Equipment Anomaly Detection in Power Grids Using Deep Learning
Zhixiong Shi, Qi Zhao, Lei Su, Yun Su, Nannan Yan · 2021 International Conference on Intelligent Computing, Automation and Systems (ICICAS) · 2021
With the rapid development of smart grids, intelligent inspection play an increasingly important role as a key part of smart grid construction, with the generation of a large amount of image data. How to establish an automated and efficient mechanism for detecting device anomalies on these big data is the focus of this paper. First, mask region-based convolutional neural networks (Mask R-CNN) is used in visible light images for equipment segmentation to detect all the equipment which need to be tested. Second, the temperature of these equipment are extracted from the corresponding infrared images. Then, different machine learning methods are used to classify anomaly and normal equipment. Testing results showed that Mask R-CNN can detect various electrical devices of different categories in one image simultaneously within 80 milliseconds and achieve the mean average precision (mAP) of 94.82% on a standard test set, which is of great significance in real-time power transmission line inspection. This equipment anomaly detection method combining visible light and infrared images not only refines the range of feature extraction, which reduce the noise of extracting features from the entire image and improve the classification accuracy, but also locates the anomaly equipment precisely in the image.