Fuel Efficient Deep Learning Model for Traffic Aware Power Optimization in Telecom Networks

Abstract

The recent boom in telecom infrastructure with a particular intensity in developing territories has resulted in the high growth of energy consumption with diesel generators (DGs) being the lifeline of backup energy supply at off-grid or unstable grid sites. Nevertheless, overdependence on diesel leads to exorbitant costs of operations and degradation of the environment. The following paper proposes a new capability of power optimization in telecommunication towers based on the prediction of the traffic within the network and the dynamic priorities of the fuel location. The suggested approach which is known as Traffic-Aware Fuel Optimization Algorithm (TAFOA) employs a Long Short-Term Memory (LSTM) model used to forecast the telecom traffic in every cell site after 20 hours. According to the estimated traffic, TAFOA determines the high priority seen cell sites that can probably have the most booming traffic and assigns DG fuel to the same. The algorithm combines the real-time traffic, grid power availability, and battery status, as well as the efficiency of DG to take the most suitable power source, and the minimum fuel usage to take advantage. The deployment of experimental results on synthetic telecom datasets confirms the competence of the model in the reduction of operational costs, efficient use of energy, and continuity of service. The incorporation of deep learning in telecom energy management offers a future-proof range of energy management by means of the optimization of fuel consumption and carbon footprint. The suggested TAFOA approach attained a total accuracy of 95.4% in traffic-aware fuel optimisation inside telecommunications networks.