Optimization -informed design of office buildings in Egypt’s new capital using supervised machine learning regression and genetic algorithms

In Egypt, the construction of the New Administrative Capital (NAC) is projected to add over 1.54 million sqm of floor area to the country’s commercial building stock. Given the significant impact on energy consumption of this vital sector, there is an urgent need to ensure that design is optimized to achieve energy performance targets while delivering thermal comfort standards. However, limited research addresses such optimization-informed design approaches in Egypt, representing a major gap in current literature. To address this, the paper investigates the potential of developing and applying an optimization-informed design approach focusing on three key parameters: building form, envelope and orientation. A novel Building Envelope and Architectural Form Performance Optimization (BEFPO) framework based on Multi-Objective Optimization (MOO) and Machine Learning (ML) methods was developed. This framework considered Energy Use Intensity (EUI), Annual Solar Exposure (ASE), spatial Daylight Autonomy (sDA), and View Percentage to the Outdoor (VPO). Existing studies have been largely limited to optimizing simple hypothetical forms due to modelling and computational challenges. This study advances the field by applying the BEFPO framework to the complex L-shaped Cellular Office Building Typology (COBT). Twelve ML regression models were trained to predict performance outcomes, and results were validated against ground truth. Findings showed significant performance improvements, (in comparison to the base case, without any optimization applied): 8.82% (EUI), 16.5% (ASE), 7.7% (sDA), and 45.2% (VPO). The Bayesian Ridge Regressor (BRR) achieved the highest accuracy for EUI, while the Gradient Boosting Regressor (GBR) outperformed for other objectives, confirming BEFPO’s accuracy (