<![CDATA[The advancement of Artificial Intelligence (AI) and Machine Learning (ML) has introduced various approaches to processing time series data, particularly for energy consumption forecasting. One of the most prominent architectures is the Recurrent Neural Network (RNN) and its variants, such as Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM), which are designed to capture temporal dependencies in sequential data. This study examines the development, characteristics, and performance of RNN and its variants across various domains, with a specific focus on building energy consumption forecasting. The reviewed research spans from 1990 to 2024 and was selected based on relevance, citation count, and novelty of contribution. The findings indicate that LSTM and GRU generally outperform standard RNNs in handling long-term dependencies, while BiLSTM is effective for complex data patterns. However, challenges such as the need for high-quality data, computational complexity, model interpretability, and integration into Energy Management Systems (EMS) remain significant barriers. This study reaffirms the importance of RNN and its variants in energy prediction systems while opening opportunities for further research on hybrid architectures and the development of more user-friendly interfaces.]]>
Copyrights © 2025
