Example 1 Costs are proprietary information and each generator may freely choose its offer price. Due to this fact the price for energy is affected by the bidding strategies of the market participants. In the following we give a basic example of how strategic bidding behavior can have a significant impact on the energy clearing price. We assume two sets 4 The ECP is the time weighted average of the real-time marginal prices during an hour. NEPOOL Market Rules and Procedures Section [13] defines the real-time marginal price under normal system conditions as the next least expensive MW that may be dispatched above the Desired Dispatch Points. Under capacity deficiency conditions the Real-time Marginal Price (RTMP) is the price of the last most expensive MW Under Excess Generation conditions, the RTMP is the price of the last least expensive MW In this thesis, we assume normal system conditions and assume the real-time marginal price to be the price of an additional MW. 15

26 of offers for Example 2, Set 1 and Set 2, given in Table 2.3. We display the supply curve resulting from offer Set 1 in Fig. 2.6, starting with A offering to supply 30 MW at a price offer Table 2.3: The two offer sets in Example 2 SET 1 SET 2 quantity (MW) price ($/MWh) offer quantity (MW) price ($/MWh) A 30 5 A 30 5 B B C C D D of 5 $/MWh, B offering 50 MW at 10 $/MWh, C offering 25 MW at 15 $/MWh and a small offer D over 10 MW at 30 $/MWh. With a system load of again 100 MW, offers A, B and C will be considered. We notice that offer D will not be needed to serve the system load. offers B and D come from the same company $/MWh D 30 ECP 1 A 5 B 10 C MWh/h Figure 2.6: Determination of the ECP for Example 2, Set 1 16

27 Now we assume that offers B and D come from the same company. The offers B and D may be outputs of two different generating units or two different blocks of the same unit. In Set 2 we consider the case, where in trying to maximize their profits, the company engages in strategic bidding, declaring only 40 MW in offer B, instead of 50 MW. This reduction results in the need for an additional 10 MW to serve the system load. To accomplish this goal, offer D will also be considered. Fig. 2.7 shows the new situation and the resulting new energy clearing price. The new energy price being paid to all units dispatched in merit order is the offer price of the next MW that would be dispatched in merit order, were the load raised by one MW. D offered to sell 10 MW and is currently dispatched at 5 MW. At an increase of load, D would supply the next MW. Therefore, offer D determines the energy clearing price and its value changes to 30 $/MWh. ECP 2 $/MWh D price of offer D determines the energy clearing price A 10 5 C B ECP 1 : 15 MWh/h Figure 2.7: Determination of the ECP for Example 2, Set 2 The strategic bidding behavior of offer B, in which 10 MW of capacity is withheld, doubles the system-wide price for energy from 15 $/MWh in the base case to 30 $/MWh. 17

28 This example is useful in illustrating the great influence a particular offer has on shaping the ECP. Note that with a small block of generation, offer D can establish the uniform price paid to all the suppliers. 18

29 3. CONGESTION This chapter provides a workable definition of congestion, which we use to quantify its impacts. We describe the transmission-constrained market in New England and the metrics used by the ISO-NE to measure congestion. In addition, the market power mitigation procedure used by the ISO-NE in connection with the occurrence of congestion is explained. 3.1 A workable definition So far we discussed the dispatch of generation and the determination of energy prices in the transmission-unconstrained market. Due to various considerations with many primarily of a reliability nature it is not always possible to use the dispatch determined in the transmission-unconstrained market. In fact, the requirement to explicitly consider certain constraints leads to the redispatch of generation. Congestion occurs whenever one or more constraints are violated under which the system operates in the normal state or in any of the contingency cases in a list of specified contingencies. These constraints can either be physical limits like thermal or voltage limits or specified limits to ensure system security and reliability [1,2]. We illustrate the basic notions of congestion with respect to the simple 2-bus system of Fig.3.1 in Example 3a. The system has a seller at bus 1 and a seller at bus 2. A buyer intending to buy 150 MW is located at bus 2. The seller at bus 1 offers to sell power at a price of 5 $/MWh and the seller at bus 2 offers power for 10 $/MWh. 19

30 no line flow limit bus 1 bus 2 Seller 1 ~ Seller 2 ~ 5 $/MWh 10 $/MWh 150 MW 0 MW Buyer 150 MW Figure 3.1: Unconstrained 2-bus system of Example 3a The transactions are determined as in Examples 1 and 2 for the transmissionunconstrained market (Section 2.2). Seller 2 offers more expensive power than seller 1 and will therefore not be dispatched. Seller 1 sells 150 MW to the buyer at bus 2. Thus the total costs per hour are $750. Now, we assume a more realistic situation in Example 3b where we impose a limit on the power transfer on the line between bus 1 and bus 2. Suppose the limit is 100 MW as illustrated in Fig The optimal dispatch point to minimize total costs is still the same as in the previous example: the seller at bus 1 at 150 MW and the seller at bus 2 not dispatched. But in this case a transaction of 150 MW between the seller at bus 1 and the buyer at bus 2 is not feasible, since it would result in a line overload of 50 MW. limit of 100 MW bus 1 bus 2 Seller 1 ~ Seller 2 ~ 5 $/MWh 10 $/MWh 100 MW 50 MW Buyer Figure 3.2: Constrained 2-bus system of Example 3b 150 MW 20

31 To eliminate this overload, we reduce the sale from seller 1 by 50 MW and instead dispatch the higher priced power of seller 2. With this new dispatch, total costs amount to $1,000. The additional constraint on the transmission line led to congestion and an increase in system costs by 33%. We can basically measure congestion in two different metrics. Congestion is, on the one hand, the difference in megawatts of the power scheduled to flow on a transmission line and the actual transfer which is allowed on the line without violating any constraints. On the other hand, we can determine congestion costs as the difference in the costs for securing power to serve the system load without having to consider any constraints and the costs to serve the load without violating existing limits. In the 2-bus system in Example 3b, congestion in the amount of 50 MW 5 occurs, resulting in congestion costs of $250. Transactions in a 2-bus system like the one we illustrated in Example 3 are straightforward. However, real systems are much more complex and require a higher effort to analyze. To go only one step further from the previous example, we want to illustrate the case of a power system with five buses, five sellers and three buyers. Fig. 3.3 shows the network for the 5-bus system of Example 4a and Table 3.1 gives the quantities requested by the buyers and the prices at which the sellers are willing to sell. We assume a lossless system. The buyer with the highest demand is at bus 5, intending to buy 450 MW. The seller offering for the highest price, 40 $/MWh, is also located at bus 5. Seller 4 has the lowest price offer with 5 $/MWh. The following scenarios are simulated with PowerWorld software using its Economic Dispatch and OPF functions [15]. 5 As mentioned in Chapter 2, congestion is measured in units of energy; since we focus on one hour of dispatch, we use MWh/h, or, for short, MW. 21

