Estimation of State of Charge and State of Health of Marine Lithium Battery based on BP Neural Network

Abstract

For the sake of better ensure the safety of a marine system reliability, economy, and power, on the basis of marine lithium battery discharge data, BP neural network is able to grasp the state of charge and health in this battery module. Owing to some stochastic factors of BP neural network weights and limens, the error fluctuates greatly, so a genetic algorithm is lead into improve its results. The improved GA-BP algorithm reduces the error of SOC estimation from 4.2% to 2.3% and of SOH from 0.52% to 0.24%. It indicates that the GA-BP neural network method for forecasting SOC and SOH of the marine battery management system has minor errors and high stability, which is feasible.