Heat pumps are one of the key technologies for the mitigation of carbon emissions in the heating sector. Therefore, they are an essential modelling component when analyzing future energy systems. The focus of this work are groundwater heat pumps and their integration into energy system optimization models. Three different integration approaches are introduced and compared to each other. The approaches differ in the representation of heat pumps efficiency: constant, time dependent and both time and spatially dependent. The comparison of the proposed approaches is conducted using an energy system model of the residential heating sector in Munich and two optimization scenarios: 70 % and 95 % emission reduction compared to 2014. Assuming constant efficiency of heat pumps throughout the year produces incorrect optimization results, whereas adding spatial component to the temporal one does not change cumulative results significantly. Thus, the second approach is the most suitable one due to its lower complexity compared to the last approach. However, if spatially distributed results are required, then the last approach is necessary. Finally, these integration approaches are applicable to all types of heat pumps.