Hybrid CNN–LSTM Model with Attention Mechanism and Feature Optimization for Electricity Theft Detection in Smart Grids

Key Points

• The research aims to improve the detection of electricity theft in smart grids using a hybrid CNN-LSTM model.
• Developed a hybrid model combining CNN and LSTM with an attention mechanism.
• Optimized features using PCA to enhance model performance.
• Evaluated the model's accuracy on electricity theft data.
• The hybrid model achieved significantly improved accuracy over traditional methods.
• Temporal patterns of electricity usage were effectively captured.
• Feature optimization using PCA notably enhanced detection capabilities.