A data-driven coulomb counting method for state of charge calibration and estimation of lithium-ion battery

Abstract

Due to increasing concerns about global warming, greenhouse gas emissions, and the depletion of fossil fuels, electric vehicles (EVs) powered by lithium-ion batteries have been placed on the forefront as alternative vehicles. Although lithium-ion battery has noticeable features, including high energy and power density, its highly nonlinear and dynamic nature needs to be continuously monitored by an effective battery management system (BMS). Accurate state of charge (SOC) estimation plays an essential role in BMS. The accuracy of coulomb counting method, as one of the simplest methods to estimate SOC, is strongly affected by the sensor accuracy, initial SOC and actual capacity. In the case of ignoring sensor accuracy, a data-driven coulomb counting method is proposed in this paper. Firstly, based on the incremental capacity analysis (ICA), the conventional battery voltage based IC curves are transferred to the SOC based IC curves, where the calibration point can be extracted to correct the erroneous initial SOC. Through the maximum information efficient (MIC) analysis, four voltage values are determined as the input of Gaussian process regression (GPR) model to realize the estimation for actual capacity. Finally, these two calibrated parameters are applied to modify the coulomb counting method. The robustness and feasibility of the proposed method are evaluated using experimental data under fast capacity degradation, which indicates that the data-driven coulomb counting method can calibrate the erroneous parameters and provide on-line satisfactory estimation accuracy for SOC.