Consumer Behavior-Enhanced CNN-BiLSTM-Attention Model for Residential Electricity Consumption Forecasting

Accurate short-term prediction of residential electricity consumption is critical for smart grid integration, energy efficiency, and proactive building management. This study presents a hybrid deep learning framework that integrates convolutional neural networks (CNN), bi-directional long short-term memory (BiLSTM), and an attention mechanism, enhanced by dynamic consumer behavior indicators and optimized using genetic algorithms. The proposed CNN-BiLSTM-Attention model achieves state-of-the-art forecasting performance, with a root mean square error (RMSE) of ۰.۰۷۸ and a ۲۰.۵% improvement over baseline models. Genetic algorithm-based hyperparameter tuning yields an additional ۵.۱% performance gain. The model is validated across a public benchmark dataset, demonstrating its adaptability and robustness in real-time smart building energy management scenarios. Attention mechanism offers interpretable insights into consumption patterns, enabling actionable decisions for energy optimization and demand response. Extensive ablation studies and comparative evaluations with existing methods confirm the model's efficacy and practicality for real-world deployment. This work advances intelligent, adaptive load prediction systems for residential environments, supporting the broader goals of sustainable energy management and next-generation smart grid systems.

کلیدواژه ها:

Attention Mechanism ، Consumer Behavior ، deep Learning ، demand response ، Electricity consumption forecasting ، Energy management ، Genetic algorithm ، Neural Networks ، Smart buildings ، Time series prediction