32 Figure 3.3: Unconstrained 5-bus system of Example 4a Table 3.1: Offers and demand quantities for Example 4 bus quantity (MW) seller s offer requested by buyer price ($/MWh) In the first case of Example 4 we consider the transmission-unconstrained system and again determine the optimal dispatch as in Examples 1 and 2 of Section 2.2. The result is 22

33 straightforward. Only the generator that offers to sell power for the lowest price will be dispatched (assuming a high enough output limit). In this case, all transactions will take place between the buyers and the seller at bus 4. The seller at bus 4 will sell in total 850 MW, resulting in hourly total system costs of $4,250. The values on the lines in Fig. 3.3 show the respective line loading. To accommodate the transactions between seller 4 and buyers 3 and 5, the transmission lines connecting bus 4 with buses 3 and 5 experience a higher power flow than the other lines, carrying 395 MW and 455 MW, respectively, as compared to 5 MW, 32 MW and 64 MW on the remaining lines. Next, we consider the line flow limits of 100 MW on each line of the network for Example 4b. We use the optimal power flow function of the PowerWorld software, which minimizes total system costs subject to transmission constraints. We expect the power flow of the system to change, because the flows on two lines in the transmissionunconstrained solution violated the line limit of 100 MW. The results of the optimal power flow for the 5-bus system of Example 4b are shown in Fig The pie charts show the percentage of line flow on each line according to its limit. As we can see, each line except that between buses 1 and 2 has flows exactly at the limiting values. The dispatch differs significantly from the transmission-unconstrained case in Example 4a, since seller 4 cannot be the only one selling its power, but also sellers with higher offer prices have to be considered. The new costs per hour are $19,000, which is more than four times those of the case without transmission constraints. 23

34 Figure 3.4: Constrained 5-bus system of Example 4b We see that the presence of congestion can substantially change the dispatch within the system, forcing inefficient and expensive units to run. This results in significantly higher costs of securing the power needed to serve the system demand. Assume we try to purchase power from one or more sellers with the lowest offer prices to minimize our total costs of serving the system demand. If sufficient transmission facilities connecting the sellers with the buyers do not exist, line limits will be violated, if we keep the intended dispatch. If, on the other hand, there were sellers who offer power at low prices located close to the buyers, we would not need high transmission capacity to serve the system demand. Therefore, the problem of congestion lies in insufficient 24

35 transfer capability of the transmission system. We can see that generation and transmission are substitutable. Congestion is caused by the simultaneous lack of sufficient transmission capacity and economical generation. The problem of controlling operations at the lowest cost so that the system remains secure (no limits are violated), was present before restructuring. Why is congestion more significant now in the restructured system? Before restructuring, utilities were vertically integrated, owning and controlling both generation and transmission [1]. A utility minimized costs by optimally dispatching generation, knowing all the constraints in the transmission system. Conflicts between security and economics could be solved by one decision-making entity. The method used for the solution of the problem is the optimal power flow tool 6. It resulted in a first-best solution by maximizing the social surplus or minimizing the total production costs [2]. In a competitive market, however, where transmission and generation are unbundled, many market players offer to sell power and make use of the transmission grid. With a growing number of independent competitors, it is a challenge for the transmission system operator to maintain system reliability while minimizing costs. To ensure system reliability without being able to rely on a single entity with control over all facilities, efficient rules must be established to coordinate the actions of all market participants. Congestion management is concerned with approaches for dealing with this problem. Various schemes have been adopted by the system operators in the U.S. regional systems and in other jurisdictions [2]. 6 An explanation of the optimal power flow method can be found in any standard textbook, e.g. see Wood and Wollenberg [16]. 25

36 3.2 The Congestion Metrics used by the ISO-NE To understand how ISO-NE measures congestion, we have to first explain the concept of out-of-merit-order dispatch. We started the explanation with the merit order dispatch under which the system dispatches generation blocks in the order of ascending offer prices in Section 2.2. Due to reasons of reliability requirements, it is not always possible to maintain such dispatch. To avoid any violations of the network constraints either in the normal state or in any of the contingency cases in the list of specified contingencies, the ISO must redispatch the outputs of units. Since this necessitates the use of some of the offers of higher priced units to serve load and overcome the limitations due to network constraints, such an output pattern is termed out-of-merit-order dispatch. The energy clearing price is the offer price of the generating block that would serve an additional MWh of energy when merit order dispatch is used. The prices of the generating blocks that are dispatched out of merit order so as to relieve congestion, cannot contribute toward the determination of the energy clearing price. All blocks with offer prices above the ECP are paid the price for which they offered to sell. We illustrate the effect of out-of-merit-order dispatch in Example 5. Consider the system of Fig. 2.4 with the same set of offers as in Example 1 in Section 2.2 (Table 2.2). The difference is that now offer B cannot be considered due to transmission constraints not allowing B s power to be transported over the lines connecting it to the load. Assume we need to dispatch offer E to relieve the congestion and serve the system load. The new dispatch is shown in Fig

37 $/MWh E 30 ECP C A MWh/h Figure 3.5: The supply curve with out-of-merit-order dispatch incorporating offer E It is important to distinguish between the in-merit status of offer A and C and the outof-merit status of offer E. Since E is only dispatched so as to prevent violation of the power transfer limit, its price cannot be used in the determination of the energy clearing price. The ECP is still at 15 $/MWh, as it was in the transmission-unconstrained example. An interesting side effect of the determination of the ECP is that in certain cases the transmission-constrained ECP may be lower than the unconstrained ECP. In Example 6 illustrated in Fig. 3.6, consider the case that the power transfer limit constrains the ISO to $/MWh E 30 ECP A 5 B MWh/h Figure 3.6: The supply curve resulting in a lower ECP with transmission constraints considered 27

