Abstract: The SOC is one of the important parameters in battery management system. In order to improve the accuracy of the ampere-hour integral method for estimation of SOC and solve the problem that estimation error increases with time, an error prediction model of ampere-hour integration method is established in this paper by using Extreme Learning Machine Algorithm. The model takes working current of battery as input and the corresponding ampere-hour integration method SOC estimation error as output. The error prediction model is integrated with the ampere-hour integration method to correct estimated value of the SOC of the ampere-hour integration method. An on-line estimation method of SOC of lithium-ion battery based on Ampere-hour integration method and Extreme Learning Machine Algorithm is therefore developed. Results from our simulation show that, compared with the ampere-time integration method, the integrated method can effectively reduce estimation errors of the SOC and overcome the problem that the estimation errors resulting from the ampere-time integration method increases with time.

Key words: SOC, ampere-hour integral method, extreme learning machine, error correction