Building retrofit multiobjective optimization using neural networks and genetic algorithm three for energy carbon and...

Area of Science:

• Building Science
• Sustainable Engineering
• Computational Optimization

Background:

• Industrial buildings require energy-efficient retrofits to meet global climate goals.
• Existing structures present challenges for improving energy performance and occupant comfort.
• A need exists for integrated approaches to optimize retrofit strategies.

Purpose of the Study:

• To develop and validate a multi-objective optimization framework for industrial building retrofits.
• To assess the performance of different surrogate modeling techniques for computational efficiency.
• To identify optimal retrofit solutions balancing energy, comfort, and carbon emissions.

Main Methods:

• Building performance simulation using DesignBuilder for energy, comfort, and emissions.
• Development and comparison of Backpropagation Neural Networks (BPNN) and Support Vector Regression (SVR) surrogate models.
• Application of Non-dominated Sorting Genetic Algorithm III (NSGA-III) and entropy-weighted TOPSIS for multi-objective decision analysis.

Main Results:

• BPNN showed higher predictive accuracy than SVR for surrogate modeling.
• The optimized retrofit strategy reduced thermal discomfort hours by 10.06%.
• Significant reductions were achieved in energy density index (35.45%) and life-cycle carbon emissions (28.86%).

Conclusions:

• The proposed optimization framework is effective for low-carbon renovation of existing industrial buildings.
• The study provides a practical decision-support tool for balancing competing retrofit objectives.
• The integrated approach facilitates sustainable and energy-efficient building upgrades.