Soft computing approaches for energy-efficiency modelling: An sight into multi-attributes utilization

Cooling load is the amount of heat that needs to be removed from a room, structure, or system in order to maintain the appropriate temperature and humidity level. In this study, we focus on using five (5) different hybrid algorithms; Feedforward Neural Network (FNN), Radial Basis function network (RBF), Generalized Regression Neural Network (GRNN), Nonlinear Autoregressive with Exogenous inputs (NARX) and optimized Interaction Linear Regression (ILR) algorithm for cooling load prediction. The datasets are pre-processed by normalization, scaling, partitioning, input-output correlation analysis, sensitivity analysis, and feature scaling. The performance of the algorithm was evaluated using Pearson correlation coefficient (PCC), Nash-Sutcliffe efficiency (NSE), Mean Absolute Error (MAE), the Mean Absolute Percentage Error (MAPE), the Root Mean Square Error (RMSE), and Mean Square Error (MSE). Overall, RBF-M1 outperformed other models by achieving 1.00 for NSE and PCC, respectively. RBF-M1 also achieved MSE, RMSE, MAE, and MAPE values approaching zero. The study highlights that the different hybrid algorithms efficiently predict cooling loads whose accuracy is essential for architecture and in designing a suitable HVAC for residential, commercial, and industrial buildings.