INDOOR CLIMATE MONITORING AND REFINEMENT OF THE THERMAL MODEL OF A PUBLIC BUILDING BASED ON MEASUREMENT DATA

Abstract

The article presents a methodology for refining the thermal model of a public building based on indoor climate monitoring and energy consumption measurements. The relevance of the study stems from the uncertainty of the actual thermal properties of existing buildings and the operating conditions of their heating and ventilation systems, which may lead to substantial discrepancies between calculated and measured energy consumption. The aim is to develop and validate an integrated approach combining on-site measurements, uncertainty quantification, sensitivity analysis, and model calibration.

A simplified hourly resistance–capacitance (RC) model using a 3R1C scheme is proposed for simulating the thermal regime of a typical four-storey educational building in Vinnytsia, Ukraine. The uncertainty of nine input parameters is assessed using the Monte Carlo method, while global sensitivity analysis is performed using Sobol indices. The results indicate that the annual space heating demand ranges from 92 to 137 kWh/m² within the 5–95% uncertainty interval. The infiltration air change rate is identified as the dominant parameter, accounting for 64% of the variance in heating demand. Wall thermal transmittance and the indoor temperature setpoint have substantially smaller effects, whereas solar gains and internal heat gains play an important role in determining overheating conditions.

The study demonstrates that simultaneous calibration of all influential parameters against heat meter data may produce physically unrealistic parameter estimates due to the compensation of errors. To overcome this limitation, a two-stage calibration procedure is proposed. First, directly measured parameters, including infiltration, ventilation airflow, and indoor temperature, are fixed in the model. The remaining influential parameters are then adjusted through calibration. In a virtual experiment, this approach reduced the monthly normalized mean bias error (NMBE) from 14.5% to 1.0% and reproduced the wall thermal transmittance with an error of 2.6%.

The proposed methodology improves the reliability of building energy simulations and provides a basis for substantiated decisions concerning thermal modernization and the optimization of heating and ventilation systems.