A Primer on Pricing, CMSC, and Gaming in a Simple Electricity Grid

Transcription

1 Draft: February 12, 2014 A Primer on Pricing, CMSC, and Gaming in a Simple Electricity Grid INTRODUCTION Purpose This memorandum uses a simple three-node model of an electricity grid to explore key features of the Ontario electricity market and the recent controversies over alleged gaming behaviour. The working of the uniform price / two-settlement system and its associated Congestion Management Settlement Credits (CMSC) payments is explored. The examples developed explain in an easily understood way the means by which market participants could target higher CMSC payments via offering and bidding strategies, and how competition and rivalry among market participants for CMSC opportunities may be affecting the situation. The memorandum will also identify the key recommendations made by the Market Surveillance Panel (MSP) in respect of gaming situations, and the market rule amendments made by the IESO in response. Before we begin to develop the three-node model it is important to understand the distinction between uniform pricing and locational pricing. Uniform Pricing and Location Pricing on an Electricity Grid The MSP has made many recommendations over the life of the market relating to pricing, CMSC, and alleged gaming behaviour. Perhaps their most fundamental recommendation is that the province move towards some form of locational pricing and away from the uniform price / two-settlement system. It is important to understand what these pricing systems are as their fundamental features will be present in every example developed in this memo. Uniform pricing and locational marginal pricing (LMP) are two alternative pricing and congestion management schemes for a wholesale electricity market. They concern the pricing of the wholesale electricity commodity but also have implications for the rationing of transmission capacity on the grid, and what kinds of incentives will exist to reinforce the grid and locate new generation and load. An electricity transmission grid has locations, or nodes, where generators and loads (such as LDCs and industrial loads) connect. Transmission grids always exhibit transmission constraints (limits on how much a given transmission line 1

2 can transmit) and line losses (some of the energy injected at a generation node dissipates as waste heat before being withdrawn at a load node). The constraints and line losses mean that the cost to the system of supplying energy at different nodes can be different. In a uniform pricing system such as Ontario s all loads pay the same commodity price and the cost differences of supply across nodes are socialized into uplift charges. In a full LMP system the cost differences of supply among nodes are reflected in different prices for each node. These nodal prices are what generators get paid, and loads pay, depending on which node they are located at. The Shadow of LMP Most electricity markets in the U.S. and some in other parts of the world use nodal pricing or LMP whereas the Ontario market does not. Why has Ontario not opted for an LMP system? After all, analysts generally agree that at least in theory LMP provides better signals for generation and load investment location, and for grid reinforcement investments. Perhaps the origins of electricity markets in the U.S. help to explain the adoption of LMP south of the border. Electricity markets in the U.S. evolved from a process where separate investor-owned utilities (IOUs) who covered their own service territories formed power pools for trading of electricity. The motive to trade electricity in a pool would arise because each IOU had its own cost structure deriving from its endowments of transmission and generation assets. These cost structures would lead to differing bundled electricity costs across the IOUs service territories. Differing prices would provide the motive to trade in a pool. The original power pools later evolved into the wholesale markets of today such as PJM, and ISO-New England. The pre-existing cost differentials apparent across the IOUs composing a market footprint would live on in the market era as differences in transmission rates and locational prices for the commodity. There probably was not a sudden change from a market wide uniform price to geographically varying LMPs. Thus there would have been more limited rate shock in the transition to an organized market. Ontario by contrast has not generally exhibited locational differences in pricing in the century plus history of the electricity industry here. As a result the introduction of full LMP in the Ontario setting is a different matter. The Market Design Committee held to the view that Ontario should move to LMP but only after a transition period of 18 months after market opening. They cited concerns over equity and potentially greater complexity with an LMP system as reasons for a transition period. However once the uniform system was in place, it ha proven very difficult to move away from it. 2

3 While it may be true that LMP is more complicated to run 1, it is also true that Ontario s grid does exhibit locational differences in the marginal cost of supply, and these locational costs are used in the operation of the system. In fact, every five minutes in the IESO-administered market a market clearing price (MCP) is calculated on a uniform basis, but so are shadow prices Ontario s version of nodal prices - at every node on the grid. These shadow prices are not used to settle payments for market revenues but they are used to determine whether or not CMSC is paid. Now we can begin to consider examples of market pricing within the simple three-node grid model. A THREE NODE GRID The three node grid device is frequently used in expositions of locational pricing. This memo will use it to also explain the operation of uniform pricing and the associated side payments we call CMSC. The memo will also use it to consider potential gaming strategies and what may happen when participants compete over gaming opportunities. The three nodes are each an equal distance from each other (thus the grid is an equilateral triangle). We also assume that there are no line losses on the grid. This assumption allows us to focus on congestion constraints and makes computations very easy. For now we assume that two of the nodes have only generation, and one of the nodes has only load. Figure 1: A Three Node Grid 1 LMP advocates such as William Hogan maintain the opposite that LMP is simple. Deviating from LMP is actually more complicated in the view of Hogan and others. 3

