Building energy consumption prediction and optimization using different neural network-assisted models; comparison of different networks and optimization algorithms

The consumption of energy in buildings holds considerable importance within the realm of overall energy usage. This underscores the critical nature of employing efficient strategies for managing energy. Accurate assessments of energy consumption in buildings serve as a central factor in improving energy efficiency and providing guidance for energy management choices in the context of residential buildings. Therefore, predicting and optimizing building energy utilization has become a popular area of research because it has the potential to greatly improve how efficiently energy is used in buildings. In this study, in the first step, four different models, including three artificial neural network frameworks and a regression model are expanded to predict cooling and heating loads. After selecting the best network, using four optimization techniques, the hyperparameters of the selected network are tuned and the best hybrid model is obtained. Furthermore, the multi-objective optimization process is extended to define the optimal conditions using Particle Swarm Optimization, and Biogeography-Based Optimization optimizers and LINMAP, TOPSIS decision-making approaches. The findings of this investigation underscore the enhanced efficiency conferred by the BBO algorithm on the Extreme Learning Machine. Specifically, an increase in the correlation coefficient from 0.9959 to 0.9969 for cooling load estimation and from 0.9973 to 0.9993 for heating load estimation reflects the improved alignment of results from the ELM-BBO model with actual experimental data. These values surpass those of all other models, indicating that the ELM-BBO model demonstrates the best performance among the hybrid models. Importantly, the results obtained underscore the overall effectiveness of the selected optimizers in delivering accurate outcomes.