38 accept only half the block of offer B and no part of C. Then, as before, offer E is dispatched out of merit order. The transmission-constrained ECP is determined by the price of offer B to serve an additional MWh of demand using in merit order dispatch. The ECP is reduced from 15 $/MWh to only 10 $/MWh. The ISO-NE uses two congestion measures to quantify congestion. We illustrate the use of the two measures with respect to Example 5 illustrated in Fig One measure of congestion is the measure of the energy/hour (MWh/h) that must be redispatched from higher priced blocks so as not to violate transmission constraints. It is customary to express this quantity in MW rather than MWh/h. The New England system defines congestion as the amount of energy supplied by the out-of-merit-order blocks used in order to ensure no violation of transmission constraints. In Example 5 illustrated in Fig. 3.5, there is consequently congestion of 50 MW, which is the dispatched capacity of the out-of-merit-order block of offer E. The second measure is congestion costs. These are defined as the costs of the energy produced by blocks dispatched out of merit order. These costs arise due to the higher prices for the energy generated by the out-of-merit order blocks; these blocks are paid the offer prices. The ISO-NE refers to these costs as congestion uplift costs. The formula for the total costs is given by uplift costs = [( offer priceb- ECP )*κ b] b where κ denotes the capacity of block b used in out-of-merit-order dispatch and the summation is over all the out-of-merit blocks b in the redispatch [17]. 28

39 In Example 5, the offer price of the generation dispatched out of merit order is 30 $/MWh and the ECP is 15 $/MWh. With 50 MW of congestion, this results in congestion costs of $750 (Fig. 3.7). 50 MW $/MWh offer price : 30 E C congestion costs [(30-15)*50] = 750 A ECP: 15 MWh/h Figure 3.7: Example of ISO-NE calculation of congestion costs Section 24 of the NEPOOL Open Access Transmission Tariff [11] requires that the total congestion costs be allocated to all market participants in proportion to their share of the network load. This socialization of costs is controversial, since all consumers are billed for congestion, even those who are not located in the congested area. 3.3 The ISO-NE Market Power Mitigation Procedures The main purpose of restructuring the electricity system into a competitive market is to increase benefits to consumers. But replacing the regulated rates for electricity by competitive prices also raises the problem of market power. Definitions of market power 29

40 can be found in the literature [6,18,19]. The exercise of market power denotes the ability of a supplier to charge prices above competitive market levels for sustained periods. This is precisely the opposite of perfect competition where each participant is a price taker. An entity with market power is a price setter. In general, a high concentration of the ownership of resources can facilitate the exercise of market power. However, the transmission network plays a vital role in the functioning of competitive markets. In fact, it is possible for market participants to exercise market power without a dominant position of market concentration, but solely because of the occurrence of congestion [20]. If an area is isolated from low priced energy supply due to congestion, local suppliers may have the potential to exploit this situation for their own benefits. Any exercise of market power interferes with the competitiveness of markets and can thereby significantly reduce the benefits to consumers that are the goal of introducing competition to electricity markets. Since the beginning of operations in 1999, the ISO-NE practiced market power mitigation by monitoring the offers of generating units that are necessary to relieve congestion and may therefore have the potential to exploit this situation and exercise market power by significantly raising their offer prices. Whenever the ISO-NE assumed that a supplier exercised market power, the respective generating unit did not receive the full offer price for its energy, but a reduced price determined by the ISO-NE. These actions were taken to foster competition by monitoring and, if necessary, modifying any behavior that is deemed anti-competitive or otherwise interfering with efficient market operations. 30

41 FERC found that the existing market rules allowed the ISO-NE too much discretion in determining when market mitigation will be imposed and required the ISO-NE to disclose clear criteria for market power mitigation actions. The ISO-NE then proposed the Revised Market Rule 17 of NEPOOL Market Rules and Procedures [13], which is in effect since May, We ignore the mitigation procedures undertaken before this date and we focus on the Revised Market Rule. It provides the ISO with a clear framework used for detecting and mitigating the exercise of market power. The two basic ways of exercising market power are physical and economic withholding. Physical withholding is the act of falsely declaring resources as unavailable or submitting operating parameters, which will prevent the block from being dispatched, although absent such false declarations the block would be dispatched in economic merit order. A prime example is the scheduling of maintenance: sellers can exercise market power by strategically planning outages [7]. The ISO-NE has promulgated regulations that require the sellers to provide certifiable reasons for generation outages and that require that current and historical data be compared to detect and analyze changes in the resource s availability. Economic withholding occurs, if a resource s offer price is deliberately raised to such a high level as to effectively remove it from the merit order dispatch. The interaction of offers with the consideration of transmission constraints also provides market participants with the opportunity to exercise market power. Consider a so called load pocket a region which is connected to inexpensive generation by transmission lines, but where one or more local generating units are physically required to serve the load in this area. Since the sellers of the area s generation know that their offers are dispatched irrespective of their prices, nothing prevents them from submitting 31

42 high-priced offers to maximize their profits. Consider again the five-bus example of Fig The seller at bus 5 can raise its offer price to any value, since the transmission constraints allow a limited amount of power imports and this amount is less than the load requirements of the bus. According to the regulation in Market Rule 17 of the NEPOOL Market Rules and Procedures [13] each out-of-merit-order block be subject to two market power screens -- a structural and a price screen. The ISO-NE must first apply a structural screen to determine if all requirements could have been met without the block in question. If this is the case, the blocks of alternative resources are identified. The rules require the identification of three or more independently controlled competing bidders that would satisfy the transmission constraints. The ISO may raise this number to five bidders, if the constraint is reasonably foreseeable 7. This screen therefore ensures that the block has passed to receive its offer price. If the block is unable to pass this screen, an additional price screen must be applied by the ISO-NE. This second screen distinguishes between blocks of resources that are generally dispatched in merit order and those that seldom are in merit order. In either case the offer price of the block is compared with a screen price, which is a percentage of a reference price whose level depends on the number of hours the block was operated out of merit order during the past 90 days. The reference price is the weighted average of the in-merit offers during the most recent 30 days. If the block s offer price lies below the screen price, it may receive its offer price. Otherwise, the offer price must be replaced by the particular screen price, or under certain cases by a price negotiated by the offerer and the ISO. The costs resulting from the prices actually paid to 7 NEPOOL Market Rules and Procedures, Section c. It is not explained in any greater detail as to under which circumstances a constraint would be classified as being reasonably foreseeable, leaving room for disputes to arise. 32

