Anomaly detection technology plays an important role in the field of industrial production and intelligent manufacturing, and has received extensive attention. The goal of anomaly detection is to detect potential faults or anomalies in a timely manner by monitoring and analyzing the operating status of equipment, systems, or processes, so that appropriate measures can be taken to repair or adjust. In our work, an anomaly detection network based on multi-attention mechanism is proposed, which captures the temporal, spatial, and physical connections in industrial data through multi-head attention mechanism and graph attention mechanism to improve the performance of anomaly detection. Finally, we verify the performance of the algorithm through a real dataset collected from industrial energy storage battery system. Analysis of the experimental results shows that the algorithm can effectively capture spatio-temporal correlation information in the data.