4 NW NE S The two generation nodes are called NW and NE (for Northwest and Northeast respectively) while the load node, S, symbolizes the southern Ontario load centres. The generator, G1, located at NW is assumed to have a marginal cost of generation of $20 / MWh, while G2, located at NE has a marginal cost of generation of $30. These costs are the offer prices that the generators submit into the electricity market. Both generators have 100 MW of capacity. Dispatch of the System No Constraints G1 is first in the merit order 2 and thus will be used to meet the first amounts of load at S. Under the assumptions we have made about the grid (equal line lengths, no losses, and, for now, no constraints) the energy injected at NW to meet load at S will follow very specific paths to arrive at S. Specifically, if total load at S is equal to 80 MW, then G1 must inject 80 MW. Of the 80 MW injected at NW two thirds of that amount (or 53.3 MW) will travel along the NW S transmission route, while one third (or 26.7 MW) will travel along the NW NE- S transmission path. This two to-one partitioning of the energy injection reflects the fact that the NW-NE-S transmission path is twice as long as the NW-S path and thus has twice the resistance. 2 The merit order is the list of generator offers stacked from lowest to highest price. It is thus the supply curve of generation in the market. The terms merit order, supply curve, and stack are frequently used interchangeably. 4

5 Figure 2: Dispatch with no Constraints Injection = NW NE Load = 80 S Once load at S exceeds the 100 MW capacity of G1 then the G2 generator which is next in the merit order is dispatched. The fact that energy is injected at two nodes has consequences for energy flows on the grid. Suppose load has risen to 130 MW. Thus 100 MW are injected at NW and 30 MW are injected at NE (see Figure 3). Figure 3: Injections at Both Nodes 5

6 = 23.3 Injection = 100 Injection = 30 NW NE = = 76.7 Load = 130 S Along the NW to NE line the energy injected at the generation nodes flow against one another. Thus the net flow is the difference between the two flows. Along the two other sides of the triangular grid the flows from each generator are additive. Pricing In the Grid with No Constraints Now it is time to discuss how much loads will pay and generators will get paid in the system as developed so far. Since we have not introduced any transmission constraints at this point there is no distinction between LMP and uniform pricing. There would also be no need for a two-schedule system. The MCP for the first 100 MW of load would be $20. This would also be the shadow price at all nodes on the system. The Shadow Price The Shadow Price at a node is the marginal cost of serving another MW of load at that node. In a system with no transmission constraints or line losses each node could be served by the generator that is at the margin in the merit order. Thus the shadow price at all nodes would be that generator s offer price. If transmission constraints are present then generation will have to be redispatched around the constraint the merit order will have to be departed from. This will result in locational differences in shadow prices. 6

7 TRANSMISSION CONSTRAINTS Now we introduce transmission constraints. As noted above all transmission lines have limits beyond which they cannot transmit energy. Two major types of limits are: Physical limits: as the energy flow on a line increases the line will actually heat up and begin to sag. In the extreme the line could melt. Security limits: the system operator may place a limit on a line that is more restrictive than the physical limit. This is because a contingency elsewhere in the system say a failure of another transmission line could suddenly increase the energy flow on the line in question. Thus power flow on this line is restricted so that in the event of a contingency the higher flow on the line will not reach an intolerable level. In Figure 2 above a failure of the NW NE line would suddenly increase the flow of power on the NW S line from 53.3 MW to 80 MW. In the rest of this memo we will assume that there is a 50 MW limit on the NW S transmission line. 7

8 Figure 4: 50 MW Limit on NW S Line, Violated Injection = NW NE Limit = Load = 80 S With a limit of 50 MW on the NW S line the dispatch of 80 MW from G1 at NW is no longer feasible despite the fact that G1 has 100 MW of capacity and is first in the merit order. In fact, at a load of 75 MW the flow of energy on the grid would just reach the point of congestion as 50 MW would flow from NW to S and 25 MW would flow along the longer NW NE S route. How does the system operator handle this situation? It must depart from the merit order and re-dispatch around the constraint. The more expensive G2 must be activated even before G1 s capacity is exhausted. So if load rises from 75 MW to 76 MW what must the system operator do? It will not be sufficient to leave G1 at the 75 MW output level and inject 1 MW from G2. If the operator dispatched this way, the limit on NW S would be violated. To see why assume the operator orders this dispatch. Of the 1 MW injected by G2, two thirds of it would flow along the NE S path, while one third of it would flow along the NE NW S path. Thus the flow along the NW S path would reach 50 1/3 MW (or which we will abbreviate to 50.3 in this memo). Instead the system operator must dispatch G2 to 2 MW and order G1 to reduce output by 1 MW. This re-dispatch will leave total net injections at 76 thus just meeting load. More importantly, it ensures that the flow of energy along the constrained NW S path is kept at no more than 50 MW. How is this so? The extra two MW from G2 increases flow along the NW S path by two times one third of a MW or two thirds of a MW (2 * 1/3 = 2/3). At the same time the loss of 8

9 one W from the G! generator reduces flow along the NW S path by one times one third of a MW (1 * 1/3). As load increases beyond 76 MW the system operator must continue to constrain down the G1 generator and constrain up the G2 generator. Indeed there is appoint of maximum transfer capability on this system where G1 is constrained off all the way to zero and all load is supplied by G2. This point of maximum transfer capability would be 150 MW (assuming G2 has 150 MW of capacity). If load at S reaches 150 MW all generated by G2 then 100 MW would flow along the NE S path while 50 MW would flow along the NE NW S path and the constraint would not be violated. SHADOW PRICES We noted earlier that as long as the system is unconstrained (in our examples, as long as load is 75 MW or less) there would be no difference between uniform pricing and locational pricing, but that once the system is congested (ie., once load exceeds 75 MW) a pattern of locational differences in prices will emerge. Before getting into the specifics of uniform versus locational pricing let s examine concretely the behaviour of shadow prices in system as we have developed it so far. Figure 5 50 MW Limit with Corrected Dispatch Injection = = 20 Injection = 10 NW NE = 50 Limit = = 30 Load = 80 S 9

