Research And Realization Of Health Evaluation For EMU Battery

China's high-speed railway is developing rapidly,and people pay more and more attention to the efficient operation and safety of EMUs.In the safe operation of the EMU,the battery is a key component of the EMU.Whether the state is normal is the cornerstone of the safe operation of the EMU.Although the planned repair of the existing battery can guarantee the safe operation of the high-speed railway,there will be an over-repair problem during the maintenance process,which will cause an increase in the maintenance cost and even reduce the efficiency of the high-speed railway transportation.With the increasing density of high-speed EMUs,there is an urgent need to change the maintenance program from "planned repairs" to "state repairs",that is,to conduct battery status assessments,predict battery repair dates,reduce maintenance costs,and improve EMU operation and maintenance effectiveness.At present,the existing battery health assessment models at home and abroad still have the problem of low accuracy.Therefore,based on the data-driven method,this paper proposes an optimized model for evaluating the health status of the battery.The current capacity of the battery and the remaining service life of the battery are studied by analyzing the relevant historical data during the operation of the battery.The main research work is as follows:(1)Aiming at the problem of simply using the current,voltage,temperature and other basic characteristics of the battery in the current capacity degradation model.This paper combines the direct characteristics of the battery with the indirect characteristics of the battery discharge time,equal pressure drop discharge time.It is proposed to use Kendall Rank correlation coefficient and Spearman correlation coefficient to analyze the correlation between battery characteristics and capacity,and select the appropriate battery characteristics according to the correlation level to form a sample set to complete the model training and testing.(2)Put forward the GARF battery capacity estimation model.The generation of decision trees in the random forest algorithm and the number of selected decision trees both affect the computational cost of the model.The genetic algorithm is used to optimize the parameter combination of the number of features and the number of decision trees to reduce the calculation cost of the model and improve the model's estimation Precision.Finally,use the EMU battery data to verify the feasibility of theoptimized model.(3)The attention mechanism is used to optimize the Bi GRU neural network,and a Bi GRU-Attention battery life prediction model with higher prediction accuracy is proposed.The two-way GRU changes the internal structure of the traditional GRU,using both positive and negative directions for propagation,and uses past and future data to predict the current capacity of the battery at the same time,which further reduces the prediction error.The introduction of the attention mechanism is to assign different attention to the input sequence,so that the model focuses more on learning the key parts,and the training model handles the learning ability of the time series prediction model,thereby obtaining a more accurate battery life prediction model.Finally,based on the Bi GRU-Attention battery life prediction model,the data of the EMU battery is tested to verify that the prediction accuracy of the model has indeed improved.