Abnormal Battery Location Recognition and State Estimation in Lithium Battery Pack

There are many problems in the abnormal diagnosis of the lithium battery pack, such as incomplete research structure, insufficient positioning accuracy of abnormal batteries, and inadequate combination of diagnosis and treatment. To deal with these problems, this paper systematically achieves the goal of precise positioning, state estimation, and decision-making processing of abnormal batteries in a complete series-parallel battery pack. It also provides effective basic methods and exploration ideas for lithium battery energy storage systems to achieve intelligent diagnosis, intelligent prediction and precise positioning. Firstly, the attenuation model of the single battery is built from the perspective of the three coupling characteristics of heat, electricity and attenuation, and then build a 3*3 small battery pack model. Then study the impact of different degrees of attenuation of batteries in different positions in the battery pack on the overall thermal and electrical characteristics of the battery pack, and compare it with the normal battery pack to analyze and extract the characteristic variables that characterize the attenuation battery. Next, based on the LS_SVM and BP neural network algorithm, the feature classifier and regression feature model are established respectively to locate and estimate the attenuation rate of abnormally attenuated batteries, and conduct simulation tests. The simulation results prove that all the accuracy and effectiveness of establishing the classifier and regression model are above 95%. Finally, relevant solutions are proposed for abnormally attenuated batteries.