EXHIBIT 99.448 January 1998 EQUILIBRIUM AMONG POWER MARKETS COMPETING FOR TRANSMISSION JAMES G. KRITIKSON AND ROBERT B. WILSON ABSTRACT. We demonstrate that in a power system in which several markets or "power exchanges" for energy compete for customers and for transmission services, as in California, there are multiple equilibria. In one of these each market accepts the transmission usage charges imposed by the system operator, whereas in the other each market uses an internal system of usage charges to curtail its transmission demand so that the system operator sees no congestion. Because the latter equilibrium enables each market to avoid payments of usage charges, we expect that this equilibrium will prevail. INTRODUCTION The restructured power industry in California allows several power exchanges to compete for customers and for transmission services. The enabling legislation created the Independent System Operator (ISO) to manage transmission, and the Power Exchange (PX) to provide daily and hourly markets for energy. The PX is not a monopoly, however. The thee investor-owned utilities are required to trade though the PX for the first few years, but other producers and consumers can trade though any other qualified "scheduling coordinator" (SC). Besides the PX, the entities that can operate as SCs include private energy markets that compete with the PX for customers and for transmission services. The basic requirement imposed on a SC is that its schedules submitted to the ISO are balanced; i.e., injections must equal extractions from the transmission system. The ISO treats all SCs equally; e.g., each SC, including the PX, is responsible for its own losses and ancillary services, and each pays the same usage charge ($/MWh) for transmission between zones. Each SC can participate in all of the markets run by the ISO. These markets include day-ahead and hour-ahead markets for transmission and a real-time market for supplemental energy to meet the ISO's requirements for system balancing on a short time frame; in addition, the ISO runs central markets for ancillary services from which each SC can purchase whatever resources it does not provide itself To simplify exposition, however, we concentrate here on the day-ahead market for delivery in a single hour. The PX's day-ahead market is a pool in which all transactions are settled at an hourly clearing price for energy, plus a transmission usage charge specific to each hour and zone (which need not be exactly the same as the ISO's usage charges). In contrast, other SCs can base their transactions on other market designs, such as bilateral contracting. Thus, the restructured California market is unique in its provision for various power exchanges, each operating as a SC, to compete for customers based on different market designs and the extent and quality of auxiliary services. In principle, all SCs are on an equal footing, except for the PX's initial monopoly on trading by the utilities. Presumably their competition for customers will reveal the relative efficiency of their alternative market designs, each of which has inherent strengths and weaknesses. For instance, as a pure-exchange market-clearing mechanism the PX is limited in its ability to take a net position long or short in the market, or to conduct arbitrage, but it has the advantage that its market structure meshes exactly with the ISO's markets. Similarly, a private bilateral market must be carefully managed to ensure efficient utilization of transmission, and to reconcile long-term contracts with the near-term forward markets for transmission, but it has the advantage of more flexibility (e.g., a SC can trade in the PX or with other SCs, but the PX cannot easily trade directly in another SC). In this article we examine a basic theoretical issue at the core of the vigorous debate that surrounds the California restructuring. This issue concerns the outcome of the SCs' competition for transmission services. Will the outcome be efficient? How can the PX or any other SC conduct an energy market in advance of the ISO's transmission market? We begin with a description of how the ISO actually conducts its transmission market, and then we address this issue with a theoretical prediction about the nature of an equilibrium among the strategies used by the various SCs. This prediction will be tested empirically in the next few years by the actual operations of the California market. THE ISO'S TRANSMISSION MARKET The IS 0's day-ahead market for inter-zonal transmission operates as follows. At a prescribed time each morning, each SC submits a balanced schedule, called its preferred schedule, together with "adjustment bids" whose role will be described below. For the PX, the preferred schedule is the result of financially binding transactions in its energy market, but for other SCs there is no requirement that the preferred schedule is based on completed transactions. The adjustment bids are options offered to facilitate congestion management by the ISO; they may be, but need not be, bids that were rejected in the energy market. Using these preferred schedules and adjustment bids, the ISO uses an optimal power flow (OPF) program to calculate a modification to each preferred schedule, called the advisory re-dispatch, and the associated inter-zonal transmission charges obtained from this optimization. As the name indicates, the advisory re-dispatch schedules and the associated usage charges are not binding on the SCs; they are merely indicative of what the ISO's final schedules and usage charges will be if none of the preferred schedules are modified (and