Research On Status Monitoring Of Lithium-ion Batteries Under Multi-time Scale

Lithium-ion batteries are widely used in mobiles,computers,electric vehicles,microgrid,and other mass consumer products because of their unique advantages of high energy density,low production cost,and long cycle life.However,they also have temperature-sensitive,highly nonlinear,and aging characteristics,which make the state monitoring and safety control of lithium-ion battery management system challenging.Accurate and reliable state monitoring is of great significance to ensure the safe,reliable,and durable,and efficient operation of lithiumion batteries.From the point of view of multi-time scale,it is of great significance to study the theory and method of state monitoring of lithium-ion battery and to realize accurate and reliable monitoring of battery state,which is of great significance to ensure the safe,reliable,and lasting and efficient operation of a lithium-ion battery.To realize the multi-time monitoring of battery status,this paper studies the two key technologies of charge state estimation at micro-time scale and remaining useful life prediction at macro-time scale,the specific studies include:Because of the problem that the inaccuracy of the battery model leads to a decrease in the accuracy of charge state estimation in a microscopic time scale.Firstly,through the analysis of dynamic electrical behavior and the equivalent circuit model of the lithium-ion battery,two core elements of model parameter identification and open-circuit voltage curve are determined to ensure the accuracy of the lithium-ion battery model.Secondly,the second-order equivalent circuit model parameter identification formula based on the recursive least square method with forgetting factor is derived to realize the off-line identification of model parameters.Finally,an open circuit voltage curve fusion method based on incremental test and the low current test is proposed to improve the accuracy of the open-circuit voltage curve and to describe the electrical dynamic behavior of lithium-ion battery accurately.It provides a model basis for improving the accuracy of charge state estimation.Aiming at the problem that noise uncertainty and system nonlinearity lead to the decrease of estimation accuracy in charge state estimation at micro time scale,an on-line estimation algorithm for the charged state of lithium-ion battery based on neural network state estimator is designed.Firstly,to solve the problem of network structure error in traditional neural network state estimator,the adaptive neural network algorithm is studied,and the state estimator based on GMDH neural network is proposed;Secondly,the evaluation criterion of hidden layer neurons and the optimization method of network structure based on principal component analysis are designed to solve the problem of out-of-control of GMDH neural network.The proposed method can effectively reduce the influence of process noise on the estimation results and realize high precision state estimation of charge.Aiming at the problem that the small amount of data available in the early used batteries and the lack of capacity degradation characteristics lead to the difficulty of realizing the prediction of the remaining useful life in the macroscopic time scale,a method for constructing discharge capacity difference curve is proposed.The data features for remaining useful life prediction are extracted.Based on the feature extraction and ridge regression algorithm,the data-driven residual service life prediction model is established,and the parameters of the ridge regression algorithm are optimized by cross-validation and learning curve.The proposed datadriven model only uses the early cycle period data of the battery to achieve high accuracy of the remaining useful life prediction.Aiming at the problem that the capacity regeneration leads to the decrease of the prediction accuracy of remaining useful life in macroscopic time scale,by analyzing the capacity degradation behavior of lithium-ion battery with capacity regeneration phenomenon,a multimodel prediction algorithm is proposed.Firstly,the degradation behavior of battery performance is described by capacity degradation process and capacity regeneration process;Secondly,a hybrid iterative prediction model based on correlation vector machine and grayscale model is established to predict the continuous degradation process of capacity;The capacity regeneration model based on exponential function is established to predict the capacity regeneration process;Finally,the iterative prediction strategy based on the above two models is designed to describe the degradation process of a battery health state by using various models.The proposed method can accurately describe the phenomenon of capacity regeneration and improve the long-term prediction accuracy of the remaining useful life.