Hybrid Model for Forecasting Indoor CO2 Concentration

Indoor CO2 concentration is considered a metric of indoor air quality that affects the health occupants. In this study, hybrid model was developed for forecasting varying levels in residential apartment unit presence occupants by controlling ventilation rates heat recovery ventilator. model, mass balance equation single zone as white-box combined with Bayesian neural network (BNN) black box model. During learning process BNN estimated an aggregated unknown rate and transferred estimation to mass-balance equation. A parametric study conducted changing prediction horizons from 5 15 min, performance compared stand-alone The showed better than on experimental dataset living room bedroom. average MBE CVRMSE horizon min were 0.65% 5.23%, respectively, whereas those 0.99% 9.30%, respectively.