A framework for state-of-charge and remaining discharge time prediction using unscented particle filter

As a typical complex system, the lithium-ion battery system is characterized by strong coupling and nonlinearity, which brings great challenges to its modeling, state estimation, and control. The modeling and state estimation especially the state-of-charge and remaining discharge time are key issues for the battery management system. This paper details a framework for observation of the battery state-of-charge and remaining discharge time by using the unscented particle filter. First, an equivalent circuit model considering hysteresis is presented and verified at different temperatures. Then the framework for observation of the battery state-of-charge and remaining discharge time is proposed using the unscented particle filter in order to improve the observation accuracy. The recursive method is employed to predict the probable future current considering the historical data. In addition, the prediction results of the probable future current with different forgetting factors are compared and analyzed in order to select the optimal parameter for the remaining discharge time prediction. Finally, experiments under different dynamic driving cycles at different temperatures are carried out to verify the proposed method. The performance of the unscented particle filter and the extended Kalman filter are compared and analyzed. The experimental results indicate that the proposed unscented particle filter method has high accuracy and fast convergence under dynamic driving cycles.