A Comparative Analysis of BiLSTM-GRU and RNN Models for Smart Home Energy Consumption Forecasting

Abstract

Energy conservation in smart homes is crucial for optimizing electricity usage, reducing costs, and contributing to sustainability efforts. This study evaluates the performance of two deep learning models, the Bidirectional Long Short-Term Memory with Gated Recurrent Units (BiLSTM-GRU) and the traditional Recurrent Neural Network (RNN), in forecasting smart home energy consumption. The research employs comprehensive data preprocessing, model implementation, training, and evaluation to determine the superior model. Experimental results demonstrate that the BiLSTM-GRU model consistently outperforms the RNN model across multiple forecasting windows, offering higher accuracy and robustness. The study provides insights into model selection for energy management applications and discusses the trade-offs between model complexity, interpretability, and computational efficiency.