Traction Transformer Cooling System Fan Research On The Prediction Method Of Filter Clogging

The purpose of the research project on service safety law of EMU electrical system is to determine the failure mechanism and failure mode of key components by means of theory,signal and data analysis,and to further determine the deterioration law and service life of components.Find the optimal maintenance strategy.The traction transformer is one of the key components,and the fan filter of the traction transformer cooling system is the key factor affecting the health status of the traction transformer,and its clean state greatly affects the operation status of the traction transformer.Therefore,the use of advanced data monitoring equipment to obtain data,the use of big data analysis and prediction technology to achieve the predictive cleaning of the filter has become an important issue in the study of the service law of the key components of EMU.Based on the PHM data of the real operation monitoring of a certain type of motor car in 2020,this paper puts forward a scheme for predicting the blockage of the fan filter in the traction transformer cooling system based on data mining,and develops a prototype system for predicting the residual life of the filter,which realizes the real-time residual life prediction and early warning of the filter.Firstly,a fuzzy comprehensive evaluation method optimized by improved combination weighting method is constructed to evaluate the health state of fan filter in traction transformer cooling system,which provides a basis for subsequent deterioration trend discrimination and residual acceptance prediction.After carry on the data analysis,outlier processing,data denoising,time-frequency domain feature construction and local time window feature construction of the original monitoring data,and use random forest algorithm for feature screening to improve the data quality.Then the prediction model is constructed from the two directions of machine learning and deep learning respectively.An improved DE-GWO-SVR machine learning algorithm and a deep fusion model of 1D-CNN-Attention-LSTM are proposed to improve the accuracy of residual life prediction of fan filter in traction transformer cooling system.Finally,a set of residual life prediction system of fan filter in traction transformer cooling system is developed,which transforms the research results of this paper into practical application to realize the residual life prediction and early warning of the filter.The completeness of the function of the development system is verified based on the existing data.