Artificial neural networks as a fast tool to assess the thermal comfort and performance of a wind catcher room in warm climate

This study evaluates the performance of an Artificial Neural Network-based model (ANN-BM) to predict occupant thermal comfort and the efficiency of a wind-catcher room in warm climates. The ANN-BM predicts the Nusselt number on heated and cooled walls (Nuhw and Nucw), air exchange effectiveness (AEE), predicted mean vote (PMV), and predicted percentage of dissatisfied (PPD) in naturally ventilated buildings. Predictor variables include outdoor temperature, temperature difference (ΔT) between walls, air velocity, relative humidity, clothing insulation and activity level, selected through independence and importance analysis. The model achieved an R2 of 0.960, with mean absolute errors of 2.83 (Nuhw), 3.05 (Nucw), 0.006 (AEE), 0.2677 (PMV), and 6.58% (PPD), demonstrating superior accuracy compared to traditional methods and multilinear models (R2 = 0.83). Additionally, computation time was reduced by over 99% compared to CFD simulations. This highlights the model’s effectiveness and efficiency in evaluating ventilation, thermal performance, and thermal comfort in naturally ventilated buildings.