10 Figure 5 above shows the dispatch from Figure 4 corrected to respect the 50 MW constraint on the NW S line. What can we say about the shadow prices at the three nodes in the system? Recall the definition of shadow price: The Shadow Price at a node is the marginal cost of serving another MW of load at that node. We said earlier that the NW and NE nodes were generation-only nodes. However, loads could be installed at either of these nodes. If there was an extra MW of load at NW it could be supplied directly by the G1 generator as long as G1 s capacity is not exhausted. This direct supply of an extra MW would not use the transmission system since the extra MW supplied and demanded are located at the same node. Therefore the shadow price at NW is $20 / MWh. The same holds true at NE so the shadow price there must be $30 / MWh. But what is the shadow price at the S node? We saw earlier that once the point of congestion was reached the system operator was forced to depart from the merit order and re-dispatch the system around the constraint. We also saw that this required for each extra MW demanded at S, that two extra MWs be injected at NE, and one less be injected at NW. This means that the marginal cost of supplying extra MWs at S is: Shadow price at S = 2 * $30 1 * $20 = $40 / MWh. Figure 6 below repeats Figure 5 with the shadow prices included: 10

11 Figure 6: Dispatch, Flows, and Shadow Prices when Load = 80 Injection = = 20 Injection = 10 NW 20 NE 30 Limit = = = 50 Load = S The shadow prices shown above reflect the marginal system cost of increasing load by one MW at each node. The shadow price pattern shown here will recur in most if not all of the examples that follow. These shadow prices are related to generator offers (which are assumed to reflect generator marginal costs of production) into the market. Some observations: The shadow prices exist as above regardless of what congestion pricing system is in place (uniform versus LMP). Indeed in Ontario shadow prices are used to dispatch the system and determine when CMSC payments are made. The shadow prices are closely related to generator offer prices but not necessarily equal to them. In this simple example two of the three shadow prices are equal to a generator offer price while the third one is a linear combination of the two generator offers. Question for reader: If load at S is equal to 80 MW and G1 s offer price is $15 / MWh, while G2 s offer price is $35 / MWh, what will the shadow prices at NW, NE, and S be? SETTLEMENTS: UNIFORM AND LOCATIONAL MARGINAL PRICES 11

12 Now we are ready to discuss how Ontario s uniform pricing system compares to a locational marginal pricing system. Locational Marginal Pricing In a full LMP system the nodal prices would be the same as the shadow prices analyzed above. Thus the G1 generator would be paid $20 / MWh when producing while the G2 generator would be paid $30 / MWh when producing. Loads at S would pay a $20 / MWh price whenever the system was not congested but this price would rise to $40 / MWh whenever load exceeds 75 MW and the system is congested. 3 Does this mean that the shadow prices in Ontario today would be the LMPs if LMP were to be adopted? Probably not as the shift in payment flows brought about by such a change might induce changes in generator offering behaviour and thus result in a set of LMPs different from the shadow prices observed under the uniform price regime. 4 Proponents of LMP suggest that it provides better signals for investments in generation and load, as well as incentives for grid investments. A glance at Figure XX above supports this view. The NW node has more generation than can be used once the system is generated. It also has the lowest price for energy which would discourage new generation investment from locating there and would incentive new loads to locate there. The opposite is true at S where the location price is high and there is no generation present. In addition loads located at S would support reinforcement of the NW S transmission line so as to mitigate the effects of congestion and allow more of NW s low-priced generation to reach S. Uniform Pricing and the Two Settlement System While the shadow prices we have analysed so far will figure importantly in the examples to follow it will be in the context of a uniform pricing system resembling the one in Ontario rather than an LMP system. Ontario s uniform pricing system requires the use of two dispatch algorithms each which create their own dispatch schedule. The market algorithm (or market sequence) produces a market 3 Note that this pattern of prices gives rise to congestion rents. If load equals 80 MW in this example, the system operator will collect $3,200 each hour while paying generators a total of only $1,400 plus $300 equals $1, The explanation here for why LMPs under nodal pricing might be different from shadow prices under uniform is frequently offered. However it seems to hint at defects to the competitive process. Generators marginal costs are unlikely to change greatly in a switch from uniform to LMP. If competition is working properly offers would be forced towards marginal cost. If offers can suddenly change because the pricing regime changes, one reason might be that competition is not effectively disciplining offers to equal marginal cost. 12

