Test Before You Sleep: A Policy-Reactive Digital Twin for RAN Carrier Switch-Off

Abstract

Energy consumption is a dominant operational cost and a chief sustainability concern for Mobile Network Operators (MNOs), and the Radio Access Networks (RANs) alone account for around 73% of total network energy expenditures. Switching off capacity Radio Frequency (RF) carriers at times of low service demand is the most immediate means to reduce the energy footprint of RAN infrastructure. However, every switch-off decision also removes wireless capacity and risks degrading Quality of Service (QoS). The possibility of upsetting customers precludes operators from trialing aggressive carrier control policies on live networks, making it arduous to adopt innovative RAN energy efficiency policies, such as those based on Machine Learning (ML). In this paper, we present a policy-reactive RAN Digital Twin (DT) that reproduces at scale realistic traffic demands, post-switch-of handovers, energy costs, and QoS performance. The DT lets an external policy submit switch-on/off actions, redistributes load across the RAN according to such actions, and outputs energy consumption and potential QoS degradation over an open interface. We validate each component of the proposed DT against operator measurements and the literature, instantiate the twin on a real city-scale deployment with 3,054 carriers, and expose the energy-QoS trade-offs that characterize baseline threshold policies.