Stochastic OPF and Transaction Monitoring in Deregulated Power Markets

点击翻译按钮获取中文摘要

The deregulated electric power industry requires new optimal power flow (OPF) and state estimation tools to maintain high levels of security while operating near full capacity in competitive markets. Stochastic OPF problems incorporate operating contingencies and random effects such as uncertainty in demand or energy prices. No operator in today's markets can afford the cost of operating so that its system can withstand fully every contingency, regardless of the probability of its occurrence. Because so many contingencies are included, these problems are massive, to large for direct methods. We propose the use of Monte Carlo-based importance sampling coupled with Benders decomposition, techniques that produce subproblems of tractable size. The proposed work will first develop Monte Carlo-based stochastic OPF algorithms for a DC network subject to random failures of lines and generators. The second stage will add uncertainty to loads, requiring separate importance sampling for the two sources of uncertainty. Finally, the stochastic model will be extended to include day-ahead purchase of spinning reserve in the face of uncertain loads and energy prices. The algorithmic methods for these three classes of OPF will then be extended to a full AC network. A number of public discussions have identified a growing problem arising from undisclosed bilateral energy transactions that distort a network's planned operating point as they pass through it. The task of identifying such "rough" transactions can be formulated as a problem in hypothesis testing in power system state estimation. The potentially daunting number of undisclosed buyer-seller combinations to be tested can be accommodated by careful problem formulation and by efficient computational techniques such as low rank updates. The central conceptual step is proper use of all a priori information.The proposers will draw on their experience in state estimation-observ-ability, efficient computation, bad data detection and identification-and in time-coupled dynamic dispatch, unit commitment, and price-responsive load modeling to develop efficient algorithms for both classes of problems by exploiting common aspects of their problem structure. This work will lead to new estimation and hypothesis-testing paradigms, robust approached to stochastic OPF, and efficient, practical computation algorithms.