Machine learning methods to assist structure design and optimization of Dual Darrieus Wind Turbines

An optimal structure of wind turbines, especially the Dual Darrieus Wind Turbines (DDWTs), conduces to the development and promotion of wind power industry. However, the conventional structure optimization in the previous researches is basically relied on the numerical or experimental methods. In view of the balance between solution accuracy and computational cost, this study develops a bi-level structure design and optimization model based on the algorithms of machine learning. The first hierarchy is the prediction model, including the non-parameters algorithms such as Artificial Neural Network (ANN), Adaptive Neuro-Fuzzy Inference System (ANFIS) and Support Vector Machine (SVM) to play a role as the objective function in the overall model. The second hierarchy, as the optimization model, is composed of the metaheuristic algorithms such as Particle Swarm Optimization (PSO), Simulated Annealing method (SA) and Genetic Algorithm (GA). Furthermore, the dataset used to train this hybrid model is generated by Orthogonal Test (OT) method and Computational Fluid Dynamics (CDF) simulation to produce the representative sample. Results reveal that the Hybrid Structure Design and Optimization Model (HSDOM) can reach to the optimal combination of chord ratio, radius difference and offset angle, with an identical accuracy compared with the outcome of OT-CFD model.(c) 2021 Elsevier Ltd. All rights reserved.

Automated summary

This study develops a bi-level structure design and optimization model based on machine learning algorithms to optimize the optimal structure of wind turbines. By combining prediction models and optimization models, it achieves a balance between solution accuracy and computational cost. The training dataset, generated by orthogonal testing and computational fluid dynamics simulation, produces results with the same accuracy as traditional models.