no alteration in transmission limitations occurs in the interim). After receiving its advisory re-dispatch and the associated usage charges, each SC has an hour to decide whether it chooses to (a) adhere to its preferred schedule, (b) accept the advisory re-dispatch, or (c) submit a revised schedule. Adjustment bids cannot be changed but they can be withdrawn for use in a revised schedule, and they can be traded among SCs. Finally, at noon the ISO repeats the OPF calculations to determine the final set of adjustment bids used to achieve inter-zonal balancing, and the day-ahead inter-zonal usage charges that are now binding on each SC. Subsequent deviations from these final schedules are reconciled in the hour-ahead and real-time markets at newly calculated prices (except that the cost of intra-zonal balancing during real-time operations is absorbed by the ISO). The OPE works as follows. Its objective is to minimize the cost of achieving inter-zonal transmission feasibility. This cost is nil if the aggregate of the SCs' preferred schedules (plus anticipated loop flow, etc.) implies inter-zonal flows within specified limits, but if some of these limits are exceeded then the program selects adjustment bids to reduce the flows until they are within the limits imposed by line capacities and security considerations. Each adjustment bid applies to a particular supply unit or demand node and the zone in which it is located, and it is classified as either an increment (me) or a decrement (dee). An inc from a supply unit specifies a supply curve indicating the schedule of energy prices ($/MWh) at which that unit's output rate can be increased at the option of the ISO. Similarly, a supply dec specifies a demand curve indicating the schedule of energy prices at which that unit's output rate can be decreased, interpreted as a buy-back of supplied energy previously committed in the preferred schedule. Thus a supply inc entials a payment to the supplier and a supply dec entails a payment from the supplier. Demand incs and decs are analogous, but we will mostly ignore their role here. Transmission feasibility can be achieved by some combination of two basic operations, simplified here by ignoring loop flow so that one can suppose that one zone is an import zone and another is an export zone. The flow across the lines between an import zone and an export zone can be reduced by either: (a) invoking supply ines in the import zone and the same quantity of supply decs in the export zone, or (b) invoking demand decs in the import zone and the same quantity of demand ines in the export zone. As mentioned, the OPF selects the least cost combination of ines and decs that achieve transmission feasibility, but this is subject to an important constraint that is the novel feature of the California design. The novel constraint is that each SC's advisory re-dispatch can use only incs and decs submitted with its preferred schedule. This constraint is enforced in the OPF by appending the constraint, one for each SC, that its advisory re-dispatch must also be balanced, with injections equaling extractions, just like its preferred schedule. This constraint ensures that the advisory re-dispatches do not implicitly trade energy among the SCs by incrementing one SC's schedule and matching it with a decrement from another SC's schedule. On the other hand, this constraint allows trades "in kind," in the sense that invoking adjustment bids from one SC enables another SC to avoid adjustments. The "in kind" feature of these implicit trades is evident in the following consequence. Suppose that the PX's adjustment bids are relative to an energy price of $20 /MWh so that supply ines are offered at higher prices and decs are offered at lower prices, whereas another SC's adjustment bids are relative to an energy price of $30 /MWh. The key fact is that this $10 /MWh difference is irrelevant in the OPF calculations: the same result would obtain if all the SC's adjustment bids were decreased by $10 /MWh so that they are relative to the same energy price of $20 /MWh. One way to see this is to realize that from the perspective of the OPF (after appending the novel constraint) all that matters is the difference, say in basic operation (a) above, between the offered price of the supply inc in the import zone and the matching dec in the export zone. For instance, if the SC's pair of adjustment bids invoked are a supply inc at $32 /MWh in the import zone and a dec at $28 /MWh in the export zone, then the cost of these adjustments perceived by the OPE is only the net cost of $32-$28=$4 /MWh for the pair. Exactly the same net cost would result if the SC's adjustment bids were standardized around the PC's energy price of $20 /MWh, so that the calculatiOn of the net cost would be ($32-$l0)-($28-$10)= $4 /MWh. The advisory re-dispatch received by a SC consists of its preferred schedule as modified by its offered adjustment bids that are invoked in the OPF calculations. The ISO's associated inter-zonal usage charge is the most expensive of the adjustment options invoked among all the SCs. For example, it might be that the most expensive of the PX's adjustment pairs invoked in the OPE calculation costs $2 /MWh, obtained from a supply inc offered at $21 IMWh and a dec at $19 /MWh, whereas among all the SCs the most expensive pair costs $4 1MM/h, obtained from a supply inc offered at $32 /MWh and a dec offered at $28 /MWh. In this case the ISO reports an inter-zonal usage charge (for flows in the direction from the export zone to the import zone) of $4 IMWh. Thus, the PX is subject to the system-wide usage charge of $4 1MM/h even though internally no pair was invoked that is so expensive. The usual reason for this disparity is that the ines and decs are offered as discontinuous step-functions, so when the PX's cheapest pair after the $2 1MM/h pair is exhausted costs $5 /MWh for the next step, the OPF selects the cheaper pair offered by another SC at anet cost of $4 1MM/h. In the hour between receipt of the advisory re-dispatches and the final submissions due at noon, the SCs can trade adjustment bids among themselves. This provision gives the SCs a last chance to arbitrage differences between their basic energy prices. Usually the PX cannot trade profitably in this informal market for adjustment bids, however, since it cannot trade adjustment bids for money (which would require it to take a net position), and all the potential gains from trade "in kind" were exhausted in the advisory redispatches. The OPF calculation run after the final submissions is analogous, except that each SC's preferred schedule might be replaced by its advisory re-dispatch schedule or by a revised schedule. The analog of the advisory re-dispatch is now the final schedule for each SC, and the newly computed inter-zonal usage charges are now binding financial obligations payable to the ISO, which conveys them to the owners of the transmission assets. Notice that if all SCs adhere to their preferred schedules then the final schedules will be the same as the advisory re-dispatch schedules, since the OPE will again find the same set of adjustment pairs to be the cost minimizing ones. Similarly, if every SC accepts its advisory re-dispatch then no further adjustments are necessary. SELF-MANAGEMENT OF TRANSMISSION CONGESTION This brings us to our main subject, which is the role of self management of transmission congestion by the SCs. To understand congestion self-management (CSM) it is easiest to consider first the case that the PX is the only SC. In this case, traders in the PX might strongly prefer that usage charges be imposed by the PX rather than the ISO. For instance, suppose that if the PX adheres to its preferred schedule then the ISO's usage charge will be $4 /MWh, paid to the owners of transmission assets (which include the utilities that trade in the PX as well as those municipal utilities that choose to join the ISO). The PX traders and management know in advance the amount of this usage charge because it is reported to them along with the advisory re-dispatch. Consequently, they can also foresee that if the PX itself imposes a $4 /MWh usage charge then this will sufficiently curtail transmission demand to prevent the ISO from imposing any usage charges at all. Thus, by imposing its own usage charge of $4 /MWh and submitting at noon a revised schedule (e.g., its advisory re-dispatch) that in effect invokes its own adjustment bids, the PX can earn a surplus ($4 /MWh minus the average cost of the adjustment pairs invoked) that would otherwise have been paid to transmission owners. This surplus must ultimately be refunded to the traders in the PX, but if this is done in a way that preserves the inter-zonal price differences then this refund will not disturb the traders' incentives. [[It should be noted that this scenario depends on the current plan in California for the ISO to convey usage charges to the owners of transmission assets. An alternative scheme weakens the incentives for CSM. In this scheme the PX traders operate essentially as a cooperative. Transmission congestion contracts (TCCs, in the form of contracts for differences compared to a strike price) are auctioned to the traders and the proceeds conveyed to the transmission owners; thereafter, usage charges are payable to the owners of TCCs, which in effect are the traders themselves, so the incentives for CSM are muted]] The question we address here is whether the incentives for CSM persist when several SCs compete for transmission access. The matter is more complicated in this case because of a fundamental free-rider problem. For example, the PX prefers that other SCs curtail their demand for transmission to reduce or eliminate the ISO's usage charges, thereby enabling the PX to obtain its preferred schedule without invoking any adjustments. When the PX and all other SCs take this position, the net result is that no reduction in usage charges occurs, and each winds up with its advisory re-dispatch as its final schedule, and each pays usage charges to the ISO. EQUILIBRIUM IN THE TRANSMISSION MARKET There are several candidates for an equilibrium among the SCs' strategies. The three pure symmetric candidates correspond to the three options available in responding to the advisory re-dispatch. We label these: (A) All SCs adhere to their preferred schedules. (B) All SCs accept their advisory re-dispatch schedules. (C) All SCs submit revised schedules. Asymmetric mixtures among these are also possible, but here we consider only these three candidates. Although (C) is not specified exactly, what we have in mind is CSM, in the sense that (as in the PX example above) each SC submits a revised schedule that reduces congestion by invoking sufficient adjustment pairs to reduce its transmission demand to the amount in its advisory re-dispatch. For the PX, which operates as a pool, there is essentially only one way to do this, which is to impose its own internal usage charges. This follows from the fact that any pricing rule that yields an efficient allocation of trades subject to transmission limitations is equivalent to a uniform clearing price for energy, plus a system of usage charges for inter-zonal transmission. This is in fact