A novel multiple training-scale dynamic adaptive cuckoo search optimized long short-term memory neural network and multi-dimensional health indicators acquisition strategy for whole life cycle health evaluation of lithium-ion batteries

Abstract

State of health evaluation of lithium-ion batteries has become a significant research direction in related fields attributed to the crucial impact on the reliability and safety of electric vehicles. In this research, a dynamic adaptive cuckoo search optimized long short-term memory neural network algorithm is proposed. The aging mechanism of the battery is described effectively by extracting and selecting high correlation health indicators including voltage, current, charging time, etc. A dynamic adaptive strategy is introduced to the cuckoo search algorithm to stabilize the step size and improve the global search ability. The hyperparameter optimization and noise filtering problems of the long short-term memory model are solved and the accuracy of the algorithm is improved by taking advantage of the established dynamic adaptive cuckoo search algorithm. The accuracy and effectiveness of the proposed method are verified based on the seven groups of battery aging datasets from the National Aeronautics and Space Administration and the University of Maryland. Compared with the long short-term memory and convolutional neural network long short-term memory, the mean absolute error of the results obtained by the proposed algorithm is kept under 2%, the root mean square error is less than 3%, and the average absolute percentage error is less than 3%. The results indicate the algorithm has better fitting performance, stronger robustness, and generality.