基于神经网络和混沌特征选择的短期负荷预测方法

Short-term Load Forecasting Based on Neural Algorithm and Chaotic Feature Selection

在激烈竞争的电力市场中,短期负荷预测(Short-term Load Forecasting, STLF)是电力系统高效运行的主要研究热点。针对具有高度波动性非线性信号的电力负载,提出了一种基于神经网络和混沌智能特征选择的预测方法,通过选择最佳候选输入集作为特征参数,将其作为预测输入数据。预测引擎采用一种多层感知层,具有差分进化的学习算法。输入通过的候选者进行相关性分析。对电力市场实际数据进行测试,并与其他STLF技术进行对比,结果表明该方法具有较高的准确度,可在不同电力系统中推广应用。

In the deregulated competition environment of power market, short-term load forecasting(STLF) is a main research point of efficient operation of power system. In view of the power load with highly fluctuating nonlinear signals, a prediction method based on neural network and chaotic intelligent feature selection is proposed in this paper, and the feature selection method is also proposed by selecting the best candidate input set. It is used as the input data for prediction. Candidates enter the correlation between the measured value and the target value by using correlation analysis. The prediction engine is a multi-layer perception layer(MLP) NN, which has the learning algorithm of mixed LM and differential evolution. Finally, it is tested in the electricity market and compared with some latest STLF technologies.