the rule used by the PX, which bases all settlements on zonal prices that are equivalent to a uniform state-wide energy price plus usage charges represented by the differences between zonal prices (if one zone is designated as a reference zone, then its zonal price is the same as the. state-wide energy price). Our main proposition is that each one of these three candidates is in fact an equilibrium. That is, given that one SC, say the PX, expects the others to follow the strategy indicated in (A), then the PX cannot benefit from following a different strategy -- and similarly for (B) and (C). But there is an important corollary. The equilibrium (A) entails payment of usage charges to the ISO, whereas the others do not. Consequently, we expect that over time the most likely equilibrium is (B) or (C), and due to its simplicity, we expect (B) to prevail. The following subsections consider each candidate in turn. EQUILIBRIUM (A) Suppose that all but one SC adheres to its preferred schedule. For convenience, suppose the exception is the PX. We now ask whether it could be better for the PX to choose one of the other two options. To see that neither of these options is better than adhering to its preferred schedule, observe that in both alternatives the traders in the PX forego gains from Wade in order to reduce congestion, and a portion of the benefits accrue to traders in the other SCs who do not contribute to reducing congestion, since these other SCs adhere to their preferred schedules. Moreover, and this is the key, every adjustment pair that the PX accepts (as in B) or invokes voluntarily (as in C) merely eliminates the ISO's need to invoke an adjustment pair in the PX or some other SC. Thus, each self-management effort made by the PX is nullified by the open door for further transmission it provides to other SCs. To see this argument in operation, consider an example in which there is only one SC besides the PX and for each of these two SCs its preferred schedule implies transmission of 150 MWh from the export zone to the import zone, whereas its advisory re-dispatch implies transmission of 100 MM/h, for which the associated usage charge is $4 [MM/h. Suppose further that the energy supply and demand curves in each are linear, so that for each the potential gain from trade lost because of limited transmission capacity is (150-100 MWh)x($4 /MWh)/2 $100, which is in addition to the $400 that each must pay in usage charges for the 100 MM/h it transmits, so the actual total is $500 for each. Now assume that the other SC adheres to its preferred schedule that implies 150 MM/h of transmission, and consider the decision problem of the PX. Since supply and demand curves are linear, it suffices to consider just one case, which we take to be the one in which the PX curtails its transmission demand from 150 MWh to 100 MWh as indicated in the advisory re-dispatch. This reduces the aggregate demand for transmission from 300 MWh to 250 MWh, but there remains an excess demand of 50 MM/h, so the ISO still needs to invoke adjustment pairs to achieve feasible schedules. Presumably the PX has used up its least expensive adjustment pairs in curtailing its transmission demand by 50 MM/h, so those pairs remaining all cost more than $4 1MM/h. Therefore, for its final schedules the ISO invokes adjustment pairs only from the other SC, and in doing so it uses those 50 MWh of adjustments with costs increasing up to $4 /MM/h. Thus, the net result is that the usage charge remains $4 1MM/h and the final schedules allow 100 MM/h of transmission by each SC. Each SC pays $400 in usage charges and foregoes $100 in potential gain from trade. Thus, if the other SC adheres to its preferred schedule, then the PX cannot gain by accepting the advisory re-dispatch, nor by submitting a revised schedule. These arguments are subject to several qualifications. For instance, they assume that more than one SC contributes to congestion. And they assume that each SC's marginal cost of transmission reduction (the cost of its marginal adjustment pair) is the same; e.g., in the above example, if the PX's marginal cost is $4 1MM/h but the other SC's marginal cost is $3 /MM/h then voluntary reduction by the PX might indeed reduce the usage charge from $4 /MM/h to $3 /MWh. In general, our argument is that o If each SC is a perfect substitute for the other in terms of the cost of its marginal adjustment pair, then each cannot do better than to adhere to its preferred schedule when it expects the others to do so. This equilibrium indicates that vigorous pursuit of gains from CSM might be fruitless in the California context. This reflects a presumption that the SCs competing for transmission access will have, at the margin, marginal costs of adjustment pairs that are closely comparable. There may be occasional situations in which one's marginal cost is significantly higher than others (e.g., because the others are all at capacity limits), so its CSM efforts could reduce usage charges to the level of the highest among the other SCs. As we shall see, however, there is a more basic reason why CSM might actually succeed, which is that there is another equilibrium such as (B) or (C) in which all SCs are better off than in equilibrium (A). EQUILIBRIA (B) AND (C) Next we demonstrate that (B) and (C) are equilibria. We lump these together because they are similar when (C) is interpreted as each SC's submission of a revised schedule that approximates its advisory re-dispatch. So, we focus on (B), in which each SC accepts its o advisory re-dispatch. If the SCs were to collude, acting like a single entity, then the