<![CDATA[Accurate short-term electricity load forecasting (STLF) is critical for efficient energy management, demand response, and grid stability in modern residential environments. This study presents an empirical investigation of a hybrid convolutional neural network–long short-term memory (CNN–LSTM) model applied to the UCI Electricity Load dataset, integrating convolutional layers to extract localized temporal features and stacked LSTM layers to model long-term dependencies across households. The model is trained using Python and TensorFlow on a GPU-enabled workstation, with preprocessing including normalization, sliding-window sequence generation, and train-validation-test splitting. Performance is evaluated through mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and k-fold cross-validation. Comparative benchmarking against LSTM, CNN-GRU, and multi-scale CNN-LSTM architectures demonstrates superior accuracy, stability, and generalizability of the proposed hybrid model. Sensitivity and interpretability analyses further reveal critical temporal patterns, feature contributions, and operational insights, facilitating actionable energy management decisions. These results substantiate the hybrid CNN–LSTM approach as a robust, interpretable, and operationally relevant solution for STLF applications.]]>
Copyrights © 2026
