Complex Power Quality Disturbance Recognition Research Based on Deep Complementary Fusion of 2-D Coding Transition

Accurate identification and classification of power quality disturbances (PQDs) is important for monitoring power system stability. In this study, a new intelligent identification method for power quality disturbances (PQDs) is proposed. First, 2-D image coding methods, namely, Gramian angular field (GAF) and recurrence plot (RP), are introduced to convert 1-D PQD signals into 2-D images and realize the feature-enhanced display. Second, the GAF and RP images are fused using a correlation-driven feature decomposition fusion (CDDFuse) model to realize the deep complementary enhancement of different image information. Then, multiscale feature attention (MSFA) MobileNetV2 is proposed for PQD identification, which can quickly and accurately identify the disturbance features in the signal. In comparison with other popular methods, the proposed method achieves higher accuracy in recognizing PQD, with a validation accuracy of up to 99.3%. Finally, PQD experimental platform is established, and the IEEE PES dataset is utilized to further certify the reliability and effectiveness of the proposed method.