AbstractThis paper proposes a method using neural networks to calibrate numerical models. The approach passes the output of numerical model to a neural network for calibration. An experimental study was conducted using a simulation of unheated and uncooled indoor temperature of a sports hall. The proposed neural network-based model improves the results and produces more accurate calibrated indoor temperature. Furthermore, the developed calibration method requires only measurements of indoor temperatures as the necessary inputs, thus significantly simplifying the calibration procedure needed to model the building performances.