Model Predictive Control

Control of HVAC systems may reduce congestion of the electricity grid on district level by shifting energy demand of buildings and increase the self-consumption of local photovoltaic energy. To achieve an optimal control of ventilation, occupant behaviour should be taken into account. To describe occupant behaviour, usually black box models are used and typically need large amounts of high quality training data. Alternatively, use of physical relations allows for a good predictive power requiring less training data.

In recent years, many multi-type package air conditioning systems for buildings have become widespread in office buildings in Japan, and there are many cases where one air conditioning space is shared by using several indoor air conditioning units. The advantages of multi-type package air conditioning system are that it is possible to operate and control individually for each indoor unit, and that the user can arbitrarily change the temperature setting of the indoor unit.