A prediction model for CO2 concentration and multi-objective optimization of CO2 concentration and annual electricity consumption cost in residential buildings using ANN and GA

Environmental pollutants in the air have long been a great threat to the health and life of human society and the volume of these pollutants is rapidly increasing. Human beings spend most of their time in closed environments which highlights the demand for appropriate indoor air quality. This favorable air quality makes sense when the concentration of pollutants such as CO2 that have penetrated the building space is reduced. This paper aims to predict a model for CO2 emissions and optimize it for 6 cities of the U.S with different climates. Firstly, using the time series of GMDH type of artificial neural network, the amount of these pollutants was predicted monthly and annually from 2020 to 2025. The results show that the amount of CO2 pollutants during this period increases by 1-3% and 1.25-1.8% on a monthly and annual basis, respectively. In this research, to solve this huge predica-ment in the residential sector, 5 design variables are considered, which are the thermostat set point temperature of the air conditioning system for cooling and heating, the clothing insulation level of the residents' clothes for winter and summer seasons, and the amount of clean air that is transferred from the outside to the inside of the building by the air conditioning system. The goal is to simultaneously minimize CO2 emissions and annual electricity consumption costs of the building and improve the thermal comfort of building occupants. Therefore, design variables and objective functions in JEPLUS software are defined. Afterward, they are analyzed based on the building's energy performance using EnergyPlus software. The elicited data are then transferred to JEPLUS + EA software, where they are optimized by the NSGA-II algorithm, which finally discovers the most optimal states so that users can select any state that is in line with their goals.