SOC Estimation of Battery Pack Considering Cell Inconsistency

Range anxiety problem has always been one of the biggest concern of consumers for pure electric vehicles. Accurate driving range prediction is based on accurate lithium-ion battery pack SOC (State of Charge) estimation. In this article, a complete SOC estimation algorithm is proposed from cell level to battery pack level. To begin with, the equivalent circuit model (ECM) is applied as the model of battery cell. ECM parameters are identified every 10% SOC interval through genetic algorithm. The dual extended Kalman filtering (DEKF) algorithm is adopted for cell-level SOC and ohmic resistance R₀ estimation. The estimation accuracy of cell SOC and R₀ is verified under NEDC dynamic working condition. The cell-level SOC estimation error is below 1%. However, cell inconsistency can always result in inaccurate cell SOC estimation inside the battery pack. The impact of initial SOC inconsistency and internal resistance inconsistency between cells on battery pack SOC is specifically analyzed. Considering cell inconsistency, dual time-scale dual extended Kalman filtering (DTSDEKF) algorithm based on “Specific Cell and Difference Model” (“S&D model”) is introduced to help accurately estimate cell SOC inside the battery pack: In the first time scale, DEKF algorithm is used to estimate current SOC and R₀ of “Specific Cell”. In the second time scale, the correction of the cell inconsistency is added, which involves the SOC difference between the remaining cells and the “Specific Cell”, as well as R₀ of all cells. Finally, the DTSDEKF algorithm based on “S&D model” is verified through NEDC dynamic working condition. The SOC estimation error of each cell inside the battery pack is below 1%.