An innovative clustering technique to generate hybrid modeling of cooling coils for energy analysis: A case study for control performance in HVAC systems

Despite past studies, no comprehensive models or empirical correlations cover all aspects of performances of cooling coils under different flow regimes (laminar, transition, and turbulent). Moreover, the cooling coil is characterized by a highly nonlinear dynamic subject to multiple inputs, coupling between the latent and sensible heat transfer modes, uncertain disturbances, and strong dependence of the overall heat transfer coefficient on the flow type, all causing significant challenges when it comes to modeling. Therefore, a hybrid layer structure model was adopted in this study to overcome these challenges. The new approach used two different optimization methods, Neural Networks' Weights and Takagi-Sugeno (TS) fuzzy, and the hybrid layers tuned by the Gauss-Newton algorithm (GNA). The proposed model covered three types of fluid flow to represent the dynamic behavior of the water-side and air-side heat transfer coefficients, each of which was divided into seven clusters and had its unique TS consequence. This study also administered meaningful fitness tests in the responses of the eleven independent variables that serve as its inputs. Furthermore, its application shows the control performance saving more than 44% of HVAC system energy. Based on the results, it was concluded that the proposed model is suitable for estimating energy and cost savings for electric power and water flow rate efficiency. In addition, the response of all types of output flow can be evaluated when changing eleven independent variables that are manipulated by three different controllers.