基于神经网络与混沌理论的电力系统短期负荷预测

Short-term Load Forecasting Based on Chaos Theory and Neural Network in Power System

短期负荷预测是电力调度部门的重要工作之一,负荷预测的精度直接影响到电网的安全、经济和稳定运行。本文针对目前负荷预测中单一预测理论精度较低的问题提出采用BP神经网络与混沌理论相结合的算法,以变步长和附加动量法进行改进,同时以混沌时间序列来确定网络结构,从而克服了算法对大量训练样本的依赖,提高预测精度和速度。对咸阳区域电网负荷的实际预测结果表明了该方法的有效性。

Short-term load forecasting is an important work for electric power dispatching center. The accuracy of load forecasting has effected safe, economic and stable operation of network. In view of low accuracy of current load forecasting with single forecasting theory, the method of chaos theory combined with neural network in short-term load forecasting is put forward, and is improved by variable step size and additional moment method. Chaotic time series is used to determine network structure, and overcomes the dependences on large amount of samples. The accuracy and speed of forecasting is increased. The load forecasting in Xianyang power network shows the effectiveness of this method.