Design Mining Interacting Wind Turbines.

Area of Science:

• Engineering
• Renewable Energy
• Computational Intelligence

Background:

• Previous research introduced surrogate-assisted evolutionary algorithms for vertical-axis wind turbine design.
• This method physically instantiated prototypes using 3D printing and evaluated them under fan-generated wind, avoiding mathematical models and assumptions.

Purpose of the Study:

• To extend initial findings by exploring alternative surrogate modeling and evolutionary techniques for vertical-axis wind turbine design.
• To compare the accuracy of different modeling algorithms for estimating individual fitness.
• To investigate the impact of temporal windowing on surrogate model training samples.

Main Methods:

• Comparison of various surrogate modeling algorithms for fitness estimation.
• Exploration of temporally windowed surrogate model training data.
• Introduction of a novel surrogate-assisted approach with enhanced local search.
• Examination of alternative coevolution collaboration schemes.

Main Results:

• The study provides a comparative analysis of surrogate modeling algorithm accuracy.
• It evaluates the influence of temporal windowing on training data for surrogate models.
• A new surrogate-assisted design approach and coevolution strategies are presented.

Conclusions:

• The research offers advancements in surrogate-assisted evolutionary algorithms for vertical-axis wind turbine design.
• Findings contribute to more efficient and assumption-free design optimization methods.
• The study paves the way for further exploration of enhanced local search and coevolutionary schemes.