机构地区: 浙江大学数学系
出 处: 《电力系统自动化》 2004年第11期17-21,共5页
摘 要: 电力工业的市场化改革突出了市场清算价格(MCP)预测的重要性。文中以浙江电力市场为背景,提出了一种基于小波分解和神经网络的MCP预测方法。该方法对原电价数据进行了预处理,将经小波分解去除细节后的重构电价序列作为输入参数,并且依据“重近轻远”的原则及考虑到电价具有星期的周期性的特点,重新设计了神经网络拟合误差的代价函数。对浙江电力市场下一交易日的MCP进行了预测,预测精度达到90%左右。 Reforms of power industry to open electricity markets make market clearing price (MCP) forecasting more important. According to Zhejiang circumstance, this paper presents a wavelet-improved neural network method, which inherits the advantage of neural network model. In addition, by adding a data preprocess procedure, price series after subtracting details with wavelet transform are choosed as an input parameter, and error function of ANN is redesigned by 'emphasizing recent data' principle and considering 'week period' in MCP series. Application to the real power system shows the accuracy of the proposed method is about 90% .