关键词: 锂电池, 荷电状态, 蝗虫优化, BP神经网络

Abstract:

Accurate estimation of battery state of charge (SOC) is an important guarantee for the safe operation of energy storage devices. This work proposes a joint algorithm based on an improved grasshopper optimization algorithm and a backpropagation neural network (UGOA-BP). The algorithm introduces a uniform distribution function based on the standard grasshopper optimization algorithm, updates the control parameters, and builds a new system of random adjustment. The location update increases the population diversity and makes up for the limitations of the weak global search capability of the locust optimization algorithm. Historical data of energy storage equipment from a new energy company was used. The complete and partial discharge process data set of battery SOC from 100% to 0% and 52% to 49%, respectively, were selected. The proposed model, the grasshopper optimization algorithm BP neural network (GOA-BP), and the traditional BP neural network model were compared, tested, and analyzed from two dimensions. The simulation results show that the absolute errors of the UGOA-BP predicted values are all in the range of [-0.050, 0.050], with -0.046 maximum absolute error and 0.001 average mean square error. The average mean square errors of the GOA-BP model and the BP neural network are 0.009 and 0.067, respectively. Thus, the model prediction accuracy is better than other methods, with reasonable accuracy and engineering application value.

Key words: lithium-ion battery, state of charge, grasshopper optimisation, BP neural network