【摘要】 由于传统径向基函数(radial basis function, RBF)神经网络容易陷入某个局部最小值，会导致网络的性能受到限制，发生预测结果误差较大和预测结果不准确等问题。为此，提出一种改进RBF神经网络的1 000 kV特高压输电线路损耗预测方法。通过斜率灰色分析法筛选气候特征参数，并组成影响特高压输电线线损特征体系，利用遗传算法改进RBF神经网络参数对预测模型进行训练，实现对特高压输电线路的损耗预测。测试结果表明：所提方法在四个区域的预测误差值较其他方法更低，对输电线路损耗预测具有更高的准确性和可靠性。更多还原

【Abstract】 Due to the tendency of traditional radial basis function(RBF) neural network to fall into a local minimum, the performance of the network is limited, resulting in large prediction errors and inaccurate prediction results. Therefore, a loss prediction method for 1 000 kV ultra-high voltage(UHV)transmission lines using an improved RBF neural network was proposed. By using the slope grey analysis method to screen climate characteristic parameters and form a system that affected the line loss characteristics of UHV voltage transmission line, the genetic algorithm was used to improve the RBF neural network parameters for training the prediction model, achieving loss prediction of UHV transmission line. The test results show that the proposed method has lower prediction error values compared with other methods in four regions, and has higher accuracy and reliability in predicting transmission line losses.更多还原