基于改进RBFNN的1 000 kV特高压线损预测

针对特高压输电线线损与特征参数间关系复杂的特点，提出一种联合聚类优化算法（Canopy -K- means）和自适应二次变异差分进化（adaptive second mutation differential evolution，ASMDE）算法改进的径向基神经网络（radial basis function neural network，RBFNN）模型，用于特高压输电线线损的预测。通过理论分析确定特高压输电线线损的特征参数，采用Canopy-K-means聚类算法进行聚类，以此确定径向基（radial basis function，RBF）神经网络的隐藏层节点，从而确保RBF神经网络具有较优的隐藏层中心。用特征参数和线损的样本数据训练ASMDE算法优化的RBF神经网络，拟合出线损与特征参数之间复杂的非线性关系。以华中地区某特高压输电线路的历史数据为例，仿真验证了所提方法的实用性和有效性。

Prediction of 1 000 kV UHV Line Loss Based on Improved RBFNN

In view of the complex relationship between UHV transmission line loss and its characteristic parameters, this paper proposes a radial basis function neural network (RBFNN) model improved by use of the Canopy-K-means clustering algorithm and the adaptive second mutation differential evolution (ASMDE) algorithm to predict the UHV transmission line loss. The characteristic parameters of UHV transmission line loss determined by theoretical analysis are clustered through Canopy-K-means clustering algorithm to determine the hidden layer nodes of radial basis function (RBF) neural network, subsequently ensuring the RBF neural network to have a better hidden layer center. The RBF neural network optimized by ASMDE algorithm is trained with the sample data of characteristic parameters and line loss, so as to fit the complex nonlinear relationship between line loss and characteristic parameters. Finally, the historical data of a UHV transmission line in Central China is taken for simulation, and the results have verified the practicability and effectiveness of the proposed method.