Abstract:

A multi-model adaptive Kalman filter algorithm was designed to capture battery characteristics and estimate state of charge (SOC) accurately. This method obtained a variety of characteristics of the battery through an improved hybrid pulse power characteristic (HPPC) test, and established two dynamic models to
describe the temperature characteristics and the rate characteristics of the battery respectively, and used a fusion algorithm based on conditional probability to combine different dynamic characteristics information within the battery. The results show that the average estimation error is less than 1% under dynamic stress testing conditions. The multi-model filter estimation method integrates the different dynamic characteristics of the battery, which can better explain the nonlinear characteristics of the battery in complex environment, maintaining high accuracy over the entire charge-discharge interval and the more complex usage conditions, also improving the accuracy and robustness of the SOC estimation.

Key words: lithium-ion battery , state of charge (SOC) , temperature characteristic , rate characteristic ,  multiplemodel , adaptive Kalman filter