利用多因素小波变换和多变量时间序列模型的日前电价预测  

Day-ahead Electricity Price Forecasting Based on Multi-factor Wavelet Analysis and Multivariate Time Series Models

作　　者：谭忠富[1] 张金良[1] 

机构地区：[1]华北电力大学电力能源经济研究所

出　　处：《中国电机工程学报》2010年第1期103-110,共8页Proceedings of the CSEE

基　　金：国家自然科学基金项目(70571023);教育部新世纪优秀人才项目(NCET-06-02-08)~~

摘　　要：电力市场中,市场出清电价具有较强的波动性、周期性和随机性,实践证明单一的电价预测模型很难提高预测精度。针对该问题,提出一种基于多因素小波变换和多变量时间序列模型的日前电价预测方法。利用小波变换将历史电价序列和负荷序列分解和重构成概貌电价、细节电价和概貌负荷、细节负荷。用概貌电价和概貌负荷作变量建立多元时间序列模型,预测未来概貌电价;用单变量时间序列模型预测未来细节电价。将概貌电价和细节电价的预测结果求和作为最终的预测电价。采用上述方法对美国加州电力市场日前电价进行预测,并与对比模型进行了详细的比较分析,结果表明该方法能够提供更准确的预测电价。The electricity market prices are highly volatile, seasonal and stochastic. The previous literature shows that it is difficult to improve the accuracy of price forecasting by applying one model alone. Hence a novel hybrid model for day-ahead electricity price forecasting was presented in this paper. The proposed model is based on multi-factor wavelet analysis and multivariate time series models. Historical prices and loads were decomposed and reconstructed into approximate price series, detailed price series, approximate load series and detailed load series. The future approximate prices were forecasted by multivariate time series models based on historical approximate prices and loads. The future detailed prices were forecasted by univariate time series models. The final forecasted prices are the sum of the predicted approximate prices and detailed prices. This proposed method was applied to forecast the day-ahead electricity prices in California electricity market. The comparisons of forecasting results between the presented method and other methods show that the proposed method can provide more accurate forecasted prices.

关 键 词：小波变换 日前电价预测 电力市场 市场出清电价 

分 类 号：F123.9[经济管理—世界经济] TM73[电气工程—电力系统及自动化]