基于DEEDA-CWLS-SVM的弓网电接触系统模型预测研究

Research on Model Prediction of Pantograph Catenary Electric Contact System Based on DEEDA-CWLS-SVM

在弓网电接触系统中,回路电流、运行速度、接触压力与载流稳定系数、电流效率、磨损率之间存在一定的函数关系,且该函数关系为多输入-多输出关系。因此,通过自行研制的实验机模拟弓网电接触系统的实际运行,提出一种基于柯西分布加权的最小二乘支持向量机(CWLS-SVM)进行弓网电接触系统的输入输出模型预测,考虑到CWLS-SVM模型的正则化参数C与高斯核参数σ难以选取的问题,采用差分进化分布估计(DEEDA)算法对以上参数进行寻优,利用实验数据进行预测模型的训练与检验。结果表明,采用本文算法建立的预测模型具有更高的预测精度,泛化能力强,为弓网电接触系统模型建立提供理论支持。

In pantograph catenary electric contact system,there is a certain function existed among loop current,running speed,contact pressure and carrier stability factor,current efficiency and wear rate,and the function relation is a multi-input- multi-output relationship.Therefore,the actual operation of pantograph catenary electric contact system is simulated by self-developed testing machine.A least squares support vector machine based on Cauchy distribution weighted（CWLS-SVM） is proposed to predict the input and output of the model of pantograph catenary system.To solve the problems that regularization parameter and Gaussian kernel parameter are difficult to select,estimation of distribution differential evolution（DEEDA） algorithm is adopted to optimize these parameters,and the experimental data are used to train and test the forecasting model.The results show that the prediction model using the algorithm has higher prediction accuracy,generalization ability,provides theoretical support for the establishment of pantograph catenary electric contact system model.