Research On Remaining Useful Life Prediction Of The Lithium Ion Battery

The battery has been widely used in both military and civilian fields.Estimating state of health(SOH)and predicting remaining useful life(RUL)exactly have a great significance to improve the safety and extend the useful life of battery.In this paper,we took the lithium ion battery as the research object and investigated the SOH estimation methods and RUL prediction methods.The specific contents are as follows:1)We introduced the structure and working principle of lithium-ion battery,clarify the basic concept of the commonly used performance parameters for lithium ion batteries and characteristics.The effect of discharge current,the use of ambient temperature and the number of cycles on aging of lithium-ion battery is also analyzed.2)At present,we can only collect battery voltage,charge and discharge current and ambient temperature and other measurable parameters through the sensor in the practical application.It is difficult to calibrate and update the parameters of the battery model using these data.In this paper,a method of estimating the online SOH of the battery based on BP neural network and double square root cubature Kalman filter(DSRCKF)algorithm is studied.Firstly,the terminal voltage and discharge current data of the battery are collected as the training data.Then,the battery model is trained by the BP neural network algorithm.Next,the battery model obtained by offline training is embedded in the double square root cubature Kalman filter(DSRCKF)algorithm.The two Kalman filters estimate the battery SOC and SOH while updating the parameters of the neural network model.Finally,NASA data is used to validate the method.The experimental results show that the BP neural network model can be updated on-line and the robustness of the model can be improved,the on-line estimation of SOC and SOH can be also realized.As the BP neural network model can be updated,the battery model can accommodate the current battery usage environment,which greatly improved the SOC and SOH prediction accuracy.3)Due to processing technology,design error,functional differences and the use of environmental factors,batteries of the same type are usually exist differences between different individuals.So this paper studied the reliability assessment method based on linear Wiener degradation process and the remaining useful life(RUL)prediction method based on Wiener degradation process of random effects.Firstly,we use the linear Wiener process to build the model,and the model parameters are estimatedbased on the maximum likelihood method.Then,the remaining useful life of the battery is calculated by the life reliability function given by the linear Wiener process model.However,in order to solve the problem of inconsistency between batteries of the same type,this paper use drift parameters in the Wiener process model to characterize the influence of stochastic factors on the battery.Then,a random effect Wiener degenerate process model is established and model parameters is estimated by EM algorithm.Next,to get the battery remaining useful life(RUL),we use the remaining useful life(RUL)distribution function.Finally,we use NASA experimental data and the data which is obtained from the independent experimental platform to verify and compare these two methods.The experimental results show that the remaining useful life(RUL)prediction method based on the random effect Wiener degradation process can better predict the remaining useful life(RUL)of the battery,and with the increase of training data,the prediction accuracy is higher.