Grey-box models for energy consumption prediction in large scale buildings: the case of an Italian hospital facility

Abstract

The present study aims to develop a lumped parameter RC-type model using a Grey-box approach to represent the dynamic thermal behaviour of a healthcare facility with an approximate usable area of 9,000 m², composed of several pavilions of different construction types and ages. The main objective is to investigate the maximum idealization limit of the thermal and dynamic behaviour of a complex structure in terms of both building configuration and internal activities. The model incorporates a comprehensive collection of construction, stratigraphic and environmental characteristics of the facility. The model uses the monthly energy consumption for air conditioning for the winter season as thermal input data. The identification of the unmeasured and unknown free contributions, which constitute part of the unknown components in the Grey-box model, was performed through Bayesian optimization, minimizing the sum of the quadratic deviations between the predicted and calculated values of the thermal contributions due to the internal air conditioning systems. The study identified four main components of the thermal behaviour related to the central node: transmission losses through the envelope, contributions due to solar radiation through transparent surfaces and two unknown shares depending respectively on the usable internal surface and the exchange of air volumes with the outside. The validity of the model was verified for the heating period from 2022 to 2024 (including part of 2025) using the CV(RMSE) and NMBE indicators according to the ASHRAE guideline 14 (“Measurement of Energy, Demand, and Water Savings”) respecting the limits of 15% and ±5%. The seasonal analysis allowed to evaluate the impact of each factor on the thermal balance and to identify the most effective areas of intervention. The proposed model can be used to predict energy consumption in similar complex structures, facilitating the planning of energy efficiency measures.