ENERGY FORECASTING MODELS FOR POWER AND TELECOMMUNICATION INFRASTRUCTURES: A SYSTEMATIC REVIEW

Abstract

Energy demand for power and telecommunication infrastructures has risen in recent years owing to technological advancements. The downtimes caused by an inadequate energy supply to critical infrastructure pose a great risk to daily activities. Hence, knowledge of future energy demands is pertinent to minimizing losses and operational costs, while ensuring consistent and reliable services. This article comprehensively reviews different models for predicting energy requirements by power and telecommunication infrastructure. The findings reveal that, while traditional models such as linear regression are simple to implement, models utilizing machine learning (ML) and deep learning (DL) techniques demonstrate superior performance in predicting energy consumption, yielding more precise outcomes. It has also shown that ML and DL models, including long short-term memory (LSTM), convolutional neural networks (CNN), Gated Recurrent Units (GRU), and hybrid architectures, are particularly effective for handling the complexities of long-term forecasting and adaptive systems. Thus, this current study offers valuable insights for academia, researchers, and energy personnel in network planning of the power and telecommunication industries to improve energy efficiency and cost management by analyzing historical data, identifying complex patterns, enabling real-time adaptations, and accurately forecasting their energy requirements. Researchers can also build upon the identified gaps to enhance the existing models and improve productivity.