Improvement of SOC Estimation of Lithium Ion Batteries Considering Sensor Signal Error

The state-of-charge estimation of lithium-ion battery is affected by sensor signals. In this paper, neural networks are used to judge the error pattern, and the state update process of the extended Kalman filter method is improved using the judgement. The results show that for the recognition of voltage signal error, the neural networks used have a good effect, with validation accuracy of more than 70%. Through the method proposed in this paper, the maximum estimation error of the state of charge is reduced by about 1% under the designed voltage signal error.