Research On Fault Early Warning Of Urban Rail Power Supply System Based On Power Data

The rapid development of urban rail transit industry puts forward high requirements for the reliable operation of urban rail power supply system and key equipment.However,the current equipment operation system is mainly planned repair,it is difficult to find the latent fault of equipment.As an important manifestation of the operation state of the power supply system,power data contains rich system information.Therefore,based on the power monitoring data,this thesis takes the urban rail power supply system and equipment as the research object,and carries out specific research from two perspectives of ’ load forecasting ’ and ’ multidimensional feature similarity measurement ’.The main research contents are as follows:Firstly,different fault early warning methods are analyzed and compared to determine the feasibility of identifying the abnormal operation of urban rail power supply system and equipment through fault prediction and probability statistics.Corresponding data processing methods are adopted for the abnormal manifestations of power monitoring data.Secondly,a load anomaly warning method for urban rail power supply system based on CNN-LSTM load forecasting is studied.Aiming at the problem of large fluctuation and randomness of urban rail power supply system load,a hybrid load forecasting model based on CNN-LSTM is studied to improve the accuracy of load forecasting.The model combines the feature extraction ability of CNN model and the temporal relationship processing ability of LSTM model,which is beneficial to realize high precision load forecasting.In order to further optimize the model,the grid search algorithm is used to optimize the model parameters to obtain the optimal load forecasting model.Compared with the traditional LSTM prediction model,the prediction model is used to predict and visualize the node load of different festivals,and better prediction results are obtained.Based on the prediction results of CNN-LSTM model,a fault warning strategy is designed to realize the fault warning of node load anomaly.Then,a fault warning method of urban rail power supply key equipment based on multidimensional feature similarity measurement is studied.Aiming at the problem that the single-dimensional monitoring data are easy to be submerged in the early fault stage and the early warning accuracy is low,a fault early warning strategy combining Gaussian Mixture Model,similarity measurement method and entropy weight method is studied.This method first builds the Gaussian Mixture Model of the multi-dimensional monitoring data of a single device,optimizes the modeling process by using the EM algorithm,and obtains the multi-dimensional feature sequence that characterizes the operation state of the device.The similarity of different single devices is measured by multidimensional features,and the operation similarity between two devices is obtained.Finally,using the entropy weight method to obtain the comprehensive similarity of single equipment and group equipment.Then,according to judge whether the comprehensive similarity has a trend change,the early warning is achieved.The effectiveness of the proposed method is verified by measured and simulated data.Finally,the research process of two fault early warning methods for urban rail power supply system designed in this thesis is summarized,and the future research direction is pointed out.41 figures,17 tables,52 references.