Application Research of Intelligent Inspection Technology Based on Multi-Source Data Fusion in Digital Power Grid Construction

Abstract

Intelligent inspection technology has become a key support to ensure the safe and stable operation of the power grid. In view of the shortcomings of traditional inspection methods in terms of efficiency, accuracy and adaptability, this paper proposes a multi-source fusion architecture for intelligent inspection for digital power grids. By integrating multi-source heterogeneous information such as Unmanned Aerial Vehicle (UAV) remote sensing data, sensor timing information, GIS geographic data and equipment operation and maintenance text, combined with dynamic threshold adjustment, spatiotemporal correlation analysis and cascade attention mechanism, a complete processing process covering data collection, feature extraction, multi-source fusion and decision output is constructed. Experimental verification shows that the model’s F1 score in fault detection reaches 94.8%. Moreover, the performance retention rate in a 5 dB strong noise environment reaches 85.4%, the inference speed reaches 98 frames/second, and the practicality score is 0.83, the generalization ability across data sets is 89.6%, and the scalability retention rate on a thousand-node scale is 93.6%, which significantly optimizes the timeliness of inspections and reduces the consumption of human resources. The multi-source fusion method can not only effectively improve the accuracy of power grid equipment condition monitoring and fault prediction, but also enhance the robustness and practicability of the system in complex environments, and promote the inspection technology to be adaptive and reliable. direction evolution.