EMD-NLSSA-SVM Based Bow Network Contact Resistance Prediction Study

With the development of electric locomotives,the advantages of electrified railways,such as energy efficiency,speed,and safety,are increasingly valued.The good service performance of the bow network system is an important guarantee for the safe operation of locomotives and the quality of the received current.Due to the increase in the operating speed of electric locomotives and the addition of locomotive lines,the unstable contact performance between the conductors of the bow network leads to an increase in the impact vibration of the bow network system,which causes the pantograph slide plate to break,making the locomotive unable to run stably.Therefore,the stability of the electrical contact in the bow net system becomes very important.Contact resistance is an important parameter of electrical contact performance,and its variation can affect the performance of electrical contact as well as the contact state.Previous research on contact resistance usually adopts theoretical or numerical modeling methods,but these methods need to make corresponding simplifications to contact resistance,which makes the accuracy of the model used cannot reach the industrial requirements.In recent years,intelligent predictive modeling has received much attention from scholars,so this paper adopts an intelligent predictive modeling approach to improve the accuracy of bow network contact resistance models and,at the same time,provides a new approach to bow network contact resistance models.This paper presents the first method to combine the Empirical Mode Decomposition(EMD)with a multi-strategy hybrid Improved Sparrow Search Algorithm Optimised Support Vector Machine(NLSSA-SVM)model to predict the contact resistance value of the bow network system.The method not only improves the prediction accuracy of the bow network system contact resistance but also provides a new direction for the establishment of the bow network system contact resistance prediction model.Since Support Vector Machines(SVM)have the advantage of higher prediction accuracy,this paper chooses SVM as the main method for the contact resistance prediction model.However,there is a problem of difficult parameter selection in SVMs,and improper parameter selection will affect the prediction accuracy of the model.Therefore,this paper introduces a multi-strategy hybrid improved sparrow algorithm to find the optimal parameters for SVMs.Firstly,the empirical mode decomposition method is introduced in this paper to decompose the experimentally collected contact resistance data,which is non-linear and non-stationary.The empirical modal decomposition method decomposes the nonlinear contact resistance data into several gradually smooth Intrinsic Mode Functions(IMF)and residual components,which improves the depth of data mining for subsequent prediction;secondly,the Sparrow search algorithm(SSA)is prone to Based on this,this paper uses multiple strategies to improve the sparrow search algorithm,to improve the global exploration ability and the population diversity of the sparrow search algorithm,and verifies that the multi-strategy improved sparrow search algorithm is better than other intelligent optimization algorithms.Then,the optimized support vector machine model is obtained by optimizing the parameters of the support vector machine with the improved sparrow search algorithm,and it is verified through experiments that the optimized support vector machine model with the improved sparrow search algorithm has higher prediction accuracy compared with the unoptimized support vector;finally,the optimized support vector machine is used to build individual prediction models for each eigenmodal function of the empirical modal decomposition,and the individual components after prediction are reconstructed to obtain The final prediction results were obtained,and the empirical modal decomposition combined with the optimized support vector machine model was experimentally verified to have higher prediction accuracy than other prediction models,and the model was found to be feasible and practical for contact resistance prediction.There are 43 figures,9 tables,and 61 references in this thesis.