A Q-Lattice and EnergyPlus coupled modelling approach to estimate HVAC energy consumption and natural ventilation usages

The U.S. energy Information Administration (EIA) estimates that about 94.8 quadrillion Btu of energy is produced from various energy sources like petroleum, natural gas, coal, renewable, and nuclear power plants. Out of this total generated energy, mostly from non-renewable sources, about 12 quadrillion Btu energy is consumed by residential buildings and about 9.50 quadrillion Btu of energy is consumed by commercial building. Some estimates go further to state that about 40% of all energy generated are used by the building sector.

Out of all building system components, HVAC (Heating, Ventilation and Air Conditioning) equipment consume significant portion of this energy. Some studies estimate that about 30-50% of the energy consumed by a building is used by the HVAC system. The data obtained from energy simulation during preliminary design phase of the building is used for the HVAC system sizing. Some studies estimate about 50-150% discrepancy between actual and predicted energy usage is observed because most of the building simulation do not take dynamic Occupant Behavior (OB) changes into consideration.

In this study, we propose to integrate the OB predicting Q-Lattice model into EnergyPlus. The Q-Lattice model is trained using the newly introduced python package called ‘feyn’. Unlike other black box models like DNN and LSTM, the Q-Lattice model is fully interpretable as it describes the relationship between dependent and independent variables using the basic mathematical operations. To train the model, we collected data from 16 residential dorms located at Upstate New York for two years. We propose to train two Q-Lattice models; the first model estimates the HVAC energy usage independent of EnergyPlus.

The second model estimates the window opening behavior of the students. The second model will be integrated into EnergyPlus using EnergyPlus API to simulate the HVAC energy usage during the winter months in residential dormitories. The second model was also integrated with the infiltration and window airflow prediction DNN model developed in past study. After that, the HVAC energy usage will be evaluated using EnergyPlus with Q-Lattice embedded inside it.