Optimizing solar-powered multi-generation systems for sustainable energy management in university buildings using Neural Networks and Genetic Algorithms

Abstract

University campuses operate like small cities, with complex energy demands driven by diverse activities, making energy management particularly challenging in extreme climates. Traditional energy systems often rely on fossil fuels, leading to inefficiencies, high costs, and increased carbon footprints. This study optimizes a solar-powered multi-generation system for a university building in the severe hot climate of Dezful, Iran, using a two-step approach. First, the Building Energy Optimization Tool (BEopt) was used to select optimal building materials, leading to a reduction in total energy demand. Then, Artificial Intelligence algorithms (Neural Networks and Genetic Algorithms) were applied to enhance system performance and cost-effectiveness. The Engineering Equation Solver (EES) validated the thermodynamic performance of the optimized system. This optimized system generates 49.55 gigawatt-hours (GWh) of electricity annually, along with 19.47 GWh of heating and 13.93 GWh of cooling, producing a surplus of 34.28 GWh (69% of total generation). Compared to the base case, electricity generation increased by over 224%, cooling by 120%, and heating by more than 3500%, while CO 2 emissions decreased from 37.18 to 35.34 metric tons/year. The optimized system operates with an exergy efficiency of 23.44% and a cost rate of 14.93 $/h. What sets this study novel is the integration of demand-side improvements with AI-enhanced multi-generation systems and thermodynamic validation, creating a practical model for achieving net-zero energy on university campuses. This methodology is scalable to more complex energy networks, including district energy systems, with future work focusing on real-time weather integration and advanced forecasting to improve adaptability and resilience.

Topics & Concepts