关键词: 动态主元分析, 数据拟合, BP神经网络, 负荷预测, 电力系统

Abstract: The result of the load forecasting based on PCA neural network cannot meet the accuracy required with the autocorrelation between data being ignored. An augmented matrix is therefore given, of which the number of dimensions is determined by adding the observed data for l time past together, and the forecast data after augmentation are taken as the input variables in BP neural network. Then, a forecasting model is developed on the basis of DPCA-BP neural network, with its architecture described. The model can exclude efficiently the data irrelevant to dependent variables from the independent variable system so as to intensity the autocorrelatability between data of the independent variable system. Compared the two methods of load forecasting with each other, the exemplification results show that the load forecasting model we developed is better than that based on PCA-BP neural network, since it provides higher interpretability and forecasting accuracy.