利用神经网络PRPE算法进行短期负荷预报

Parallel Recursive Prediction Error Algorithm for Short-term Load Forecasting

本文根据神经网络的并行递推预估误差训练理论,建立了一种负荷预报的新算法——PRPE预报算法。该算法属利用多层前传网进行负荷预报的范畴,由于该算法有效的权重训练方式,使其收敛速度大大高于传统的BP算法。尽管该算法的训练过程比较复杂,但对于时段负荷预报这类单输出系统,总体训练时间较BP算法成倍减少,且预报精度较高。可用于在线短期负荷预报。

This paper introduces a new parallel recursive prediction error algorithm for short -term load forecasting. This algorithm enables the weight in each neural of the network to be updated in an efficient parallel manner. Exmples demonstrate that it has batter convergence properties than the classical back propagation algorithm, and is a batter kind of artificial neural network for on-line short -term load forecasting.