Design of intelligent inspection system for 500 kV substation based on multi-source data fusion

INTRODUCTION: The safe and stable operation of 500 kV substations is essential for ensuring power grid reliability. Traditional manual inspection methods suffer from low efficiency, limited coverage, and poor real-time performance, making intelligent inspection technologies increasingly important. OBJECTIVES: This study aims to develop an intelligent inspection system for 500 kV substations based on multi-source data fusion to improve equipment status monitoring, fault diagnosis, and operational reliability. METHODS: The proposed system integrates infrared thermal images, visible-light images, sound signals, and vibration data for comprehensive equipment perception. A multi-modal deep learning network with an attention mechanism (AMM-Net) is designed for adaptive feature extraction. In addition, pixel-level, feature-level, and decision-level fusion strategies are combined with a trust-based distributed Kalman filtering algorithm (Trust-DKF) to improve robustness and anti-interference capability. RESULTS: Experimental results show that the system achieves 94.7% accuracy and 93.1% F1-score in equipment status recognition. Under noise and occlusion interference, performance decreases by only 10.7%. The edge-device inference time is optimized to 28.9 ms with low energy consumption of 0.12 J per operation. CONCLUSION: The proposed system significantly improves the efficiency, accuracy, and reliability of intelligent substation inspection.