Forecasting of System Marginal Price of Electricity Using General Regression Neural Network

摘要

In the electric market system marginal price (SMP) can help the company accept a fitness bidding strategy and yield good economic returns. But it is difficult to be forecasted because of its complexity and uncertainty. A general regression neural network was proposed to forecast SMP because its foundation of probability conforms to SMP's uncertainty. The key of smoothing parameter was optimized by an improved particle swarm optimization method and three main factors of electrical load, historical corresponding hour SMP value and current SMP tendency were considered as independent variables. The simulation from actual data showed this method is effective.