Applying modified coot optimization algorithm with artificial neural network meta-model for building energy...

Area of Science:

• Building energy performance optimization
• Computational intelligence in engineering

Background:

• Multi-criteria optimization of real building designs presents significant challenges in the building energy performance field.
• Existing simulation-heavy methods are computationally expensive and time-consuming for complex building designs.

Purpose of the Study:

• To propose a novel meta-model-based method for efficient multi-criteria optimization of building energy performance.
• To reduce the computational cost associated with validating and optimizing building designs for thermal comfort and energy efficiency.

Main Methods:

• Utilized EnergyPlus™ for building performance simulations.
• Employed a hybrid approach combining multi-criteria Modified Coot Optimization Algorithm (MCOA) with artificial neural network meta-models (ANN-MM).
• Developed an optimized sampling strategy for training and validating ANN meta-models, minimizing required simulations.

Main Results:

• The proposed ANN-MM and MCOA method achieved accurate Pareto sets for multi-criteria optimization problems.
• Demonstrated significant savings, reducing the number of required physics-based simulations by 75% compared to simulation-based methods.
• Successfully optimized a real house for thermal comfort and energy efficiency, balancing heating and cooling demands.

Conclusions:

• The meta-model-based approach offers an efficient and accurate solution for multi-criteria optimization of real building designs.
• This method provides substantial computational savings, making complex building energy performance optimization more accessible.
• The study validates the effectiveness of ANN-MM and MCOA for achieving optimal building performance trade-offs.