改进F-score特征选择的MPSO-BP神经网络短期负荷预测

Short-term load forecasting of MPSO-BP neural network based on improved F-score feature selection

针对海量数据提出一种基于改进Fisher分数(F-score)特征选择的改进粒子群优化的BP(Modified Particle Swarm Optimization and Back Propagation,MPSO-BP)神经网络短期负荷预测方法。首先采用改进F-score特征评价准则计算影响负荷预测精度各个特征的F-score值,再通过F-score Area法设定阈值筛选出最优特征子集,然后将最优特征子集作为MPSO-BP神经网络模型的输入变量完成对预测日一天24点负荷的预测,并与MPSO-BP神经网络短期负荷预测和传统BP神经网络短期负荷预测进行对比。算例表明,文中提出的短期负荷预测方法可以较好地对海量数据进行挖掘,具有较高的预测精度。

In this paper,a short-term load forecasting method of MPSO-BP（ modified particle swarm optimization and back propagation） neural network based on improved F-score feature selection is proposed for mass data. Firstly,the improved F-score feature evaluation criterion is used to calculate the F-score value of each feature affecting the load forecasting accuracy,then,the optimal feature subset was selected by the F-score area method. The forecast of 24-point load is completed by using the optimal feature subset as input variable of the MPSO-BP neural network model,as well as comparing with the short-term load forecasting of MPSO-BP neural network and he traditional BP neural network short-term load forecasting. The calculation example shows that the short-term load forecasting method proposed in this paper can be used to excavate the massive data,which has higher prediction accuracy.