A New Framework for Classification of Power Quality Signals Using GRU-based Intelligent Model in Utility Smart Grid Network

Accurately identifying power quality disturbances (PQDs) is essential for addressing issues in modern smart grids. This research introduces a deep learning framework combining a 1D convolutional neural network (1D-CNN) with a gated recurrent unit (GRU) network for efficient PQD signal analysis. The 1D-CNN extracts spatial features from input signals, while the GRU captures temporal dependencies, boosting classification accuracy. An active learning approach is employed to select representative samples from a mostly unlabeled dataset, minimizing reliance on manual labeling. Tested on both synthetic and real-world PQ data using MATLAB simulations and achieved 97.20% in real-time simulated utility smart grid network, the framework delivers faster processing, improved accuracy, and better generalization than current methods.