基于小波包分析的电力负荷预测算法

提出基于小波包分解和重构的电力负荷预测算法.算法使用具有线性相位的双正交小波对电力负荷数据进行小波包分解和重构,然后用神经网络直接对各尺度上的电力负荷分量进行预测,最后将各尺度上的预测值相加,得到实际负荷预测值.算例表明算法具有较高的预测精度,优于传统的BP神经网络,有利于分析不同时频区域的电力负荷特性,为更准确地建模和预测提供了条件.

Power Load Forecasting Algorithm Based on Wavelet Packet Analysis

This paper investigated the application of wavelet packet ananlysis in power system load forecasting,and proposed a power load forecasting algorithm based on wavelet packet decomposition and reconstruction.The algorithm used the biorthogonal wavelet which has linear phase to decompose and reconstruct the power load data,then used neural network to predict the load series in each scale.Finally,the load forecasting values of each scale are summed up to produce the load forcasting result.Case study showed that the new algorithm improves forecasting accuracy and is superiror to traditional back-propagation neural network.Furthermore,it is useful for analyzing the load characteristics in each time-frequency zone,thus it provides necessary information for more accurate modelling and forecasting.