Power Quality Disturbance Identification Based on Optimized VMD and CNN-BiGRU

With non-linear loads connected to the power grid, the harmonics they generate can cause distortion in the current and voltage waveforms, interfering with the accelerated aging of equipment. To ensure the stability of the power system, improve energy efficiency, and optimize system operation, it is necessary to identify and address disturbances in electric power quality. In order to identify power quality disturbance efficiently and accurately, this paper proposes the chaotic Crayfish Optimization Algorithm (COA) to optimize two parameters of variational mode decomposition (VMD), and to extract and select nine time-domain indexes of the Intrinsic Mode Function (IMF) of the minimum fitness function value index. The feature phasors are fed into the Convolutional Neural Network-Bidirectional Gated Recurrent Unit (CNN-BiGRU) of multi-head attention mechanism (MHA) for perturbation identification. Using Matlab platform to simulate ten types of disturbances, the accuracy rate can reach 99.72%, which verifies the accuracy of the model.