43 the out of merit order blocks are the mitigated congestion costs calculated with the formula given by uplift costs = [( offer price b- ECP )*κ b] b where offer price denotes the price actually paid to the out of merit order blocks. Note that we call the costs resulting from the total mitigated uplift costs in one hour divided by the total amount of congestion in the respective hour the per MWh mitigated costs of congestion; the formula is given by uplift costs = b [( offer price - ECP )* κ ] During some months the mitigation measures reduced congestion payments after the fact by 50% of the initial costs. Bushnell and Saravia claim that the competitiveness of the New England markets was definitely improved by the mitigation procedures of ISO- NE, as can be found from comparison with other markets with less intense market power mitigation actions [7]. A summary of the ISO-NE mitigation procedures and screening processes is provided in Fig We see that the ISO-NE is aiming at mitigating possible market power by applying the described mitigation procedure. However, this does not mean, that necessarily each entity whose offers are mitigated by the ISO-NE has really been engaged in exercising market power. Likewise there is no guarantee that the ISO-NE screening processes detect all cases of exercise of market power by market players. b b κ b b 33

44 identify out-of-merit order blocks define the particular constraint identify alternative blocks STRUCTURAL SCREEN TEST Could the requirement have been met without running the selected resource? no yes resource will receive its offer price yes Were there 3 or more independently controlled competing bidders that could satisfy the requirement? no yes PRICE SCREEN TEST Is the offer price of the resource smaller or equal the screen price? no resource receives either applicable screen price or price negotiated with ISO-NE Figure 3.8: ISO-NE market power mitigation procedure 34

45 4. CONGESTION DATA ANALYSIS In the previous chapters we have described the New England power system and the key roles of the independent system operator, the ISO-NE. We also discussed the market operations in the transmission-unconstrained market and the nature and measure of congestion. We next focus on the study of the congestion data in the New England system. To start with, we describe the available data that we use in our study. Using the two metrics discussed in Chapter 3, we quantify the congestion. Our analysis consists of the detection of monthly and hourly patterns of congestion quantities and costs. We examine the distribution of congestion across the twelve transmission areas of New England and the impacts of the ISO-NE s market power mitigation on the congestion costs. Our investigation of the relationship between generation availability and congestion is also discussed. 4.1 Data Description All the data used in this study are taken from congestion reports published by ISO-NE [21]. Each report consists of a spreadsheet that provides the following data: total MW 8 of congestion; total, marginal and per MWh congestion costs; and, the type of the constraint. The data are available in daily and hourly forms. Each daily value is the sum of the 24 hourly values and the computation is made for the total congestion and the total 8 The footnote 5 of Chapter 3 applies to all the discussion and plots of congestion in this chapter. 35

46 congestion costs. In addition, the daily average values of the congestion costs per MWh are given. The values of the unmitigated costs and the mitigated costs are given. The information about the type of congestion is very generic and not closer specified and therefore it does not provide satisfactory details on the reasons of particular congestion occurrences. The congestion reports are available for the twelve separate transmission areas of the New England system presented in Table 2.1. For three of these areas -- MAINE, NEMASS/BOST and ME/NH -- only daily but no hourly data are reported. All other areas have both daily and hourly data available. All the data available at the time of this study from the beginning of the operation of the ISO-NE in May, 1999, to June, 2001, were analyzed. A sample congestion report can be found in Appendix B. 4.2 Analysis of monthly and hourly patterns In this section we analyze data on monthly and daily congestion quantities, monthly and daily mitigated costs and monthly and daily per MWh mitigated congestion costs. The New England Power System experienced 9.3 million megawatts of accumulated congestion in the 26 months from the beginning of the restructured markets in May, 1999, until June, We used the daily congestion reports to compute the monthly congestion metrics over the period of the analysis. We give the monthly congestion quantities in MW in the chart of Fig

47 May-99 Jul-99 Sep-99 Nov-99 Jan-00 monthly congestion in 1,000 MW Mar-00 May-00 Jul-00 Sep-00 Nov-00 Jan-01 Mar-01 May-01 Figure 4.1: Monthly congestion for May 1999 June 2001 The analysis of the plot allows us to make the following findings: the monthly average is 360,000 MW of congestion and the range is from 19,000 MW to 770,000 MW; the highest values were attained in March, April and May, 2000 (620,000 MW, 720,000 MW and 770,000 MW, respectively), some lower high values occurred in December, 1999, June, 2000, August, 2000, and June, 2001 (above 500,000 MW); during the first four months of operation (May to August, 1999) monthly congestion stayed below 100,000 MW, being as low as 19,000 MW in July; these small values may be indicative of the fact that trading in the new markets was in its incipient stages and participants still had to become accustomed to the competitive environment. 37

48 We analyze the changes in congestion in the 26 months of the observed period in the graphs in Fig. 4.2 and 4.3. Fig. 4.2 shows the monthly congestion in the first 12 months from May, 1999, to April, 2000, and Fig. 4.3 shows the values for the following 14 months from May, 2000, to June, May-99 Jun-99 Jul-99 Aug-99 Sep-99 Oct-99 Nov-99 Dec-99 Jan-00 Feb-00 Mar-00 monthly congestion in 1,000 MW Apr-00 Figure 4.2: The overall upward trend of the monthly congestion for May April 2000 During the first year of operation the monthly congestion amounts were growing, as can be seen in Fig Fig. 4.3 shows that during the second year, which started out with the peak month of the 26-month period (May, 2000), the monthly values were decreasing. The last two of the observed months, May and June, 2001, experienced an increase in monthly congestion again, but data for the following months would be necessary to detect further developments. 38

49 May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 monthly congestion in 1,000 MW Jan-01 Feb-01 Mar-01 Apr-01 May-01 Jun-01 Figure 4.3: The overall downward trend of the monthly congestion for May 2000 June 2001 The peak months of congestion occurred in winter, spring and summer. To compare the seasonal behavior in each year of the 26-month period, we show the MW quantities of congestion in each month of the observed period in Fig The winter months December, January and February experienced similar amounts of congestion in each year. This can be seen from the closeness of the lines showing the congestion amounts during the different years. The months of spring, March, April and May, however, show high congestion with more than 600,000 MW each month in 2000, but relatively lower congestion with less than 400,000 MW in The summer months June, July and August also experienced very different amounts of congestion in each year. In 1999 congestion during the summer was very low, ranging below 100,000 MW. In 2000, congestion in summer was not especially low with values between 450,000 MW 39

