Research On Traction Current Feature Recognition Algorithm For Urban Rail Transit Based On IHGS-VMD And MKRVM

Under the increasingly complex driving conditions of urban rail transit,the safe and stable operation of its power supply system is inseparable from the protection of relay protection equipment.However,affected by the characteristics of the traction network,the DC feeder current of the traction substation will be specific in extreme cases,resulting in the refusal and misoperation of relay protection devices,which poses a great threat to the driving safety of urban rail transit.In view of the above problems,it is necessary to analyze the characteristics of DC feeder current by constructing the simulation model of urban rail transit traction power supply,so as to accurately identify the different states of DC traction feeder current.it is of great significance to improve the safety and stability of urban rail traction power supply.This thesis takes the Tianjin Metro Line 4 as the engineering application background,firstly models and simulates the AC/DC power supply system of urban rail transit.Based on this,in order to address the problem that the existing train model cannot reflect the transient changes of DC feeder current,the permanent magnet synchronous motor is used as the load to refine the train model.With the improved model,two extreme situations of train passing through electrical sections and remote short-circuit faults of the traction network are simulated.The DC feeder current obtained from the simulation is analyzed by Pearson correlation coefficient and current rise rate,clarifying the reasons for protection misoperation and providing support for subsequent research.Secondly,in order to further analyze the characteristics of the charging current of the needle train and the remote short-circuit current,an adaptive current analysis method based on optimized variational mode decomposition(VMD)is proposed.The effect of VMD on the decomposition of DC feeder current under different parameters is analyzed.To address the difficulty in selecting parameters for the existing VMD algorithm,an improved hunger games search algorithm(IHGS)is proposed to optimize the key parameters of VMD using correlation coefficient and energy entropy as fitness functions,so as to achieve the optimal decomposition effect of VMD.Comparative experiments of the algorithm show that the frequency spectrum distribution of each component of the current after IHGS-VMD decomposition is different under different operating conditions,which proves that the method has high adaptability and accuracy.Finally,in order to effectively extract features from the DC feeder current after IHGS-VMD decomposition,a feature extraction method based on sample space relationship is proposed.Based on sample entropy,this method quantifies sample similarity from both distance and spatial direction,and through comparative verification,it is proved that this method can effectively extract the specificity of DC feeder current.Meanwhile,considering the high dimensionality and strong sparsity of the DC feeder signal after feature extraction,a recognition model based on optimized multi-kernel relevance vector machine(MKRVM)is proposed.This model improves the solution accuracy and generalization ability of the model by constructing a mixed kernel function,and optimizes the model parameters using the IHGS algorithm to further improve the recognition accuracy of the model.Compared with other models,this model has excellent performance in both recognition accuracy and speed.Experimental results show that this method significantly improves the accuracy compared to traditional protection fault identification methods,demonstrating the effectiveness of this method.