Research on Condition Monitoring Model of Mobile Energy Storage Equipment Based on Machine Vision

This paper proposes an innovative condition monitoring model based on machine vision and multimodal fusion technology, VT-FusionNet, which is specially designed for condition monitoring of mobile energy storage equipment. The model mainly consists of three key modules: visual feature extraction module, temporal feature extraction module and multimodal feature fusion and decision module, which fully reflects the comprehensive analysis and processing capabilities of multi-source data. First, the visual feature extraction module adopts an improved convolutional neural network (CNN) to efficiently extract the visual features of the device surface through deep learning methods. Secondly, the temporal feature extraction module uses a combination of long short-term memory network (LSTM) and Transformer structure to perform time series modeling on various sensor data (such as current, voltage, temperature, vibration, etc.) generated during the operation of the equipment. LSTM is good at capturing short-term fluctuations and long-term trends in the operating status of the equipment. While Transformer, with its powerful global modeling capabilities, we need to further explore the complex associations in multidimensional time series data and provides accurate analysis for the dynamic changes of the equipment status. Finally, the multimodal feature fusion and decision module dynamically weights and fuses the visual features and time series features by introducing the attention mechanism. On this basis, the model can flexibly adjust the importance weights of features according to the needs of different application scenarios, thereby achieving more accurate state classification and anomaly detection. Through this multimodal collaborative processing strategy, VT-FusionNet demonstrates excellent performance in fusion perception and decision-making capabilities. Finally, this paper experimentally verifies that VT-FusionNet has significant advantages over existing state monitoring methods.