Research On Rapid State Estimation And Detection For Lithium-ion Power Battery

With the gradual popularity of electric vehicles(EVs),the demand for lithium-ion power batteries(Li PB)continues to rise,and the scale of decommissioned Li PB will also increase substantially.The large-scale retired Li PB will be an unavoidable problem in the development of new energy vehicles.For large-scale decommissioned Li PB,if they cannot be recycled or reused in time,serious environmental pollution and waste of resources will be caused.Therefore,in order to better promote the echelon utilization and rapid recovery of retired EVs power batteries,it is necessary to increase research on key technologies such as battery function status,health status,and fault diagnosis mechanisms.Only through on-site rapid battery status detection can the battery recycling process and echelon utilization be accelerated,thereby effectively solving the environmental pressure caused by the large-scale decommissioned Li PB.In response to the above-mentioned problems,the research carried out in this paper is as follows:(1)Rapid estimation of battery So C.The state of charge remaining(So C)estimation is essential for the battery management system(BMS).Accurate So C estimation can improve battery utilization efficiency,especially for EVs.This paper explores a simple and effective method for So C estimation of various Li PB,and proposes a recurrent neural network(RNN)model based on the gated recurrent unit(GRU)architecture to estimate the battery So C.By combining the integrated optimization method of Nadam and Ada Max optimizers,GRU-RNN can quickly learn its own model parameters.The Nadam optimizer is used in the model pre-training stage to find the minimum optimization value as soon as possible,and then the Ada Max optimizer is used in the model fine-tuning stage to further determine the model parameters.In order to verify the effectiveness and accuracy of the method,three dynamic driving condition data sets are used to train and test the GRU-RNN model,and compare with existing So C estimation methods.(2)Rapid estimation of battery So H.The state of health(So H)is estimated to be another key indicator of the BMS.Accurate So H estimation can be used to guide the timely recycling and cascading utilization of lithium-ion batteries,which is particularly beneficial to environmental protection.This paper proposes a least squares support vector regression(LS-SVR)model using radial basis kernel functions to estimate battery So H.Based on the hysteresis behavior of lithium-ion batteries,the sample data can be quickly obtained by testing the aging battery through HPPC method.Grey relational analysis(GRA)is used to select the features of the data samples,and then the sample data is used to train the LS-SVR model,and finally K-fold cross-validation is used to determine the hyperparameters of the LS-SVR model.In order to verify the proposed method,the LS-SVR model was trained and tested using Li PB samples with different degrees of aging,and compared with the existing So H estimation method.(3)Rapid estimation of battery OCV.Accurate Open circuit voltage(OCV)estimation is beneficial to the estimation of the So C and So H for the power battery.This paper proposes a ε-type support vector regression(ε-SVR)model for battery OCV estimation.According to the voltage relaxation behavior of lithium-ion batteries,data samples are collected through HPPC for test batteries with different aging degrees.GRA is used to select features of data samples,and K-fold cross-validation is used to obtain the hyperparameters of the ε-SVR model.In order to verify the proposed method,the ε-SVR model was trained and tested on lithium-ion batteries with different aging conditions.(4)Quick prediction of battery ISC state.The prediction of internal short circuit(ISC)is one of the key challenges of the BMS.Accurate battery ISC prediction can effectively reduce the risk of thermal runaway to ensure the safe use of power batteries.There are some researches on battery ISC prediction methods,but there is a lack of an easy-to-use on-site rapid measurement method.Therefore,this paper proposes a random forest classifier(RFC)model for battery ISC prediction.According to the relaxation behavior of lithium-ion batteries,data samples of ordinary batteries and batteries with ISC are measured by HPPC testing.The Matlab curve fitting tool is used to fit the relaxation curve to obtain the parameters of the battery ECM and use them as sample features.GRA is used to select the features of data samples,and grid search(GS)and ‘‘Out-of-Bag’’(Oo B)error are adopted to obtain the parameters of the RFC model.(5)Research and development of battery tester.The Arbin battery tester produced in the United States and the battery tester produced by Newwell in China are expensive and bulky.They are suitable for laboratory measurement and calibration experiments,but are not suitable for rapid on-site battery status detection.Therefore,this article introduces a rapid battery status tester based on MCU,which provides a guarantee for rapid status test for the recovery and cascade utilization of large-scale decommissioned lithium batteries.Secondly,according to the electrochemical characteristics and mechanism of lithium batteries,the factors representing the health status of the battery transfer process and the health status factors of the battery diffusion process are derived.Third,based on the HRPWM technology,the charging control and battery impedance measurement are realized simultaneously,and the function of the battery tester is verified by the impedance measurement experiment results at key frequency points.