50 and 560,000 MW each month, but summer was not the peak period of congestion during that year. Figure 4.4: Seasonal behavior of monthly congestion for May 1999 June Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec monthly congestion in 1,000 MW From this analysis of the seasonal behavior of the congestion quantities during the observed 26 months we can find no indication for a concentration of congestion on a particular season of the year. Next we examine the daily data. We illustrate the daily congestion amounts together with the monthly averages in Fig The study of the daily congestion values leads to the following findings: the average daily amount of congestion is 11,770 MW and the range is from 0 MW to 51,799 MW; the highest value was attained on June 27, 2001, with 51,799 MW, some lower high values (each above 30,000 MW) occurred during December 8, 1999, 40

51 March 13, April 27 and 28, May 1, 2, 4, 10, 11 and 16, 2000, August 9, 2000 and June 28, 2001; during the first four months of operation (May to August, 1999) daily congestion stayed below 10,000 MW with the exception of May 2, 1999, when 10,871 MW of congestion occurred; three days experienced 0 MW of congestion; these small values may be indicative of the incipient stages of the markets as described for the monthly values above. congestion amounts in 1,000 MW May Jul Sep Nov Jan Mar May Jul Sep Nov Jan Mar May-2001 daily total amount monthly average congestion Figure 4.5: Daily and monthly average congestion for May 1999 June

52 In order to examine how the daily congestion quantities account for the monthly amounts, we analyze the frequency of the daily congestion values for the 26 months of the studied period and separately for the three highest congested months, March, April and May, Such a discussion requires the introduction of a new curve. We define the congestion duration curve to be the rearranged plot of the congestion amount from the highest to the lowest value. Such a curve has a nice interpretation in terms of probability distribution if the ordinate is interpreted as a percentage of the total period under study. 60 daily congestion in 1,000 MW A 0 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in the May June 2001 period Figure 4.6: Congestion duration curve for May 1999 June 2001 For example, we rearrange the daily congestion plotted in Fig. 4.5 as the congestion duration curve for the 26-month duration, shown in Fig The 100% on the x-axis means the entire 26-month period. The point A on the curve indicates that daily congestion was at or above the 20,000 MW level about 15% of the days of the 26-month period. 42

53 The flatness over most of the congestion duration curve indicates a rather limited variability: for over 80% of the days in the 26-month period the congestion is below 20,000 MW. For the remaining 15% there is a considerable volatility making the congestion vary between 20,000 and 50,000 MW. In Table 4.1 we divided the daily congestion quantities into sections of 10,000 MW. We provide the percentage of days that fall under each section within the given period of time. Table 4.1: Frequency information for congestion over the May 1999 June 2001 period percent of days with congestion period below 10,000 MW between 10,000 and 20,000 MW between 20,000 and 30,000 MW above 30,000 MW May 1999 June March April May We also studied the volatility of daily congestion quantities in the three highest congested months. Fig. 4.7 to 4.9 show the duration curves of the daily congestion amounts during March, April and May, For these months, the curves also exhibit flatness over most of the period: for nearly 70% of the days in May, 2000, and for over 80 % of the days in March and April, 2000, the congestion is between 10,000 and 30,000 MW. We can further explain the high accumulated amounts of congestion during these months by comparing the frequency of daily congestion values in each month with that during the entire observed period. While 48% of the days in the time between May, 1999, 43

54 60 daily congestion in 1,000 MW % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in March 2000 Figure 4.7: Congestion duration curve for March daily congestion in 1,000 MW % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in April 2000 Figure 4.8: Congestion duration curve for April

55 60 daily congestion in 1,000 MW % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in May 2000 Figure 4.9: Congestion duration curve for May 2000 to June, 2001, experienced less than 10,000 MW of daily congestion, no day during March and April, 2000, and only 6% of the days in May, 2000, experienced an equally low amount of congestion. The percentage of days that had more than 20,000 MW of daily congestion is substantially higher during the three highly congested months than during the observed 26 months, as can be seen in Table 4.1. Especially the percentage of days that experienced daily congestion of more than 30,000 MW is substantially higher in these months, ranging from 6% in March, 2000, to 26% in May, 2000, as opposed to only 3% in the entire period of study. We next turn our attention to the study of congestion costs. Total mitigated congestion costs accumulated to $233 million during the observed 26-month period from 45

56 May, 1999, to June, We give the monthly mitigated congestion costs in the chart of Fig The analysis of the plot allows us to make the following findings: the monthly average is $9 million of mitigated congestion costs and the range is from $360,000 to $21.9 million; the highest values were attained in March, April and May, 2000 ($21.9 million, $19.8 million and $20.3 million, respectively), some lower high values occurred in December, 1999, June, 2000, December, 2001, and June, 2001 ($11 to $14 million); monthly mitigated congestion costs in million $ May-99 Jul-99 Sep-99 Nov-99 Jan-00 Mar-00 May-00 Jul-00 Sep-00 Nov-00 Jan-01 Mar-01 May-01 Figure 4.10: Monthly mitigated congestion costs from May 1999 June 2001 during the first four months of operation (May to August, 1999) monthly mitigated congestion costs stayed below $1.1 million, being as low as $360,000 in 46

57 July; these small values may be indicative of the same fact as described for the monthly congestion quantities above. The mitigated congestion costs show roughly the same changes over the studied 26- month period and seasonal behavior as the amount of congestion in MW that we investigated in the previous section. This can be seen from Fig. 4.11, which plots monthly mitigated congestion costs and monthly congestion quantities on the same chart. The analysis of the mitigated congestion costs allows us to make equivalent findings as from the analysis of the monthly amounts of congestion. mitigated costs in million $ per month May-99 Jul-99 Sep-99 Nov-99 Jan-00 Mar-00 May-00 Jul-00 Sep-00 Nov-00 Jan-01 Mar-01 May monthly congestion in 1,000 MW monthly mitigated costs monthly congestion Figure 4.11: Monthly congestion and mitigated congestion costs for May 1999 June

58 During the first year of operation the monthly mitigated congestion costs were growing, then decreasing in the second year. The last two of the observed months experienced increasing costs, but no assumptions about further development of the costs can be made from the available data. Note that the month that experienced the highest mitigated congestion costs is March, 2000, while the month with the highest congestion quantity is May, We discuss this observation in the following section on daily mitigated congestion costs. Similar to the monthly congestion quantities, we can find no indication for a concentration of high monthly costs on a particular season of the year. In the next section we examine the daily data. We illustrate the daily mitigated congestion costs together with the monthly averages for the studied period in Fig The study of the daily mitigated congestion costs leads to the following findings: the average daily mitigated congestion costs are $300,500 and the range is from $0 to $1.6 million; the highest value was attained on March 13, 2000, with $1.6 million, some lower high values occurred during June 27, 2001 ($1.5 million), February 28, 2000 ($1.4 million) and March 6 to 10 and 14, 2000 (each between $1.0 and $1.4 million); during the first four months of operation (May to August, 1999) daily mitigated congestion costs stayed below $0.2 million, three days experiencing $0 of congestion costs; these small values may be indicative of the same fact as described above. 48