situation would be comparable to the special case examined above in which the PX is the only SC, for which we previously established that an internal system of usage charges could prevent the payment of usage charges to the ISO and the transmission owners. Our argument here, however, is that even when there are multiple SCs, a common strategy of accepting advisory re-dispatches is a stable arrangement. Thus, we need to show that no one SC, such as the PX, can gain by reverting to an alternative strategy. Suppose then that the PX expects all other SCs to accept their advisory re-dispatches, and consider the decision problem of the PX about whether to do the same. We claim that in this case the PX cannot gain from adhering to its preferred schedule, nor from submitting a revised schedule. An example conveys the main idea. As in the example illustrating equilibrium (A), suppose that each SC's preferred schedule implies transmission of 150 MM/h from the export zone to the import zone, whereas its advisory re-dispatch implies transmission of 100 MM/h, for which the associated usage charge is $4 1MM/h. Also, the energy supply and demand curves in each SC are linear, so the potential gain from trade lost because of limited transmission capacity is $100. If the PX accepts its advisory redispatch, as do all the others, then there is no residual congestion and the ISO's final schedule imposes no usage charges. Consequently, the only cost to the PX is the foregone gain from trade of $100. Alternatively, suppose it adheres to its preferred schedule. Because the other SCs have accepted their advisory re-dispatches, the only excess demand for transmission is the 50 MM/h from the PX's preferred schedule. To eliminate this excess demand, the ISO will not invoke any adjustment pairs of the other SCs, since all those whose cost is less than $4 1MM/h were used up in constructing their advisory re-dispatch schedules, so the OPE will select only the 50 MM/h of adjustment pairs from the PX whose costs are no more than $4 /MM/h. Thus, the net result is that the PX's re-submitted preferred schedule is modified by the ISO to obtain a final schedule that is the same as its advisory re-dispatch. Moreover, the final usage charge imposed by the ISO is $4 1MM/h, since that is the cost of the marginal pair invoked from the set offered by the PX. The total cost to the PX is the $100 of foregone gain from trade plus $400 in usage charges, which is distinctly worse than the outcome if the PX had accepted its advisory re-dispatch. One can also imagine howls of protest from the other SCs, since now they too must each pay $400 in usage charges as a consequence of the PX's deviant action. The PX's other alternative is to submit a revised schedule. As the foregoing example illustrates, however, no revised schedule can do better than its advisory re-dispatch, since that succeeds totally in eliminating usage charges payable to the ISO, and any further reduction in transmission demand would unnecessarily sacrifice gains from trade that are obtainable within the available transmission capacity. Thus, our second conclusion is that: o If each SC is a perfect substitute for the other in terms of the cost of its marginal adjustment pair, then each cannot do better than to accept its advisory-redispatch when it expects the others to do so. This second equilibrium indicates that CSM might indeed succeed in California. Although both (A) and (B) are possible equilibria, (B) has the advantage for the SCs that they retain the usage charges that would otherwise be passed through the ISO to the transmission owners. However, (B) is a more fragile equilibrium than the analogous version of (C) in which the SCs curtail their transmission demands further to reduce the chance that the ISO might stilt impose usage charges. For instance, suppose that at noon it is revealed that new estimates of loop flow indicate that transmission capacity is less than previously estimated. In equilibrium (B) this results in usage charges of $4 /MM/h or more, because all of the less expensive adjustment pairs are exhausted by the SCs' advisory redispatches, so the OPE selects additional adjustment pairs that cost more. In contrast, in equilibrium (C) each SC can voluntarily restrain its transmission demand more than required by the advisory re-dispatch to reduce the risk that revisions in the estimates of transmission capacity will require further adjustments. CONCLUSION The thrust of our exposition is that the California design admits two basic equilibria, one (A) in which each scheduling coordinator adheres to its preferred schedule, and another (B) in which each accepts its advisory re-dispatch or (C) submits a revised schedule that approximates its advisory re-dispatch. These differ substantially, however, because in (A) they pay usage charges to the ISO, whereas in (B) and (C) they avoid these charges. Implementing (A) is straightforward, since each SC can pay the invoice from the ISO simply by passing on these charges to its traders. Implementing (B) or (C) is more complicated because it requires that each SC imposes internally a system of usage charges, and then refunds to its traders the difference between the revenue collected and the cost of the adjustment pairs invoked in some way that does not weaken their incentives to curtail transmission demand. The simplest policy is to refund the surplus pro rata based on each trader's volume of transactions: since this refund does not differ by zones, it preserves the zonal price differences in settlements and thereby maintains the right incentives to curtail demand for transmission.