13 schedule dispatching generation under an assumption of no transmission constraints. By contrast the dispatch algorithm (or dispatch sequence, or physical sequence) produces a dispatch schedule or physical schedule treating transmission constraints as they actually are. The market schedule is used to determine the five- minute market clearing price (MCP) and the Hourly Ontario Energy Price (HOEP). 5 The MCP is what generators are paid and loads are charged. The dispatch schedule is used to actually create dispatch instructions for the generators. Both market and dispatch schedules are then used to determine side payments paid to generators (and dispatchable loads) CMSC payments. Does this sound complicated? It is but this memo aims to present the system in its essentials to make it as easy to understand as possible. 6 How does this system work? If MCP is set using the market schedule it might be $20 / MWh even when G2 is dispatched and running at a cost of $30 / MWh and thus losing $10 for every MWh produced. This is where CMSC comes into the picture. Congestion Management Settlement Credits Market designers intended the uniform pricing system in Ontario to be a temporary stage on the way to a full LMP system. During this temporary phase CMSC side payments would ensure that generators and dispatchable loads would get the same operating profit (or operating benefit for dispatchable loads and exporters) that they would get if the market schedule was actually the true dispatch. For a generator the market sequence would produce a Market Quantity d for Injection (MQSI). The constrained sequence would produce a Dispatch Quantity d for Injection (DQSI). If the two were different a CMSC payment would be calculated based on the difference in operating profit between the two schedules: CMSC for generators = Operating Profit (MQSI) Operating Profit (DQSI) If the true dispatch from the dispatch algorithm caused a reduction in operating profit, the CMSC payment would offset it. 7 5 Note that this memo will not in general distinguish between the five minute MCP and the HOEP. We will generally refer simply to the MCP. 6 Footnote 1 noted the view of William Hogan and others that LMP is the simpler system. 7 CMSC can be negative in rare occasions when the dispatch algorithm produces a higher profit for the participant than the market algorithm does. 13

14 Operating profit in the market schedule is defined as: Operating profit(ms) = (MCP Offer price) * MQSI Similarly operating profit in the dispatch schedule is defined as: Operating profit(ds) = (MCP Offer price) * DQSI Therefore CMSC is: CMSC = (MCP Offer price) * MQSI - (MCP Offer price) * DQSI = (MCP Offer price) * (MQSI DQSI) Thus a generator such as G2 in our example receives a positive CMSC because its offer price is greater than MCP and G2 has been constrained on and therefore DQSI is greater than MQSI. In Figure XX above the CMSC payment to G2 is $100: CMSC G2 = (MCP Offer price) * (MQSI DQSI) = (20 30) * (0 10) = 100 G1 will also be eligible for CMSC but in this case it would be a constrained off CMSC payment. While the rationale for constrained on CMSC is relatively straightforward the generator s cost is higher than the MCP but the generator is needed so there must be a top-up payment the rationale for constrained off CMSC is less clear. It has, however, the effect of returning the generator to the profit it would have made under the market schedule. T CMSC G1 = (MCP Offer price) * (MQSI DQSI) = (20 20) * (80 0) = 0 14

15 In this case, because the generator s offer price is actually equal to MCP the CMSC payment nets out to zero. The table below summarizes the results of this example. Table 1 Dispatch and CMSC When Load = 80 Real Time Market Real Time Dispatch Load MCP 20 - Shadow Price 1-20 Shadow Price 2-30 Shadow Price 3-40 G1 Dispatch G2 Dispatch 0 10 G1 CMSC - Off 0 G2 CMSC - On 100 CMSC, Anomalous Behaviour, and Gaming The MSP has reported on many scenarios involving anomalous behavior where market participants are paid what seem to be quite large CMSC payments. Some of the participants involved in these scenarios are now the subjects of gaming investigations being undertaken by either the MSP or the IESO s MACD. In some cases the behavior in question has involved individual participants and the scenarios have involved offer or bid prices that seem distant from marginal cost or marginal benefit, combined with the impacts of the 3 times ramp rate assumption in the market schedule. In other cases the anomalous outcomes seem to apply in an entire region of the province and may involve competitive or- non-competitive - behavior among participants whose offers and bids take account of the offers and bids made by other participants in the same region. This more strategic or interactive anomalous behavior can be thought of as systemic anomalous behavior. The simple model used in this memorandum will not provide much insight beyond the most basic regarding the first type of anomalous behavior. However 15

16 the model can be pushed a little further to help in illuminating the more systemic type of anomalous behavior. We will cover these two topics in turn. The Incentive to Move Offers Away from Marginal Cost Consider the dispatch when load = 80 as portrayed in Figure 6 and Table 1 above. Is it possible for either G1 or G2 to increase their CMSC payments by changing their offer prices? Consider G1 at the NW node. G1 is constrained down so its market schedule of 80 MWh is greater than its dispatch schedule of 70 MWh. In this case G1 s incentive is to lower its offer price. However in the confines of this simple model this will not result in a higher CMSC payment for G1. As G1 lowers its offer price the shadow price at NW and the MCP both fall as a result. The situation is different for G2. Since G2 is constrained on above its market dispatch its incentive is to raise its offer price. G2 is not the generator setting the MCP when load is at 80. Therefore as it raises its offer price its CMSC payment does increase. This outcome remains the same in the model as load rises up to a level of 100 MWh which is the maximum capacity of the two generators. Beyond a load of 100 MWh the MCP is set by G2 s offer. In this range the ability to increase CMSC payments by moving offer prices away from marginal cost follows a pattern opposite to the previous case Figure 7: Dispatch, Flows, and Shadow Prices when Load = 110 Injection = 40 NW = -10 Injection = NE 30 Limit = = = 50 Load = S 16

