基于Relief相关性特征提取和微分进化支持向量机的短期电价预测  

Short-Term Electricity Price Forecasting Using Relief-Correlation Analysis Based on Feature Selection and Differential Evolution Support Vector Machine

作　　者：彭春华[1] 刘刚[1] 相龙阳[1] 

机构地区：[1]华东交通大学电气与电子工程学院,南昌330013

出　　处：《电工技术学报》2013年第1期277-284,共8页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(51167005);江西省教育厅科技项目(GJJ 12288)资助

摘　　要：针对电价预测中特征输入量选择的盲目性,本文通过改进传统的Relief算法,提出一种电价预测输入量的自动选取方法,并引入相关性分析来剔除冗余特征。在此基础上,采用支持向量机来建立电价预测模型并应用微分进化算法来优化选择支持向量机的参数以达到提高预测精度的目的。以PJM电力市场的真实电价来进行仿真分析,结果表明本文的特征选取方法能够很好地提取电价的短期趋势特征和周期性特征,而微分进化优化的支持向量机也获得了比常规支持向量机和BP神经网络要好的预测结果,体现了本文方法的优越性。To avoid the unreasonable inputs selection for electricity price forecasting,based on the improvements of traditional Relief algorithm,a novel feature selection method is proposed in this paper.Moreover,correlation analysis is used to eliminate redundant features.Based on the selected features,the price forecasting model is established by using support vector machine(SVM),and differential evolution algorithm is employed to determine the best parameters of SVM.Simulation results based on the real electricity price data obtained from PJM electric power market demonstrate the proposed feature selection method can extract features that reflect the short-run and periodical characters of electricity price,and furthermore,the SVM optimized by differential evolution algorithm can achieve better forecasting results than BP neural network and common SVM.

关 键 词：电价预测 RELIEF算法 相关性分析 微分进化算法 支持向量机 

分 类 号：TM734[电气工程—电力系统及自动化]