Development and performance evaluation of a minimum input model calibration methodology for residential buildings

Abstract

Energy simulations of existing dwellings are often impeded by the complexity of assigning appropriate model inputs. While data-driven calibration is an effective method to reduce variance between measured and simulated datasets, significant effort is required for monitoring and auditing. A new minimum input calibration method is proposed, where the number of inputs is greatly reduced through a three-step sensitivity analysis creating an input set with the most influential parameters on internal temperatures. The reduction in input parameters simplifies calibration and reduces the likelihood of unrealistic solutions. The proposed method is verified on two dwellings where conventional calibration techniques were compared to the minimum input calibration method using sub-hourly internal temperatures. Compared to baseline models, the variance of minimum input models reduced from 9.9% and 9.7% to 3.3% and 3.8% (CVRMSE (%)). Results indicate that minimum input model calibration can sufficiently predict thermal performance and could be applied to retrofit optimization. Acronyms: BoM: Bureau of Meteorology; BPS: building performance simulation; CD: cooling dominant; CDH: cooling-degree hours; CVRMSE: coefficient of variation of the root mean square error; D: discrete; ECM: energy conservation measure; HD: heating dominant; HDH: heating-degree hours; MAE: mean absolute error; MBE: mean bias error; MIM: minimum input model; MIS: minimum input set; MOAT: manual one at a time; N: normal distribution; NRMSE: normalized root mean square error; RMSE: root mean square error; RMY: reference meteorological year; SHGC: solar heat gain coefficient; U: uniform distribution