17 With a load of 110 MWh G1 will no longer set the MCP. IT is constrained down to an injection of 40 MWh but now will enjoy larger CMSC payments if it lowers its offer price. By contrast G2 now is setting the MCP and its attempt to raise CMSC payments by increasing its offer price will not succeed as it will lift MCP as well. This effect will cancel out any CMSC it could earn. Calculating the Re-Dispatch Around the Constraint In a real electricity grid with many generation and load nodes, different ramp rates for every generator and dispatchable load on the system, and several transmission constraints, the calculation of the physical dispatch is a large linear programming problem carried out by a very sophisticated computer program. Fortunately the same calculation in the simple three node model analyzed here is an easy exercise. As seen in Example 1 the three node grid becomes congested at a load of 75 MW. Every extra MW after 75 requires a re-dispatch of the system. Specifically, each extra MW of load requires two extra MW injected at the NE node and one less MW injected at the NW node. This re-dispatch ensures that the 50 MW limit on the NW S line is never breached. So for a load of say 100 MW what would the dispatch look like? 100 MW is 25 MW greater than the 75 MW congestion point. Thus 2 * 25 = 50 MW are needed at the NE node and = 50 MW are needed at the NW node. Verifying that this dispatch leaves the loading on the NW S line at 50 MW is an exercise left to the reader. What is the re-dispatch needed for a load of 123 MW? The answer is: 96 MW injected at NE node; 27 MW injected at NW node the reader should verify. 1 17

18 Systemic Anomalous Behavior The MSP has commented several times about one persistent and peculiar feature of the Ontario electricity market and that is the deeply negative shadow prices in the Northwest. On one level this pattern is not surprising as the Northwest is an over-generated region with only limited transmission capacity for the export of power to the rest of Ontario and outside the province. Shadow prices are expected to be low in these circumstances. However the persistence of deeply negative shadow prices in the Northwest seems to go beyond what one would expect from the normal effects arising from re-dispatching generation around a constraint. While shadow prices are not necessarily equal to the offers or bids entered at given nodes on the system, these offers and bids are key influences on shadow prices. The offers and bids typically observed in the Northwest so not appear to be consistent with normal concepts of marginal cost or marginal benefit of consumption. In addition the Northwest of Ontario garners a hugely disproportionate share of the CMSC payments made in the province. The MSP has observed some anomalous offering and bidding behavior in the Northwest. Participants seem to be competing over CMSC payments in a strategic fashion. 8 This behavior has been variously referred to as competition for CMSC or shadow price chasing behavior. Is it possible to understand these behaviours more clearly via the simple 3-node grid model used here? The answer seems to be yes. Example 1: Load = 80 MWh Now suppose there are two generators (G1 and G2) at the NW node in completion with each other. (Now the generator at the NE node is G3). Each has a capacity of 50 MW (for the same total of 100 MW at this node as in the previous examples). With load equal to 80 the dispatch at the NW node will be 70 (see Figure 6). To begin assume both generators offer at their marginal cost of $20 / MWh. We would expect them to share a dispatch of 35 MWh each while their market schedules are 40 MWh each. Profits for each generator consist of net energy revenues plus CMSC: 8 The term strategic is used here in its game theory sense where market participants make offers or bids taking into account what they think rivals will do. 18

19 Profit (G1) = Production Profit + CMSC Profit (G1) = (MCP MC) * DQSI + (MCP-PG1) (MQSI DQSI) where PG! is G1 s offer price Both generators earn zero profit as MCP = MC = PG1 = 20. Is there any way for one or both generators to use offering behavior to increase profits? The answer is no when load is at 80 MWh and the MCP is being set at this node. To see this suppose G1 lowers its offer price to, say, $19 / MWh. G1 will now get dispatched to produce 50 MWh with a market schedule of 50 MWh. G2 will have a market schedule of 30 MWh and a dispatch schedule of 20. The MCP will remain at $20 / MWh while the shadow price at the NW node will be $19 / MWh. Inspection of the profit equation will show that neither generator can reach a profit above zero as a result of G! lowering its offer price. This parallels the earlier finding shown in Figure 6 and Table 1. There is no configuration of G1 s or G2 s offer prices that will arrive at a different result. Example 2: Load = 110 MWh As before the story is different when load is at a level such that the MSP is set be generator G3 at the NE node. With load equal to 110 MWh the NW node will be dispatched to 40 MWh shared equally between G1 and G2 if their offer prices are equal. Their market schedules are each 50 MWh so each is now earning positive profits: Profit (G1) = (MCP MC) * DQSI + (MCP-PG1) (MQSI DQSI) = (30 20) * 20 + (30 20) (50 20) = = $500 for each generator. Now what happens if G1 lowers its offer price from $20 to $19? G1 will be dispatched to 40 MWh with a market schedule of 50 MWh. G2 will be dispatched to 0 MWh with a market schedule of 50 MWh. 19

20 G1 s profit is: Profit (G1) = (MCP MC) * DQSI + (MCP-PG1) (MQSI DQSI) = (30 20) * 40 + (30 19) (50 40) = = $510 Meanwhile G2 continues with a profit of $500 but it is composed entirely of CMSC payments: Profit (G2) = (MCP MC) * DQSI + (MCP-PG2) (MQSI DQSI) = (30 20) * 0 + (30 20) (50 0) = = $500 We can plot a profit function for G1 under the assumption that G2 keeps its offer at PG2 = $20 / MWh: Profit (G1) = (MCP MC) * DQSI + (MCP-PG1) (MQSI DQSI) = (30 20) * 40 + (30 PG1) (50 40) = * PG1 This function will hold for all offers less than $20 MWh. For offers above $20 up to $30 G1 s profit function looks like this: Profit (G1) = (MCP MC) * DQSI + (MCP-PG1) (MQSI DQSI) = (30 20) * 0 + (30 PG1) (50 0) = 1, * PG1 Figure 8 below shows this profit function for G1 under the assumption that G2 holds its offer price at $20 / MWh 20

