Battery identification based on real-world data

Abstract

In this paper, system identification is carried out for a 20 kWh battery using real-world measurements data. State-of-charge (SOC) and the open-circuit voltage (Voc) relationship will be obtained using least square estimation (LSE) non-linear regression. In addition, how to estimate SOC using current measurements and how to estimate the equivalent circuit's RC parameters are carried out using autoregressive exogenous (ARX) models. The respective ARX models are first derived. Estimation of the ARX coefficients is then carried out. Finally, parameter recovery is conducted to find out parameters with physical meanings, e.g., RC values. With the identified Voc and SOC relationship and RC parameters, we built a simulation model in MATLAB/Simpowersystems. With the measured current data from the real-world as the input, the simulation model gives the terminal DC voltage as the output. This output is compared with the real-world DC voltage measurements data and the matching degree is satisfactory.