An improved genetic-back propagation network constructing strategy for high-precision state-of-charge estimation of complex-current-temperature-variation lithium-ion batteries

Environmental issues have driven the booming development of lithium-ion battery technology, and improving the accuracy of state-of-charge (SOC) measurements will play an important role in prolonging battery cycle life and improving safety. In this paper, an improved genetic-back propagation network construction strategy is built to stably and accurately estimate the SOC at different temperatures. The design of the established SOC neural network estimation model is optimized by using a genetic algorithm to initialize the weights and thresholds of the backpropagation (BP) neural network. Then, the above algorithms are validated using Dynamic Stress Test (DST) and Beijing Bus Dynamic Stress Test (BBDST) datasets. The experimental results show that the MAE and RMSE of the SOC estimation results based on the BBDST dataset are 0.0133, 0.0143 and 0.0058, 0.0082 under the conditions of − 5 °C and 35 °C. Therefore, the improved genetic-back propagation network construction algorithm has a more stable iterative process and higher estimation accuracy, which provides a new method for SOC performance estimation of lithium-ion batteries.