Optimal modification of heating, ventilation, and air conditioning system performances in residential buildings using the integration of metaheuristic optimization and neural computing

Abstract This study pursues optima modification of heating, ventilating, and air conditioning (HVAC) systems embedded in residential buildings through predicting heating load (HL) and cooling load (CL). This purpose is carried out by employing four wise metaheuristic algorithms, namely wind-driven optimization (WDO), whale optimization algorithm (WOA), spotted hyena optimization (SHO), and salp swarm algorithm (SSA) synthesized with a multi-layer perceptron (MLP) neural work in order to overcome the computational shortcomings of this model. The used dataset consists of overall height, glazing area, orientation, relative compactness, wall area, glazing area distribution, roof area, and surface area as independent factors, and the HL and CL as target factors. The results indicated a high capability of all four metaheuristic ensembles for understanding the non-linear relationship between the mentioned factors. Meanwhile, a comparison between the used models revealed that SSA-MLP (ErrorHL = 1.9178 and ErrorCL = 2.1830) is the most accurate model, followed by WDO-MLP (ErrorHL = 1.9863 and ErrorCL = 2.2424), WOA-MLP (ErrorHL = 2.1921 and ErrorCL = 2.5390), and SHO-MLP (ErrorHL = 3.1092 and ErrorCL = 4.5930). Regarding the satisfying accuracy of the SSA-based ensemble, it can be a reliable tool for estimating the HL and CL for future smart city planning.