Energy Saving in 5G Cellular Networks using Machine Learning Based Cell Sleep Strategy

The optimization of fifth-generation (5G) mobile networks through energy-saving has become a significant objective for telecom operators. By reducing energy consumption, operators can improve network performance, reduce operating costs, and minimize their carbon footprint. The Operations Support System (OSS) of telecom operators can perform real-time monitoring of cells for online mobile network services. In this article, we propose two approaches for analyzing and predicting cell traffic. The first approach uses traditional statistical methods. Historical data of the cell can be used to calculate statistical metrics and perform traffic analysis. This information can be used to predict future traffic of the cell. The second approach is based on machine learning. Since a significant amount of training data is required for machine learning, we are collecting hourly Performance Monitoring (PM) data through the OSS system. Among many Key Performance Indicators (KPIs), we select five key KPIs that have the strongest correlation with the cell’s traffic. We use these five indicators as inputs in the machine learning model, Support Vector Machine (SVM) regression [1]. Using the machine learning model, we can predict the traffic of the cell. By setting appropriate thresholds, cells can switch to sleep mode without affecting the Quality of Service (QoS) and Quality of Experience (QoE) of users. We also demonstrated, through actual data from the OSS, the efficacy of our proposed cell sleep strategy in achieving relatively ideal energy-saving goals.