2-D Images Combination and Expanded-Channel ResNet Based Power Quality Disturbance Recognition Approach

With the widespread access to multiple energy systems, different degrees of power quality disturbance (PQD) have emerged in the power system, significantly impacting all stages of power generation, transmission, and consumption. Accurately identifying PQD lays the foundation for the safe operation and regulation of power quality within the power system. However, in traditional PQD recognition schemes, limited image feature information and insufficient algorithm recognition capacity are identified as key problems. To address these issues, a novel PQD recognition approach based on 2-D image combination and Expanded-Channel ResNet (ECRN) is proposed. Considering the insufficient time-frequency domain feature information, this article effectively combines Stockwell transform (ST) with Variational Mode Decomposition (VMD) to construct multi-channel PQD image that contain sufficient time-frequency domain feature information. In terms of the recognition algorithm, to better extract features from the disturbance images, this paper improves the traditional ResNet structure by expanding its channel and creates ECRN to achieve automatic and adaptive feature extraction from multi-channel images. The PQD recognition method is simulated and compared with state-of-the-art recognition approaches. The results demonstrate that the proposed method has good anti-noise performance and can more accurately extract PQD feature information to achieve higher recognition accuracy.