Prediction Of Remaining Useful Life Of Lithium Battery Based On Data Preprocessing Technology-Support Vector Machine

In order to reduce vehicle safety and reliability problems caused by power system failures,the battery management system is used for online monitoring and control of power batteries,providing decision-making reference and predictive maintenance information for system management and control.Remaining life prediction is an important part of the battery management system.As the prediction of the remaining useful life is more complicated,there are still some problems with the prediction model established by the current method.This paper combines data preprocessing technology and support vector machine regression method,based on GB/T 31484-2015 life test to obtain the number of discharges-battery capacity data,and proposes a research method of lithium battery remaining useful life prediction model based on preprocessing data-support vector machine.The main content of the paper is as follows:(1)Firstly,design the experimental plan and carry out the corresponding test according to the standard cycle life test method of GB/T 31484-2015,and the cycle times-battery capacity retention rate curve of lithium iron phosphate batteries and ternary lithium batteries are obtained and analyzed.(2)Secondly,the experimental data is pre-processed for data noise reduction.The noise reduction method used is a multi-resolution wavelet noise reduction method,in which the model parameters in the noise reduction model are obtained by optimization methods,and the objective function in the optimization problem is obtained by cross-check methods The optimization problem is solved by genetic algorithm,and the established noise reduction model is used to reduce the noise of the lithium battery cycle life data.(3)Then iterative multi-output time series prediction method is used to correct and pre-process data with large errors after noise reduction.Among them,the learning model in the time series prediction model uses the support vector machine regression model,and the parameters in the support vector machine regression model are still obtained through an optimization method,in which the objective function is obtained through a rolling origin evaluation method,and the difference between the predicted value and the experimental value at both ends of the correction area is taken as a priori knowledge as the constraint of the optimization problem,and the optimization problem is solved by particle swarm optimization algorithm.The established time series prediction model is used to correct the lithium ion battery life data.(4)Finally,a support vector machine life prediction model is established.By comparing the life prediction model of the support vector machine based on the noise reduction and correction preprocessing data,the life prediction model of the support vector machine based on the original data and the life prediction of the support vector machine based on the noise reduction preprocessing data,the validity of the SVM life prediction model based on the noise reduction and correction preprocessed data is verified.