Battery state-of-charge estimation based on fuzzy neural network and improved particle swarm optimization algorithm

The battery state-of-charge (BSOC), which is a key reference for battery management, can not be measured directly. An estimation approach for the SOC of the individual lithium-ion battery cell used in applications where the battery load i s very unstable is studied. The estimator model based on fuzzy neural network (FNN) is proposed, considering the battery terminal voltage, discharge current and battery surface temperature as inputs. Furthermore, an improved particle swarm optimization (PSO) algorithm borrowing the operation of selection and crossover from genetic algorithm (GA) is developed to train the FNN-based estimator model. This hybrid algorithm can incorporate the superiorities of the two heuristic optimization techniques. The results of simulation and experiment demonstrate that the improved PSO algorithm is more adaptive to the initial value of the parameters than the traditional training method using the back propagation (BP) algorithm. The FNN model trained by the improved PSO algorithm has better generalization capacity. At the same class of the training accuracy, higher accurate estimation value of BSOC can be obtained by employing the improved PSO algorithm. This study proposes an effective approach for estimating the BSOC.