CHEN Yanqiao, LIU Hui, WANG Chu, TAO Ye, JIN Yi, XIAO Kaiwen, DENG Aidong, NIU Hongbin, SHI Yaowei
The large-scale application of lithium batteries in energy storage stations poses severe challenges to their safe operation and maintenance. Currently, this field faces core issues such as a lack of research on fault diagnosis and early warning algorithms for energy storage station batteries, a shortage of professional experimental fault data, simplistic selection of fault features, and generalized early warning strategies without clear classification. This paper proposed a fault diagnosis and hierarchical early warning method for energy storage batteries based on multi-fault feature analysis. Relying on typical fault experimental data from 314 Ah energy storage batteries, deeper fault features integrating dynamic differential characteristics, time-domain statistics, and information entropy were constructed. Furthermore, a clear hierarchical early warning framework targeting overcharge and thermal runaway processes was designed. On this basis, a fault diagnosis model based on LightGBM and a hierarchical early warning model based on CNN were established respectively. Experimental results on the experimental dataset demonstrate that this method can effectively extract key information from fault evolution. The fault diagnosis accuracy reaches 99.10%, while the recognition accuracy rates for overcharge and thermal runaway warning levels are 98.72% and 96.57%, respectively. This provides new ideas and effective solutions for addressing safety issues in battery systems of energy storage stations.
