Improved particle swarm - untracked particle filtering for accurate battery energy state estimation with the influence of multi-parameter varying temperature constraints in Inner Mongolia power station

With the establishment of a new goal of ‟carbon neutrality” by 2060, and the gradual expansion of the application range of lithium batteries, it is critical to accurately and effectively estimate the battery energy state (SOE), one of the key indicators of lithium battery management systems. A temperature-affected 2RC equivalent model is built in this work. The least-square support vector machine model (LSSVM) and PSO-UPF are integrated, particle swarm optimization is presented, and the unscented Kalman filter is used as the suggested density function of the particle filter. This method, referred to as Particle Swarm - Unscented Particle Filter (PS-UPF), is utilized to achieve precise SOE estimation. This study validated the proposed method under three distinct working environments at varying temperatures and compared its results with standard methods to demonstrate validity and superiority. The experimental setup used a 70Ah lithium energy storage battery as the testing subject across three different operating conditions. According to the experimental data, at 5 °C, the MaxE values of the PSO-UPF algorithm under HPPC, BBDST, and DST conditions were 2.96 %, 2.79 %, and 3.15 %, respectively, while the MAE values were 0.39 %, 0.69 %, and 0.51 %, respectively. The RMSE values were 0.69 %, 0.86 %, and 0.66 %, respectively. Compared to the UPF, the MAE and RMSE values of the PS-UPF decreased by 85.8 % and 73.1 %, respectively, and compared to the traditional Particle Filter (PF), they decreased by 90.3 % and 80.5 %, respectively. Consequently, the PF algorithm's issues with particle deterioration and a lack of particle variety may be resolved by the PS-UPF method.