59 May-99 1-Jul-99 1-Sep-99 1-Nov-99 1-Jan-00 1-Mar-00 1-May-00 1-Jul-00 1-Sep-00 1-Nov-00 1-Jan-01 1-Mar-01 mitigated congestion costs in million $ 1-May-01 mitigated daily costs monthly average Figure 4.12: Daily mitigated congestion costs and the monthly averages for May 1999 June 2001 We similarly introduce the notion of congestion cost duration curves constructed analogously to the congestion duration curves defined above: With such curves, we can analyze the frequency of the daily values of congestion costs for the 26 months of the studied period and separately for the three highest congested months, March, April and May, For example, the plot in Fig shows the congestion cost duration curve for the period from May, June, We notice a similar flatness over most of the congestion cost duration curve as we did for the congestion amounts indicating a rather low variability: 74% of the days 49

60 experienced costs below $0.4 million. The remaining 26%, however, exhibit high volatility, making the daily congestion costs vary from $0.4 to over $1.6 million. daily mitigated congestion costs in million $ % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in the May 99 - June 01 period Figure 4.13: Mitigated cost duration curve for May 1999 June 2001 Table 4.2: Frequency information for daily mitigated congestion costs over the May 1999 June 2001 period percent of days with mitigated congestion costs period below $0.4 between $0.4 million between $0.8 million above $1.2 million and $0.8 million and $1.2 million million May 1999 June March April May

61 In Table 4.2 we divided the daily mitigated congestion costs into sections of $0.4 million. We provide the percentage of days that fall under each section within the given period of time. In order to examine how the daily mitigated congestion costs during the peak months account for the high monthly costs, we constructed the congestion cost duration curves of the daily mitigated congestion costs for March, April and May, These are shown in Fig to We can explain the high values of monthly costs by comparing the frequency of the daily values of the costs in each month with that during the entire observed period. The numbers for that are also given in Table 4.2. While 74% of the days during the 26 months from May, 1999, to June, 2001, experienced daily mitigated costs below $0.4 million, no day in April, 2000, and less than 20% of the days during March and May, 2000, experienced equally low costs. daily mitigated congestion costs in million $ % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in March 2000 Figure 4.14:Mitigated cost duration curve for March

62 daily mitigated congestion costs in million $ % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in April 2000 Figure 4.15: Mitigated cost duration curve for April 2000 daily mitigated congestion costs in million $ % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in May 2000 Figure 4.16: Mitigated cost duration curve for May

63 The percentage of days that had more than $0.8 million of daily costs is substantially higher during the peak months than during the entire observed period. The fact that March, 2000, experienced the highest monthly mitigated congestion costs, is caused by the high percentage of days with more than $1.2 million, which was 16% as opposed to only 1% in the entire period of study. So far, we have analyzed the monthly and daily data about congestion quantities and associated costs. The total congestion costs are a function of the amount of congestion and the per MWh mitigated congestion costs as described in Section 3.3. We next examine the nature of the monthly and daily per MWh mitigated costs of congestion. We refer to the per MWh monthly mitigated costs of congestion in Figure The analysis of the plot allows us to make the following findings: monthly per MWh mitigated costs of congestion range from $/MWh to $/MWh; the highest values were attained in March, 2000, and January, 2000, with $/MWh and $/MWh, respectively; some lower values occurred in November, 1999, December, 2000, January, 2001, and May, 2001 (32.00 $/MWh to $/MWh). Similar to the monthly congestion quantities and total costs, the monthly per MWh mitigated costs were growing in the first year of operation, from May, 1999, to April, 2000, as can be seen in Fig

64 May-99 Jul-99 Sep-99 Nov-99 Jan-00 Mar-00 May-00 Jul-00 Sep-00 Nov-00 Jan-01 Mar-01 May-01 monthly mitigated costs in $/MWh Figure 4.17: Monthly per MWh mitigated costs for May 1999 June monthly mitigated costs in $/MWh May-99 Jun-99 Jul-99 Aug-99 Sep-99 Oct-99 Nov-99 Dec-99 Jan-00 Feb-00 Mar-00 Apr-00 Figure 4.18: The general upward trend of the monthly per MWh mitigated costs for May 1999 April

65 Fig shows, that in the second year monthly per MWh costs did not decrease like the monthly congestion amounts and total costs did as we have found from our analysis above May-00 Jun-00 Jul-00 Aug-00 Sep-00 Oct-00 Nov-00 Dec-00 Jan-01 Feb-01 Mar-01 Apr-01 May-01 monthly mitigated costs in $/MWh Jun-01 Figure 4.19: The lack of a definite trend in the monthly mitigated costs per MWh for May 2000 June 2001 The seasonal behavior of the monthly per MWh mitigated costs is slightly more distinct than of the monthly congestion quantities and total costs. Fig shows the monthly per MWh costs for each month during the observed 26-month period. We can see that per MWh mitigated congestion costs were less during the summer months June to September than during the rest of the year. May, 1999, and November, 2000, experienced comparably low costs per MWh. 55

66 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec monthly mitigated costs in $/MWh Figure 4.20: Seasonal behavior of the monthly per MWh mitigated costs for May 1999 June 2001 We next analyze the daily per MWh mitigated costs. The chart in Fig plots daily per MWh mitigated congestion costs together with the monthly average values. The study of the data leads to the following findings: the average of the daily per MWh mitigated congestion costs is $/MWh and the range is from 1.49 $/MWh to $/MWh the highest value was attained on October 16, 1999, with $/MWh, some lower values occurred on July 19, 1999 (85.96 $/MWh), August 17,1999 (85.84 $/MWh) and February 28, 2000 (82.96 $/MWh); 56

67 costs below 10 $/MWh occurred nearly exclusively during the months in daily mitigated congestion costs in $/MWh May-99 1-Jul-99 1-Sep-99 1-Nov-99 1-Jan-00 1-Mar-00 1-May-00 1-Jul-00 1-Sep-00 1-Nov-00 1-Jan-01 1-Mar-01 1-May-01 Figure 4.21: Daily per MWh mitigated costs and monthly averages for May 1999 June 2001 The graph in Fig shows a concentration of extraordinarily high daily per MWh mitigated costs during the first half of the studied period. We analyze this as before with the help of duration curves of the daily values. We plot the per MWh mitigated cost duration curve for the entire studied period from May, 1999, to June, 2001, in Fig The only exception is November 16, 2000, with 9.96 $/MWh. 57

