Dynamic optimization of cooling temperature setpoints in multiple thermal zones of office buildings in connection with D3QN

Abstract

The heating, ventilation and air conditioning (HVAC) system optimization process needs to focus on both the comfort of the indoor environment and the energy consumption. In large office buildings, there are differently oriented thermal zones, connected thermal zones and multiple air handling unit systems. It becomes particularly difficult to develop control strategies that take into account the associated environmental comfort, increase overall comfort and optimizing system energy use at the same time. This study builds standard floors with multiple thermal zones in the EnergyPlus environment based on real buildings. We use a dueling double deep Q network (D3QN) strategy for determining indoor cooling temperature setpoints for multiple thermal zones in real time. Multiple optimization strategies with different combinations of action selection and reward calculation methods were designed on the target building, and the best strategy was selected in connection with multiple comfort evaluation metrics. The test results show that the screened strategy optimizes the comfort distribution of the relevant thermal zones at the same moment and the overall comfort of the indoor environment. And within the range of optional actions, the combination of actions selected by the strategy can trade-off comfort and system energy consumption.