21 Figure 8: G1 s profit function when PG2 = $20 Profit PG1 Let s assume that zero is the Minimum MCP and the minimum offer price that generators are allowed to offer. The clear message of Figure 8 is that G1 would enjoy the highest profits by lowering its offer price to the minimum possible of zero. 21

22 CMSC with Intertie Trades This section adds intertie activity to the simple three-node grid we are analyzing. With intertie transactions it is necessary to distinguish the pre-dispatch from the real-time runs of the DSO. The pre-dispatch runs of the DSO have both constrained and unconstrained sequences. These runs determine pre-dispatch market schedules and constrained schedules for all generators, dispatchable loads, and imports and exports. The runs also produce pre-dispatch market prices (in the unconstrained runs) and pre-dispatch shadow prices (in the constrained runs). The last pre-dispatch run before the dispatch hour sets the import and export schedules at each intertie. A key difference between the pre-dispatch and realtime runs of the DSO is that in the pre-dispatch runs imports and exports are fully equal to domestic resources in terms of price determination. In the real-time runs, imports and exports are fixed for the hour. Imports are placed at the bottom of the stack with an offer price of -$2,000 while exports are at the top of the demand curve with bid prices of +$2,000. These settings insure that the import and export transactions remain fixed for the hour and cannot set real time prices. Real time price determination then proceeds using the offers and bids of those resources that are dispatchable in real time. Example 1: An Import in the Northwest Suppose that everything is as before but there is a 50 MW import available at the NW interie with an offer price of $18/MWh. Offers are: Resource Offer Quantity Offer Price G G IM If load is at 80 MW the pre-dispatch market schedule will take all 50 MW offered by the import, IM1, plus an additional 30 MW taken from G1. The pre-dispatch market clearing price will be set by G1 at $20. However, this market schedule would violate the constraint on the NW S line. Thus in the constrained schedule G1 will be constrained down to 20 MW while G2 will be constrained up to 10 MW. 22

23 Because the two domestic generators G1 and G2 have dispatch schedules that are different from their market schedules they are both entitled to CMSC payments: G1 Constrained off CMSC = (RTMCP Offer) * (RTMQSI RTDQSI) = (20 20) * (30 20) = 0 G2 Constrained On CMSC = (20 30) * (0 10) = 100 G1 s CMSC payment works out to zero because G1 s operating profit in the market schedule would have been at a break-even level of zero. G2 s CMSC payment offsets the operating profit loss it would have from being constrained up at a market clearing price that is lower than G2 s offer. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM1 Dispatch G1 Dispatch G2 Dispatch IM1 CMSC 0 - G1 CMSC - Off - 0 G2 CMSC - On Example 2: An Import from the Northeast This time suppose there is an import offer of 50 MW at the NE intertie rather than the NW intertie, with an offer price of $22: Resource Offer Quantity Offer Price G IM G

24 Everything else about the example remains the same. In the pre-dispatch market schedule all load would be met by G1. However, in the dispatch schedule, this would violate the constraint so G1 must be constrained down and IM2 must be constrained up. IM2 will set the pre-dispatch shadow price at the NE node. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM2 Dispatch G1 Dispatch G2 Dispatch IM1 CMSC - ON 22 - G1 CMSC - 0 G 2 CMSC - 0 IM2 s CMSC payment is determined in the pre-dispatch sequence: IM2 CMSC = (PDMCP Offer) * (PDMQSI PDDQSI) = (20 22) * (0 10) = 22 Neither domestic generator is eligible for CMSC as their market schedules are the same as their dispatch schedules in real time. Gaming IM2 is receiving a constrained-on payment as it is more economical than the G2 domestic generator available at the same node. This importer would have an incentive to try to increase its CMSC payment by raising its offer price. It could do so without changing the schedule up to just below $30. At an IM2 offer above $30 the G2 generator would become economic and begin setting the predispatch shadow price at the NE node. This would force IM2 out of the dispatch 24

25 schedule as its offer would exceed the shadow price at NE, and IM2 would no longer qualify for a CMSC payment. Example 3: Two Import Offers at the NW Intertie In this example there are two import offers at the NW intertie in other words, there is a rising supply curve of imports at that node. Resource Offer Quantity Offer Price G G IM IM Load is unchanged at 80 MW so in this example in the unconstrained predispatch the entire load can be met via the imports alone. This sets a predispatch MCP of $19. However, 80 MW injected at the NW intertie will violate the constraint so IM2 must be constrained down and G2 must be constrained up, each be 10 MW. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM1 Dispatch IM2 Dispatch G1 Dispatch G2 Dispatch IM1 CMSC - Off 0 - G1 CMSC - Off - 0 G 2 CMSC - 0 In the real-time market schedule the two imports scheduled in pre-dispatch would enter the stack at the bottom. G1 is dispatched to 10 MW. However, in the realtime dispatch schedule G1 is constrained off and G2 is constrained on to 10 MW. 25

