基于神经网络的电力市场电价预测

Electricity Price Forecasting Based on the Neural Network

以美国加州电力市场为背景,在分析了市场清算电价(MCP)的影响因素的基础上,采用了一种基于反向传播(BP)网络预测下一日市场清算电价的方法。该方法考虑了系统供求关系、历史负荷、历史电价等对未来时段电价的影响,建立了一个单隐层的神经网络结构。预测模型融合了模糊理论,利用隶属函数对温度(最高温度、平均温度、最低温度)进行了模糊处理,将这些因素作为神经网络的输入量。在负荷高峰时段,往往存在市场外机组的调度和参与者的策略性投标等问题,这些因素共同作用容易造成电价尖峰。建立一个节假日模型来预测节假日的电价。采用美国加州电力市场的历史数据进行了训练和预测分析,结果表明该模型具有良好的预测效果。

This paper analyses the factors which may influence the MCP,adopt the method of forecasting the MCP by using the back propagation network（BPN） for California Power Market in America.The relation of supply volume and demand volume,historical price and load are considered in this method.With this method,present a neural network of simplex and implicit layer.Fuzzy theory is added in forecasting model,the temperatures（the lowest temperature,the mean temperature,the highest temperature） are the input factors of the neural network after being turned into fuzzy data.Price spikes frequently appear in the stage of load peak,because of the combined action of unit scheduling and the participator′s tactical bid.The unit is out of market.This paper presents a special model for price forecasting to weekend and public holidays.The historical loads and prices of California power market are used for training and forecasting,the results show that the proposed method is effective.