طراحی مکانیزم بازار لحظه ای و مسائل مربوط به رقابت در برق

The performance of a sealed bid-offer and an open display real time uniform price double auction in a three-node radial network and compared in terms of efficiency, price competitivity and the distribution of surplus over a demand cycle. We also compare three versus six independent generation companies who own the same aggregate portfolio assets. The environment provides a stressful market for wholesale buyers on-peak (the competitive price is above resale value) and for baseload generator units off-peak (the competitive price is less than cost). Three firms are as competitive as six under sealed bid-offer, but not the double auction.

We report new experiments which compare the sealed bid offer (SBO) market mechanism, studied in Backerman et al. (1997; hereafter BDRS), with a uniform price double auction mechanism (UPDA) that updates nodal prices and allocations continuously as new bids and offers arrive in real time down to the close when the market is “called” and all standing accepted bids and offers become binding spot contracts. We compare the performance of the SBO and UPDA institutions in terms of their impact on incentives affecting market efficiency (the ability to exhaust the gains from exchange), generator and wholesale buyer profitability, and delivery price. Under each of the two trading institutions we compare markets in which the available generator capacities and their costs are held by three versus six independent companies. We also vary the minimum loaded capacity of baseload units below which avoidable fixed cost penalties are incurred. Wholesale buyers also face large penalties if they fail to serve all of their noninterruptible demand. These nonconvexities combine to produce a very stressful market for buyers on-peak, and for sellers off-peak peak. Finally, in UPDA only, we study the impact of varying the proportion of peak demand that is noninterruptible and subject to must-serve avoidable fixed cost penalties. We do not address the effect of a transmission line constraint-this is central to the market power issue-because it is already addressed in Backerman et al., 1997. Also we do not address the so-called “loop-flow” issue in triangular and more complex networks, because (a) it is essential to first examine mechanism design and nonconvexity issues in a baseline that controls for loop flow effects, and (b) such issues constitute one of the next steps in our research program underway.

Contrary to previous comparisons showing UPDAs high performance in classical environments, it performs very poorly in the current nonconvex environment where SBO does quite well. This is explained by the greater incentives to reveal, and to avoid manipulation strategies, in SBO, where there is no feedback of within-period information and therefore more costly to attempt manipulation of prices and profits. But even the shoulder demand levels do not rescue the version of UPDA studied here. Are there rule “fixes” that would improve UPDA? We have little doubt that there are many. The other side rule used in BRS is an obvious candidate. Also needed is a close rule that will promote earlier and better revelation, e.g. a random close, or endogenous close, fixed-interval call, rule (See McCabe et al., 1993, p. 311). Under the latter the market is called when no new tentatively accepted bid or offer arrives for a specified elapsed time. Another fix would be the Wilson (1997) or other heavier-handed forms of other side rules. The severity of the must-run generator constraint (50% versus 100%) and the must-serve demand constraint (100% versus 70%) are significant and important. We chose this parameterization because it roughly characterizes the existing situation in the electric power industry, which is largely a product of command/regulatory environments and has a history of poor incentives. The ongoing deregulation of several aspects of this industry may drastically change these constraints. For this reason it is particularly important that the system not be hamstrung by rules that hardwire assumptions that the future will look like the past. Time-of-day pricing technologies including network metering and switching devices will increase the deployment of voluntary, interruptible demand options, an efficient substitute for both generator and transmission capacity. Generator technologies that are already upon us, and new ones on the horizon introduce the prospect of gradually altering the stock of generator assets, introducing more flexible cost efficient means of electrical energy production. (Thomas and Schneider, 1997). Besides reducing nonconvexities, these technologies would increase reliability by reducing dependence on supply side generator spinning reserves, and increasing dependence on demand side flexibility.