Engineering application of hybrid artificial intelligence for optimizing solar multigeneration systems in university buildings

Built environments with diverse energy demands, such as university buildings, face challenges in meeting energy requirements with renewable sources. This study develops a hybrid artificial intelligence framework to optimize a solar-based multigeneration energy system designed for complex building energy demands. The framework integrates Response Surface Methodology for design space exploration and screening, Artificial Neural Networks for capturing nonlinear system behavior, and Genetic Algorithms for global optimization, improving accuracy, convergence, and computational efficiency. Baseline and optimized models were evaluated using the Building Energy Optimization (BEopt) software and Engineering Equation Solver (EES) to assess demand-side and supply-side performance. The optimized configuration reduced electricity, cooling, and heating demands by 16.3, 15.8, and 16.9 percent, and achieved a maximum exergy efficiency of 29.23 percent, representing a 14.18 percent improvement over the baseline. Although the cost rate increased to 14.12 dollars per hour, this increase reflects the incorporation of upgraded system components and additional photovoltaic generation capacity beyond peak demand, which improves operational robustness, resilience, and performance. The system also achieved substantial reductions in site energy use and Carbon Dioxide emissions while generating surplus energy. By integrating a compression chiller, the design effectively harnessed solar energy for simultaneous cooling and heating, demonstrating a practical engineering pathway toward zero-energy performance at the campus scale. The outcomes confirm that hybrid artificial intelligence–based optimization can improve energy efficiency and environmental performance in built environments. Future extensions will examine additional renewable integrations across building types and broader climatic conditions, further supporting the framework's adaptability.