Identification of Power Quality Disturbance Characteristic Based on Deep Learning

•A single disturbance characteristics identification model based on Seq2point CNN-GRU is constructed.•A PQD identification method based on deep learning is proposed.•An evaluation method of the disturbance characteristics identification model based on DTW is designed.•The identification method is adopted after filtering the fundamental wave from the PQD data. The increasing number of new energy sources and power electronic equipment integrated into the power grid system has resulted in frequent power quality disturbances (PQDs). Hence, in this study, a PQD identification method based on deep learning is designed to address this issue. First, seven common PQD characteristic identification models based on convolutional neural network and gated recurrent unit structure with sequence-to-point are constructed. The identification model can reduce computational complexity and computation time. Second, a PQD identification process based on the PQD characteristic identification model is designed, which includes data processing, characteristic identification, identification model output judging, and identification output. The PQD identification process can identify seven single disturbances and five complex disturbances. Case results demonstrated that the PQD characteristic identification model has satisfactory identification and generalization abilities. Moreover, the proposed identification method has a high identification accuracy rate, at least 98.43%. The proposed identification method can be applied to actual power systems.