68 Note that 60.2% of all days in the 26-month period experienced daily per MWh mitigated congestion costs between 20 $/MWh and 40 $/MWh. Only 9.2% of the days experienced higher values. The percentage values are given in Table 4.3. per MWh mitigated costs in $/MWh % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in the May June 2001 period Figure 4.22: Per MWh mitigated cost duration curve for May 1999 June 2001 Table 4.3: Frequency information for daily per MWh mitigated congestion costs over the May 1999 June 2001 period percent of days with per MWh mitigated congestion costs period below 20 between 20$/MWh between 40 $/MWh above 60 $/MWh and 40 $/MWh and 60 $/MWh $/MWh May 1999 June May 1999 April May 2000 June To visualize any differences in daily per MWh mitigated congestion costs we constructed the per MWh mitigated cost duration curves for the first 12 months in Fig. 58

69 4.23 and for the following 14 months in Fig The percentage per range is given in Table 4.3. We notice that 13.9% of the days during the first year of operation experienced daily per MWh mitigated congestion costs of more than 40 $/MWh, whereas in the following 14 months only 5.2% of the days experienced equally high values. per MWh mitigated costs in $/MWh % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in the May April 2000 period Figure 4.23: Per MWh mitigated cost duration curve for May 1999 April 2000 Especially costs above 60 $/MWh occurred on 3.8% of the days of the first year, but during only 0.5% of the days in the following 14 months. Interestingly, the number of days with congestion costs below 20 $/MWh is also higher during the first year than during the second half of the studied period. This indicates a higher volatility of daily per MWh mitigated congestion costs during the first year of operation than later on. This may be again indicative of the fact, that trading in the new markets was in its incipient stages and participants had to become accustomed to the competitive environment. 59

70 100 per MWh mitigated costs in $/MWh % 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% percentage of days in the May June 2001 period Figure 4.24: Per MWh mitigated cost duration curve for May 2000 June 2001 In this section we have analyzed monthly and daily data on congestion quantities, total mitigated costs and per MWh mitigated costs. We found the months with highest congestion to be March, April and May, We could not detect any definite trend or seasonal pattern in congestion quantities and total costs over the 26-month period. Per MWh mitigated costs were slightly lower during the summer months than during other seasons. We noticed a high volatility in all congestion metrics. We are interested in finding the time of the day that congestion is most likely to occur. In this section we study the hourly behavior of congestion quantities and costs and examine, if any distinct hourly pattern is detectable. We first examine the hourly demand pattern. For this we use the hourly data for system load and energy clearing prices, which are published as the System Hourly 60

71 Clearing Price Reports by ISO-NE for each day [22]. The typical system load reaches its lowest point at night and rises around 7am, when the workday starts; it experiences a small peak in the evening around 9pm. Fig shows the average hourly load for May, ,000 hourly average system load in MW 14,000 12,000 10,000 8,000 6,000 4,000 2, hour Figure 4.25: Average hourly system load for May 1999 If we consider that the demand for energy is low at night and therefore less energy has to be dispatched, we expect the transmission network to be less intensively used and therefore with little if any congestion. To check this assumption, we determined the average hourly congestion quantities for each month and normalized them with the respective lowest hourly congestion in each month. Thus a value of 1.0 corresponds to the lowest average hourly value. We are normalizing so as to have some basis for comparing across months. We found that the assumption of lower congestion at night is true for several months, but not for all. For example, the months with the highest amount of congestion, March, April and May, 2000, experienced lower congestion during the 61

72 hours of night than during the day (Fig. 4.26). The average values during the night hours increased by nearly 40-50% for the peak daily hours in March and April, 2000, and by 20-30% in May, normalized hourly congestion quantity hour Mar-00 Apr-00 May-00 Figure 4.26: Average hourly congestion for March, April and May 2000 normalized hourly congestion quantity hour Aug-00 Nov-00 Figure 4.27: Average hourly congestion for August and November

73 Such an assumption clearly does not hold in August and November, 2000 (Fig. 4.27), since congestion was typically higher at night than during the day. Average hourly congestion quantities at night exceeded the values during the day by 40-50% in August, November, 2000, exhibited volatile average hourly congestion quantities, showing the highest peak during the night hours and a lower one during the day. Based on these observations, congestion is random during the hours of a day and fails to follow a distinct pattern. More surprisingly, the hourly congestion amount is not necessarily correlated with the hourly demand. The hourly energy clearing price is typically strongly correlated with the hourly system load: as the demand rises so does the ECP. Fig shows as an example the average hourly ECP and the average hourly system load for May average hourly ECP in $/MWh hour 16,000 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 average hourly system load in MW ECP system load Figure 4.28: Average hourly system ECP and load for May

74 Since the total hourly congestion costs do not provide very detailed information, we were interested in determining additional insights on the congestion relief financial aspects. Unfortunately, information on the hourly offers is not available. Therefore we computed the average hourly per MWh mitigated costs of congestion from the ISO-NE congestion reports for each month of the study period. We found out that per MWh mitigated costs were typically slightly higher at night than during the day. This behavior is opposite to the behavior of the ECP. A representative curve is shown for August 2000 in Fig average hourly costs and prices in $/MWh hour per MWh mitigated costs ECP Figure 4.29: Average hourly per MWh mitigated costs and ECP for August 2000 How do these findings affect the total resulting congestion costs? We first illustrate this for March and April, As we found above, these months experienced lower congestion during the hours of night than during the day. Hourly per MWh mitigated congestion costs were slightly higher during the night hours than during the day. As a 64

75 result the two months do not show a discernible pattern of average hourly total congestion costs. Figures 4.30a and 4.30b show average hourly congestion quantities and per MWh costs and the resulting average hourly total congestion costs for March and April, average hourly costs in $/MWh 60 1, hour 1,200 1, average hourly congestion in MW average hourly normalized total costs hour per MWh mitigated costs congestion Figure 4.30a: Average hourly per MWh mitigated costs and congestion for March 2000 Figure 4.30b: Average hourly normalized mitigated congestion costs for March 2000 hourly average costs per MWh in $/MWh hour 1,400 1,200 1, hourly average congestion in MW average hourly normalized total costs hour per MWh mitigated costs congestion Figure 4.31a: Average hourly per MWh mitigated costs and congestion for April 2000 Figure 4.31b: Average hourly normalized mitigated congestion costs for April

