A New Online Battery Pack Inconsistency Monitoring Method for Battery Energy Storage Systems Based on Single Cell Voltage Difference

In the current complex battery manufacturing and operation environment, monitoring dynamic voltage deviations within battery packs is a prerequisite and key to eliminating the “short-board effect” risk and ensuring safe battery operation. However, the differences among individual cells caused by battery ageing make the voltage signals of battery packs exhibit dynamic deviation characteristics, which increases the difficulty of inconsistency monitoring. To address this issue, an online monitoring method for inconsistency based on a single cell voltage differences algorithm is presented in this paper. Firstly, a calculation method of K-value based on linear fitting is proposed. This method will collect the voltage difference data of a single cell, perform linear fitting and differentiation to obtain the rate of change k, and normalize the rate of change k to obtain the K-value characterizing the inconsistency of the battery pack. Next, an improved adaptive BP neural network method for predicting the K-value is proposed. By comparing the actual K-value calculated with the K-value predicted by the improved adaptive BP neural network, the risks generated during the operation of the battery can be predicted. Integrating the K-value calculation method and the improved adaptive BP neural network algorithm, this paper further proposes an online monitoring method for inconsistency based on single-cell voltage differences. Finally, experimental data verification shows that this method’s prediction errors (MAE: 0.109%, MSE: 0.0019%, RMSE: 0.139%) are all below 0.5%, verifying its effectiveness and engineering applicability.