非线性主成分分析和RBF神经网络的电力系统负荷预测

Nonlinear Principal Component Analysis and RBF Neural Network for Power System Load Forecasting

电力系统负荷预测是电力系统规划与运行的重要内容,为提高负荷预测的精度,针对主成分分析法在涉及到多指标预测体系中降维作用不明显,且考虑不到指标间非线性关系的问题,采用非线性主成分分析法改进RBF神经网络输入量,该方法克服了数据之间相关性的约束,进一步降低了预测指标维数,兼顾了指标间非线性关系,保留了原始数据的足够信息,获得电力系统负荷预测的主成分,显著地减少了径向基函数神经网络的输入量,从而提高了电力系统负荷预测的精度。实例分析验证了该方法的有效性。

Load forecasting is essential to the operation and planning of the power system. To improve the load forecasting accuracy and to address the problem that the principal com-ponent analysis method has no obvious effects in reducing the dimension and it takes no consideration the nonlinear relationship between the indexes in the multi-index prediction system,this paper uses the nonlinear principal component analysis to improve the input of RBF neural network. This method over-comes the constraints of the correlation between the data,and reduces the dimension of forecasting indexes,and takes into consideration of the nonlinear relationship between the indexes,and retains sufficient information of the original data, and obtains the principal components of the load forecasting,and remarkably reduces the input of the radial basis function net-work and improves the accuracy of power system load forecas-ting. The effectiveness of the proposed method is validated by results of the case analysis.