76 We get a different result for August, We f ound in our analysis above, that average hourly congestion quantities were lower during the hours of the day than at night in this month. Average hourly per MWh mitigated congestion costs were slightly higher at night than during the day. In this case we get as a result average hourly total costs that are substantially higher during the hours of night than during the day. We illustrate this in Figures 4.32a and 4.32b. hourly average costs per MWh in $/MWh 60 1, hour 1,200 1, hourly average congestion in MW average hourly normalized total costs hour per MWh mitigated costs congestion Figure 4.32a: Average hourly per MWh mitigated costs and congestion for August 2000 Figure 4.32b: Average hourly normalized mitigated congestion costs for August 2000 The analysis of the hourly data in this section shows that congestion and its associated costs do not occur in a predictable pattern throughout the day. Congestion quantities do not follow the same pattern as system load. Also, total congestion costs do not follow the same pattern as the congestion quantities. Therefore, improvement of congestion costs may not be equivalent to improvement of congestion. The results of this section can be 66

77 used for further studies on effective measures for determining economically efficient levels of congestion. 4.3 The geographic distribution The occurrence of congestion is not uniform in the entire ISO-NE region. In this section, we study the distribution of congestion within the twelve transmission regions of New England based on the 26 month data used in our investigation. For the names of the transmission areas please refer to Table 2.1 in Section 2.1. Three regions are responsible for the majority of the congestion in the New England system. We first study the distribution of the total congestion in MW over the 26-month period. W MASS 0.29% ME/NH 345 & 115 kv NORTH 1.47% 0.36% MAINE 3.27% SEMA/RI 3.28% VT 5.93% NEMASS/BOST 42.90% SW CONN 19.95% CONN 22.49% Figure 4.33: Distribution of congestion in MW within transmission areas 67

78 The pie chart in Fig shows the percentage of the total amount of congestion that occurred in each transmission area. 85% of the total system-wide congestion occurred in only three of the twelve areas. The NEMASS/BOST area experienced the biggest portion with a 43% share of the total amount. The CONN area had 22% and SW CONN had 20%. The other nine areas had 15% of the share. The pie chart in Fig presents in a telling manner that the amount of congestion in the three areas NEMASS/BOST, CONN and SW CONN is nearly six times the amount of the nine other areas. To see the occurrence of congestion in these three areas we plotted the monthly congestion amounts as a function of time over the period from May, 1999, to June, These plots are shown in Fig Note that the SW CONN area basically experienced significant congestion during 2000 with peaks in March and April as well as July, August and September and a lower one in December. Figure 4.34: The contribution of transmission areas to the monthly congestion for May 1999 June

79 Significant congestion in the CONN region, however, was concentrated in a small number of months. The largest congestion quantities occurred during February to June, 2000, lower peaks were experienced in December, 1999, and February and March, The other transmission areas did not experience larger amounts of congestion during 1999, during the months of 2000 congestion values rose, but fell again during the first half of From this distribution we can detect which transmission areas were the main causes for the system-wide peaks of congestion. Since the dense NEMASS/BOST area experienced uniformly high levels of congestion from September, 1999, on, it did not account for the extraordinary high peaks of congestion. The three highly congested months in 2000 were clearly the result of high congestion in the CONN and SW CONN transmission areas. The high values in March were mainly caused by the two regions together, whereas the high quantities in April and May mostly resulted from congestion in the CONN area, together with also higher than usual congestion in the other transmission areas during May, We next examine the distribution of the total congestion costs across the twelve transmission areas. There are some additional insights we gain from the cost data. We summarize the cost distribution in the pie chart in Fig Here, 85% of the systemwide costs constitute the share of NEMASS/BOST and the Connecticut areas. The NEMASS/BOST area, which experienced 43% of the total system congestion, accounted for 53% of the resulting costs. The two Connecticut regions have a share of 32% of the costs with 19% being the share of the congestion costs for CONN and 13% the share of the costs for SW CONN. 69

80 EAST 0.14% W MASS 0.22% NORTH 0.26% ME/NH 345 & 115 kv 0.88% SEMA/RI 2.08% MAINE 3.18% VT 6.98% SW CONN 13.33% NEMASS/BOST 53.45% CONN 19.46% Figure 4.35: Distribution of mitigated costs within transmission areas mitigated average costs in $/MWh EAST NORTH MAINE NEMASS/BOST VT average costs W MASS SEMA/RI total costs CONN SW CONN period of May 1999 to June 2001 SOUTH ME/NH 345 % 115 kv WEST mitigated total costs in $ Figure 4.36: Per MWh mitigated costs and total mitigated costs for each area for May 1999 June

81 The interplay of congestion amounts given in Fig and Fig and the congestion costs in Fig need to be related to the per MWh costs plotted in Fig for each transmission area. The total costs of congestion over the 26-month period are also shown in the diagram. We observe that while the areas EAST and NORTH experienced the highest per MWh costs, they had negligible total costs. This is due to the fact that congestion in these areas occurred only during a few days, resulting in extraordinarily high per MWh costs. Of the three highest congested areas NEMASS/BOST, CONN and SW CONN, none shows particularly high per MWh costs, but each experienced congestion on a high number of days. The number of days with congestion and the number of days with a certain range of per MWh costs are given in Table 4.4 for the discussed areas. The total number of days during the observed time-span was 792. Table 4.4: Number of days with per MWh mitigated costs in certain range over the May 1999 June 2001 period area days with congestion below 120 $/MWh number of days with respective range of costs between 120 and 200 $/MWh between 200 amd 400 $/MWh between 400 and 900 $/MWh above 900 EAST NORTH NEMASS/BOST CONN SW CONN $/MWh 71

82 Figure 4.37: The contribution of the transmission areas to the total monthly mitigated congestion costs for May 1999 June 2001 The distribution of the shares of mitigated congestion costs from Figure 4.35 across time is shown in the plots of Figure In the study of these plots and in comparison with those of Fig we note that: The NEMASS/BOST area experienced a higher volatility in total congestion costs over the 26 months than in congestion amounts due to the time variability of the per MWh costs of congestion. The share of congestion amounts and that of costs of the CONN region are essentially the same across the study period with peaks experienced during the February to June, 2000, period and again in February to March,