26 IM2 s CMSC payment is determined in the pre-dispatch sequence: IM2 CMSC - Off = (PDMCP Offer) * (PDMQSI PDDQSI) = (19 19) * (30 20) = 0 G1 s would also be eligible for a constrained off payment but the amount nets out to zero as its offer price is setting the real time MCP. G2 will receive a constrained on payment: G2 CMSC - On = (RTMCP Offer) * (RTMQSI RTDQSI) = (20 30) * (0 10) = 100 Gaming in Example 3 If IM2 tries to lower its offer price in order to raise its CMSC payment it will only manage to lower the pre dispatch MCP by the same amount and will not be able to raise its CMSC. The same applies to G1 in real time. However, G2 has the potential to increase its constrained on CMSC by raising its offer price. Each $1 increase in its offer price would increase its constrained on payment by $10. Example 4: Maximum Load on System Under the assumptions used so far the grid has a maximum capacity of 150 MW. Higher levels of load can be accommodated but only if supplied locally by generation at each of the three nodes. 150 MW is the maximum transfer capability of this grid. At this level of load 100 MW would be transmitted along the NE S line while 50 MW would ship along the NE NW - S line. Energy injections at the NW node would be constrained off to zero. Suppose the offer stack is the same as in Example 3 (except that G2 must be able to offer a quantity up to 150 MW). What then are the pricing, dispatch, and CMSC implications? 26

27 Resource Offer Quantity Offer Price G G IM IM In the pre-dispatch market schedule, IM1 and IM2 will be dispatched to their maximums, and G1 will make up the remaining by supplying 50 MW. Thus predispatch MSP will be set by G1 at $20. However, in the pre-dispatch physical schedule all this generation will be constrained off to zero. G2 will be constrained on to supply the full load. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price 3 - * - * IM1 Dispatch IM2 Dispatch G1 Dispatch G2 Dispatch IM1 CMSC - Off IM2 CMSC - Off 50 - G1 CMSC - Off G 2 CMSC - 0 * - the shadow prices at these nodes would be determined by the offer prices of any local generation that can meet an increment of load at the node in question without transmitting any additional power between nodes. We assume that at the NW and NE nodes there are generators who can match the G1 and G2 offers respectively. In this example the CMSC payments will be fairly large: IM1 s and IM2 s CMSC payment is determined in the pre-dispatch sequence: IM1 CMSC - Off = (PDMCP Offer) * (PDMQSI PDDQSI) 27

28 = (20 18) * (50 0) = $100 IM2 CMSC Off = (20-19) * (50-0) = $ 50 G1 s CMSC payment is determined in the real time sequence: G1 CMSC - Off = (RTMCP Offer) * (RTMQSI RTDQSI) = (30 20) * (100 0) = $1000 Although G2 is constrained on for the full load of 150 MW its CMSC payment works out to zero as G2 s offer actually sets the real time MCP. Gaming in Example 4 The opportunities for participants to increase CMSC constrained off payments in this example are quite large. This is because in this configuration the pre dispatch MCP has been set by G1 so both IM1 and IM2 can gain higher CMSC payments by lowering their offer prices. In this case there is no limit on how low the offers can go, and how high the constrained off payments can go until the lowest allowed offer price of -$2,000 / MW is reached. 9 If the two importers lower their offer prices deep into negative territory the shadow price at the NW node will be set by lowest of the two offers. But since the MCP is set by G1 (who will be constrained off in the real time dispatch anyways) both importers are guaranteed to be dispatched in the market schedule. Thus, if both importers set their offer prices at -$1,000, each would be eligible for a constrained-off payment of $49,000. In the real time dispatch G1 has an equally attractive opportunity to profit by changing offer prices. In real time the MSP is set by G2 at $30. Thus G1 is eligible for a positive constrained off payment. The lower G1 sets its offer price the higher this payment will be. Thus if G1 sets an offer price of -$1,000 it would receive a constrained off payment of $103,000. G1 s offer of -$1,000 would set the real time shadow price for the NW node at -$1,000. This pattern of deeply negative shadow prices (both pre dispatch and real time) and very high constrained off payments has become very characteristic of the actual Northwest in Ontario s electricity market. 9 In the IESO administered markets the minimum market clearing price is -$2,000 and this is also the lowest allowed offer price for any resource. 28

29 Competition for CMSC in Example 4 The MSP has observed a tendency for market participants to compete for CMSC payments especially in the Northwest. They have not yet reported findings related to this behaviour as it is still the subject of analysis. However, the simple model presented here may be of some use in organizing thinking about competition for CMSC. Some interesting patterns are available by comparing competition in Example 4 s limiting case where maximum transfer capability on the grid has been reached with Example 5 s (below) intermediate case where load is somewhat below maximum transfer capability. As noted above the potential to game large CMSC payments in Example 5 is limited only by the minimum MCP (and offer price) of -$2,000. [NTD: Have to analyze how competition for CMSC will be affected by a congested intertie with zonal intertie pricing] Example 5: An Intermediate Load Level Example 5 is similar to Example 4 except that the load level is not so high, maximum transfer capability is not yet reached, and some generation can still be injected at the NW node. Suppose the load level is 130 MW. The offer stack is the same as in Example 4. Resource Offer Quantity Offer Price G G IM IM In the pre-dispatch market schedule the two import offers will be fully dispatched and G1 will be dispatched up to 30 MW. Thus G1 will set the pre-dispatch MCP at $20. However this dispatch would violate the NW S constraint. So in the pre-dispatch dispatch schedule G2 is constrained on to 110 MW; G1 is constrained off completely as is IM2 while IM1 is constrained down to 20 MW. 29

