Electricity price forecasting in a grid environment.
详细信息
文摘
Accurate electricity price forecasting is critical to market participants in wholesale electricity markets. Market participants rely on price forecasts to decide their bidding strategies, allocate assets, negotiate bilateral contracts, hedge risks, and plan facility investments. Market operators can also use electricity price forecasts to predict market power indices for the purpose of monitoring participants' behaviors. Various forecasting techniques are applied to different time horizons for electricity price forecasting in Locational Marginal Pricing (LMP) spot markets. Available correlated data also have to be selected to improve the short-term forecasting performance. In this study, Fuzzy Inference System (FIS), Least-Squares Estimation (LSE) and the combination of FIS and LSE are proposed. Based on extensive testing with various techniques, LSE provides the most accurate results and FIS, which is also highly accurate, provides transparency and interpretability.;Transmission congestions threaten the grid security and system reliability. Congestion costs also reduce the social welfare of electricity customers. The gross congestion cost for a transmission constraint is the product of shadow price and power flow on the constrained facility. Shadow price forecasting is seen by market operators as an additional decision-making support tool for congestion management. Similarly, different market participants may use shadow price forecasting as a tool for strategy improvement in dayahead or spot markets. The highly volatile nature of shadow prices requires forecasting models that are able to handle non-normality, spikes and volatility clustering. The proposed research is a statistical stochastic model for short- and medium-term shadow price forecasting in wholesale electricity markets based on locational marginal prices. It handles time series based on Generalized Auto-Regressive Conditional Heteroskedastic (GARCH) and nonlinear optimization of maximum-likelihood criteria. This statistical stochastic model can be utilized for short- and medium-term shadow price forecasting as well. An extended mean-reversion jump-diffusion process is developed that incorporates the correlated electricity price and shadow price. The proposed methods perform shortand medium-term shadow price forecasting and provide interpretable signals for different congestive conditions. The test cases for this dissertation are obtained from PJM Interconnection.