Fusion Equivalent Circuit Model Based on BP Neuron Network

In this paper, three equivalent circuit models 1RC, 2RC, and PNGV are explored to characterize the battery's dynamic process. It turns out that different models have their own advantages in specific local processes. The characteristics of these three models can be combined to make up for each other's shortcomings and enhance the accuracy of the overall model. However, existing fusion algorithms are overly complex. When the estimated values of the equivalent circuit model are all lower than the true values, even after fusion, the accuracy cannot be further improved. To solve these problems, a fusion model based on the BP neuron network is proposed. The three equivalent circuit models are weighted and fused through the weight calculation process of the neuron network. Finally, it is verified under three operating environments of 0°C, 25°C, and 45°C. The fusion model can dynamically adjust the fusion weights according to the estimated states of the three models at different times. The simulation results show that the fusion model has significantly smaller maximum error, MSE, and RMSE than the other three equivalent circuit models at three different temperatures, and its dynamic response is also rapid at voltage jumps.