Dynamic forecast of cooling load and energy saving potential based on Ensemble Kalman Filter for an institutional high-rise building with hybrid ventilation

Combining natural and mechanical ventilation, hybrid ventilation is an effective approach to reduce cooling energy consumption. Although most existing control strategies for HVAC systems with hybrid ventilation provide acceptable operation results, there still often exists a mismatch of demand and response from sensing, decision making, and operating. Especially when using renewable energy sources, such as solar and thermal storage, many energy-saving decisions need to be made before the actual events may happen. As a result, predictive-based controls are preferred, and the future energy loads and saving potentials from renewable measures should be evaluated in a forecasted manner. Typical prediction simulation methods are developed for designs and analysis, which may not ensure the required accuracy for modeling future events. In this study, a novel data assimilation method originating from numerical weather prediction, Ensemble Kalman Filter (EnKF), was proposed and applied for the forecasting simulations of high-rise building cooling load and energy-saving potential from its hybrid ventilation system. Similar to an accurate short-term weather prediction process, the proposed EnKF method can ensure the simulation accuracy by combining numerical simulations and measured data for short-term forecasting of future events. In the EnKF algorithm, a simulation model is adjusted according to the measuring data to output more accurate predictive results of the cooling load reduction from a hybrid ventilation system. Based on these predictions, the supply air temperature can be adjusted, and the duration of applying natural ventilation in real-time to maintain the desired comfort of building occupants with less energy consumption than existing strategies. The proposed forecasting model can be used in real life when combined with smart building controls. The results show that the proposed EnKF method improves the accuracy of the predicted velocities. The key EnKF parameters, Kalman filter gain, and the number of ensemble members are discussed as well. With the localized Kalman filter, the average RMSE and CVRMSE decrease by 46.4% and 53.5%, respectively.