Model Predictive Control Strategy for Hybrid Ventilation Building Operation

Hybrid ventilation buildings, or mixed mode buildings, represent the buildings that have the capability of running the natural ventilation and mechanical ventilation concurrently or switching between these two modes during the building operation. To establish a hybrid ventilation building with robust performance, a well-developed building control algorithm is one of the key factors that determines it. In this paper, we developed a model predictive control (MPC) strategy for maintaining an acceptable indoor environment for building occupants while minimizing the energy consumption of a hybrid ventilation building. Established based on atomic neural network models, the developed model predictive control has the capability of accurately predicting the indoor environment conditions and energy consumption in different status of hybrid ventilation operation. In the model development process, 10 prediction variables are selected first and tested to see its influence on the prediction of response we are interested in. Six variables including indoor and outdoor air temperature, the outdoor relative humidity, the season and office hour index, and the wind speed were finally incorporated into our neural network models for predictions. At last, the model validation results showed that the developed MPC could perform as expected in maintaining the occupants thermal comfort.