30 Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM1 Dispatch IM2 Dispatch G1 Dispatch G2 Dispatch IM1 CMSC - Off 60 - IM2 CMSC - Off 50 - G1 CMSC - Off G 2 CMSC - 0 All three resources at the NW intertie / node will be eligible for CMSC. IM1 and IM2 have their CMSC determined in the pre-dispatch sequence: IM1 CMSC - Off = (PDMCP Offer) * (PDMQSI PDDQSI) = (20 18) * (50 20) = $60 IM2 CMSC Off = (20-19) * (50-0) = $50 In real time G1 is constrained off while G2 is constrained on: G1 CMSC - Off = (RTMCP Offer) * (RTMQSI RTDQSI) = (30 20) * (100 0) = $1000 As in Example 4 G2 is constrained on but its CMSC payment works out to zero as G2 s offer sets the real time MCP. Gaming in Example 5 In this example, as in Example 4, the two importers have significant opportunities to increase CMSC payments by lowering their offer prices. Neither will affect the pre-dispatch MCP as this has been set by G1. Indeed G1 will also have an 30

31 opportunity to increase its CMSC payment by lowering its offer price in the real time dispatch. Competition for CMSC in Example 5 It is natural to expect other players to be attracted to the potential to earn large constrained off CMSC payments which by definition to not require the supply of actual production. Wholesale market participants offering imports at the NW intertie are the obvious candidates. How would this competition play out? Assume for now that there is no limit on the intertie 10. New importers could enter with offers lower than those already present. These offers would at some point displace the G1 generator from the per-dispatch stack. Consider the effect of a third import offer with an offer price of -$10. Resource Offer Quantity Offer Price G G IM IM IM With load at 130 MW the pre-dispatch MCP would now be set by IM1 at $19. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM1 Dispatch IM2 Dispatch IM3 Dispatch G1 Dispatch G2 Dispatch Intertie congestion and the associated congestion pricing are analyzed in the next section. There is no need to include intertie congestion here as no CMSC payments are directly associated with intertie congestion. 31

32 IM1 CMSC - Off 60 - IM2 CMSC - Off 50 - IM3 CMSC - Off G1 CMSC - Off G 2 CMSC - 0 All three importers would be eligible for CMSC in the pre-dispatch: IM1 CMSC - Off = (PDMCP Offer) * (PDMQSI PDDQSI) = (19 18) * (50 0) = $50 IM2 CMSC Off = (19-19) * (50-0) = $ 0 IM3 CMSC Off = ( ) * (50-20) = $ 870 In real time G1 and G2 would both be eligible for CMSC payments: G1 CMSC - Off = (RTMCP Offer) * (RTMQSI RTDQSI) = (30 20) * (100 0) = $1000 Again, G2 is constrained on but its CMSC payment works out to zero as G2 s offer sets the real time MCP. IM3 s negative offer price ( I ll pay you to take my energy ) results in a large CMSC payment. It also reduces IM1 s CMSC payment and eliminates IM2 s CMSC. We would expect these importers to respond in the next period by reducing their offer prices. In the simple model analyzed here this competitive process would continue ad infinitum, as importers compete to keep as large a share as possible of the CMSC pie. What is the logical end-point of this competition? As more importers rush to compete for CMSC with larger quantity offers, and lower offer prices, they will eventually begin to set the pre-dispatch MCP. Thus the MCP would plunge into negative territory as it would reflect the lowest offer 32

33 price that meets the load of 130 MW. This would shrink the CMSC pie ultimately down to zero. 11 Text box 2: MR-0239-R00, June 26, 2003 On June 26, 2003 the (then) IMO Board implemented an urgent market rule amendment MR-0239-R00 which limited CMSC payments made to generators and importers with negative priced offers. These participants had their offer replaced with a zero offer for the purpose of calculating constrained off CMSC. This urgent market rule amendment had the effect of ensuring that constrained off generators could not be paid more the MCP. In effect, this rule amendment changes the CMSC formula for generators and importers with negative offer prices to: CMSC Off = (PDMCP max(0, Offer)) * ( PDMQSI PDDQSI) MR-0239-R00 can t be analyzed without bringing intertie limits and zonal pricing into the discussion. Exports and Dispatchable Loads So far the only intertie transactions we have considered are imports; and likewise the only domestic resources considered are generators. Now we add in dispatchable loads. 11 How intertie limits and the intertie congestion pricing in the Ontario market will change this result is the subject of the next section. Briefly, the total MW that can participate in this competitive process will be limited to the transfer capability of the intertie. In addition, the intertie zone pricing that is part of the Ontario market design (LMP at the interties) ensures that MCP is not pulled down by negative priced import offers. 33

34 Exports and dispatchable loads both increase load, or demand, on the grid. They are different from the load considered so far in that they are dispatchable and must submit bids to buy energy into the market. Now we must distinguish the load considered so far which is non-dispatchable load from exports and dispatchable loads. Like imports, exports participate in the pre-dispatch market and have set schedules during real time that are not changed. As for dispatchable loads, they participate in the real time market as do domestic dispatchable generators. Example 6: Dispatchable Loads Consider Example 1 with the addition of a dispatchable load, DL1, at the NE node. This dispatchable load is bidding to buy energy at $25. Resource Offer / Bid Quantity Offer / Bid Price Load 80 2,000 DL G G IM In the pre-dispatch schedule DL1 s bid price is higher than the MCP of $20 so it will be dispatched in the market schedule. However, DL1 s location on the grid means that it costs the system more than $20 to deliver power to DL1. The shadow price at NE is $30, so DL1 will be constrained off. Pre-Dispatch Market Pre-Dispatch Dispatch Real Time Market Real Time Dispatch Load MCP Shadow Price Shadow Price Shadow Price IM1 Dispatch G1 Dispatch G2 Dispatch DL1 Dispatch IM1 CMSC 0-34