Abstract Heterogeneous Cloud Radio Access Networks (H− C R A N) is a network architecture that combines Macro Base Stations (MBS) s and Small Base Stations (SBS) s with cloud infrastructures. The dense deployment of SBSs in H-CRAN is needed to provide high data rates to User Equipments (UEs) but causes high energy consumption. Unrealistic power models lead to inefficient UE association schemes in terms of energy. In this paper, we study the joint optimization of Uplink (UL) and Downlink (DL) decoupled UE association and switching on/off the SBSs in H-CRAN by incorporating a realistic power model with the objective of minimizing the power consumption in H-CRAN. The power model encompasses static and the dynamic power consumption of MBS, the static power consumption of SBS, the power consumption of transmission links to cloud infrastructure and the power consumption of UEs. The problem is transformed into Single Source Capacitated Facility Location Problem (SSCFLP) which is NP-Hard. We then propose a heuristic algorithm based on the use of LP relaxation and solving many-to-one assignment problem with generalized assignment problem heuristics. Extensive simulations demonstrate that the proposed heuristic algorithm performs very close to optimal and achieves significant improvements in minimizing total power consumption compared to coupled UE association algorithm and algorithms utilizing the power consumption models that do not encompass MBS dynamic power consumption and the power consumption of transmission links to cloud infrastructure for various scenarios.