GOALI: Collaborative Research - A Systems Approach to Modeling Real-Time Market Prices of Electricity

该研究项目的目的是确定在放松管制的市场中寻找实时电价统计分布的合适方法。它将使用一种自下而上的系统方法，通过生产成本计算模型来代表市场。该模型的组成部分将是构成市场的发电机的容量、可靠性和成本特征、燃料价格、需求以及影响它的气象影响。作为第一个近似值，每小时的价格将等同于用于供应每小时需求的最后一个发电机组的边际成本。最初的工作将集中在通过生产成本计算模型从市场供需双方的现有信息中找出每小时边际单位的统计分布。然而，经验数据表明，在寡头垄断市场下，当供应接近需求时，现货价格会超过边际生产成本。这一现象将在基于市场生产成本表示的随机框架下使用不同的经济模型进行研究。对整个统计分布进行建模的尝试需要考虑与发电机停运和按时间顺序的负荷相关的随机过程。该项目将开发适用于大型系统的近似分析程序，并将使用蒙特卡罗模拟来评估这些近似的准确性。通过将这些模型的预测结果与历史数据进行比较，将对计算进行验证。

The purpose of this research project is to determine suitable approaches for finding the statistical distributions of real-time electricity prices in a deregulated market. It will use a bottoms-up systems based approach in which the market is represented via a production-costing model. The components of this model will be the capacity, reliability, and cost characteristics of the generators that comprise the market, fuel prices, demand, and the meteorological influences that affect it. As a first approximation, the price at every hour will be equated to the marginal cost of the last generating unit used to supply each hour's demand. The initial effort will concentrate on finding the statistical distribution of the marginal unit at each hour via the production-costing model from the available information on the supply and demand sides of the market. Empirical data suggest, however, that when the supply approaches demand and under an oligopoly market, the spot price exceeds the marginal cost of production. This phenomenon will be investigated using different economic models under a stochastic framework based on a production-costing representation of the market. Attempts at modeling the entire statistical distributions require that the stochastic processes associated with the generator outages and chronological load are accounted for. The project will develop approximate analytical procedures that are applicable to large systems and will evaluate the accuracy of these approximations using Monte Carlo simulation. The computations will be validated by comparing the results predicted by these models with historical data.