Because of the environmental policies that pursue a significant reduction of global greenhouse gas and to the ambition of the network operators to keep under control operational and maintenance expenditures, energy efficiency has recently become one of the most relevant issues for present and future research activities. This contribution deals with the management of the base station sleep mode in a cellular network. The forecasting based sleep mode algorithm, justified by daily and weekly periodic behaviour of traffic, is presented and evaluated for general deployments of second and third generation cellular networks. The energy consumption of each base station and the area power consumption are the metrics, which are considered for our simulations. The results show a significant increase of the energy efficiency during low traffic periods with respect to the usual scenario, thanks to the adoption of the proposed strategy. The forecasting based sleep mode algorithm has been also compared with a strategy based on real‐time measurements and the results show similar performances while the number of updates of the network configuration and the global complexity are decreased. Copyright © 2013 John Wiley & Sons, Ltd.