Optimization of operating parameters for combined impinging jet ventilation and chilled ceiling system using machine learning algorithms

The combined impinging jet (IJV) and chilled ceiling (CC) system, denoted as IJV/CC, is a promising ventilation strategy; however the coupling between the IJV and CC remains underexplored. This study seeks to address this gap by applying machine learning (ML) algorithms. First, a comparative analysis of various ML models, including Genetic Algorithm-Back Propagate Neural Network (GA- BP), Genetic Algorithm-Support Vector Machine (GA-SVM), Radial Basis Function (RBF), and Extreme Learning Machine (ELM), was conducted to identify the most suitable model for the IJV/CC system. Results indicate that the ELM model offers the quickest response but with the lowest prediction accuracy. The GA-BP model exhibits the best generalization ability but with the longest computational cost. Nevertheless, the GA-BP model reduces computational time by approximately 17 times compared to CFD simulations. Subsequently, the GA-BP model, combined with Spearman's rank correlation coefficient, was used to optimize the operation conditions of the IJV/CC under varying ventilation demands. The impact of the operating parameters of the IJV/CC on each ventilation performance indicator is discussed and it highlights the importance of carefully controlling the supply temperature of the IJV, as it has opposing effects on the improvement of different performance indicators. Moreover, a comprehensive graph depicting the optimal combination of operation parameters for the IJV/CC system is developed in this study. The findings not only contribute to the existing body of knowledge on the role of artificial intelligence in predicting the built environment but also provide valuable reference for the practical application of IJV/CC systems in buildings.