An AI-Driven Intelligent Health Assessment System for High-Voltage Switchgear Based on Multimodal Sensor Data Fusion

Abstract

The function of high-voltage switchgear is to maintain the stability of the power grid. The traditional health evaluation methods generally rely heavily on a single index; The accuracy of this method is relatively low, and it is difficult to detect failures because it cannot take into account multiple factors. In view of the above shortcomings, this paper has designed and developed an AI-intelligent health evaluation System based on multimodal sensor data fusion technology.The system uses the four-tier architecture of sensor layout, data acquisition and transmission, data fusion and processing, and health evaluation and decision-making, and integrates temperature, humidity, partial discharge, vibration, current, and voltage data. It uses a three-tier data fusion approach (weighted average at the data layer, principal component analysis (PCA) at the feature layer, and D-S evidence theory at the decision-making layer) to handle heterogeneous data and combines SVM models for small sample sizes and CNN models for time series data to provide all-around evaluation.Experiments based on a simulated platform show that the system can achieve an average accuracy of 96.8% and perform better than traditional single parameters by about 18.5 per cent. It can identify incipient faults 2-3 hours earlier than traditional methods, maintain 85% assessment accuracy even if one sensor fails, and complete post-fault state assessment and fault location within 500 ms after the relay protection action for short-circuit faults. The integrated three-level warning mechanism can reduce blind maintenance by as much as 60 per cent and lower the cost of maintenance by an average of 35 per cent; At the same time, it has achieved an effective grasp of complex fault conditions, improved the credibility of evaluation, provided scientific basis for preventive maintenance.