A traffic prediction based sleeping mechanism with low complexity in femtocell networks

Dynamic sleeping of base stations under off-peak time is an efficient way to reduce the energy consumption of cellular networks. However, traffic in femtocell networks is always under fluctuation, which leads to the high complexity of the existing sleeping approaches. To overcome this problem, in this paper a traffic prediction based sleeping (TPBS) mechanism with low complexity is proposed in femtocell networks. Artificial neural network (ANN) model is used to predict future traffic in each base station. Then based on prediction results, we propose a dynamic base station switching algorithm with low complexity, where base station can be switched off/on dynamically according to the traffic in real time. We show that this approach can minimize the number of active base stations while providing a satisfying service with low blocking probability to all users and, at the same time, the algorithm complexity is low compared with the Markov Decision Process (MDP) based sleeping method. Simulation results are used to evaluate the performance of our schemes considering energy saving and algorithm complexity.