BiLSTM Based Traffic Prediction for the Reduction of Energy Consumption on a Tri-Sector Radio Site of an EDGE Mobile Network

This article investigates the application of a Bidirectional Long Short-Term Memory (BiLSTM) prediction model to estimate traffic in an Enhanced Data rates for GSM Evolution (EDGE) mobile network. Mean absolute error (MAE) and root mean square error (RMSE) are used to evaluate the model. An optimization algorithm based on the proposed model was developed for the reduction of electric energy consumption associated with radio equipment and its infrastructure on a tri-sector radio site without affecting the quality of service required by the telecommunications operator. The obtained results show a reduction in energy consumption of approximately 17% on a tri-sector radio site. Furthermore, considering only the hours of low traffic (01 a.m.-07 a.m.), this reduction in energy consumption is very significant and is around 50%.