时间序列分解在短期电价分析与预测中的应用  

Application of Time Series Decomposition to Short-term Electricity Price Analysis and Forecasting

作　　者：熊高峰[1] 韩鹏[1] 聂坤凯[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《电力系统及其自动化学报》2011年第3期95-100,共6页Proceedings of the CSU-EPSA

摘　　要：为了提高短期电价预测的精度,将电价分解成工作日电价和周末电价两个时间序列,并且,通过移动平均法和离散傅立叶变换,分别将这两个时间序列分成趋势分量、周期分量和随机分量三个组成部分,然后,分别采用移动平均法、外推法和最小二乘支持向量机对这三个组成部分进行预测以求得两个电价时间序列未来的预测值。仿真结果表明,与采用传统BP神经网络法的预测方法相比,该方法具有较高的预测精度,其预测绝对百分比误差在7%以内。To improve the accuracy of short--term electricity price forecasting, original time series of electricity prices are firstly decomposed into weekday series and weekend series. And then, by using moving average method and discrete Fourier transform these two series are further divided into three components： trend corn ponent, periodic component and random component. To obtain the forecasted values of these two time series of electricity prices, these three components are forecasted using moving average method, extrapolation method and least square support vector machine, respectively. Simulation results show that, compared with traditional BP neural network method, the proposed method in this paper has higher prediction accuracy, and the mean absolute percentage errors are within 7%.

关 键 词：电价预测 最小二乘支持向量机 时间序列分解 移动平均法 离散傅里叶变换 

分 类 号：TM73[电气工程—电力系统及自动化]