Abstract

The importance of real-time fault detection and stability forecasting is growing in today's grids with nonlinear loads and renewable energy. Conventional machine learning methods perform well in classification but struggle with high-order temporal-spatial correlations in waveforms that are not steady. This study introduces a hybrid quantum-classical framework that combines deep learning, quantum feature embedding, and wavelet preprocessing.50,000 waveform samples covering normal, overload, short-circuit, and harmonic states were collected into a dataset.  Angle encoding was used to convert features into qubit states, which were further processed by a variational quantum circuit (VQC) and categorized using a CNN-LSTM model. The accuracy rate of the framework was 95.6%, and its latency was lower than that of the classical and quantum-only baselines.

Statistics