A hybrid algorithm for the state of energy estimation of lithium-ion batteries based on improved adaptive-forgotten-factor recursive least squares and particle swarm optimized unscented particle filtering

State of energy (SOE) estimation of lithium-ion batteries is the basis of electric vehicle driving range prediction. To improve the estimation accuracy of SOE under complex dynamic working conditions, this paper takes the ternary lithium-ion battery as the research object, chooses the second-order RC-PNGV model to model the polarization reaction inside the battery, and adopts the improved adaptive forgetting factor recursive least squares (MAFFRLS) method to identify the model parameters. For battery SOE estimation, an improved unscented particle filtering algorithm for particle swarm optimization is proposed, which introduces quantum theory into particle swarm optimization to solve the sub-depletion problem of unscented particle filtering and improves the accuracy and adaptability of real-time estimation of SOE in complex environments. Experimental validation is carried out by constructing different working conditions at multiple temperatures, and the results show that the maximum error of parameter identification using recursive least squares based on improved adaptive-forgotten factor is stabilized within 2%. Under the HPPC, BBDST, and DST working conditions, the MAE and RMSE are limited to within 1% when the quantum particle swarm optimized-unscented particle filtering (QPSO-UPF) algorithm is applied to estimate the SOE estimation, which indicates that the proposed algorithm has strong tracking ability and robustness to the SOE of lithium batteries.