Anomaly detection based on lightweight deep learning algorithms

Abstract The safe and reliable operation of power transmission lines is the key to sustainable stable grids. Using artificial intelligence techniques for channel status monitoring greatly improves inspection efficiency lines. However, discrimination mostly conducted on cloud side, which does not meet requirement real-time processing. In this paper, we build an intelligent system based edge computing. According specific characteristics real-world filed applications, a lightweight deep learning algorithm model designed. It then transplanted self-developed AI chip field test carried out. experimental results show that designed in paper balances well between accuracy, size inference speed. realizes local lines, promising detection are obtained.