Paper Conference

Proceedings of eSim 2014: 8th Conference of IBPSA-Canada


Recursive thermal building model training using Ensemble Kalman Filters

Brent Huchuk, Cynthia A. Cruickshank, William O’Brien, H. Burak Gunay

Abstract: Simplified building models have gained in popularity as a result of their practical applications in model-based predictive control (MPC); however, these models require accurate parameter estimates in order to give meaningful results. To this end, this simulation-based study proposes a recursive parameter estimation methodology — the so-called Ensemble Kalman Filter — as a method for predicting effective resistance and capacitance values of an office space. Accuracy of the proposed methodology to predict the response of the simulated-building was studied. Results indicate that the proposed methodology resembles a promising approach to learn building physics in a self-adaptive manner using simple generic sensor/meter networks; lending itself to the controls practitioners.
Pages: 727 - 737