ML-Optimization Procedures

ml-optimization
ml-optimization
ml-optimization
ml-optimization
ml-optimization
ml-optimization
ml-optimization
1 / 6
ml-optimization
2 / 6
ml-optimization
3 / 6
ml-optimization
4 / 6
ml-optimization
5 / 6
ml-optimization
6 / 6

To enhance the performance of optimization methods in relation to accuracy and computational requirements using neural networks

Technology & Design

Natural Sciences and Engineering Research Council of Canada (NSERC)
Description
Reliable optimization methods are critical for efficient conversion of solar irradiation to bioenergy in building-integrated photobioreactors. This paper presents a new procedure using Neural Networks to reduce the computational burden of meta-heuristic optimization methods such as Simulated Annealing, Genetic Algorithms, and Particle Swarm Optimization. This also addresses additional limitations, including platform dependencies and reduced flexibility. The developed procedure was applied to a four-variable building-integrated photobioreactors. The system selection was based on its exhaustive hourly bioenergy simulation throughout the year for a single run. The peak microalgae generation reached 2.25 kg/m2 annually with the optimal system design. On average, original algorithms required 17.2 hours, while Neural Network - Aided algorithms showed higher efficiency by demanding only 6.5 minutes. This emphasized the advantage of Neural Networks in reducing computational time and addressing optimization setup limitations in simulation software.
Publications