替代模型
控制理论(社会学)
风力发电
计算机科学
控制(管理)
数学优化
工程类
数学
人工智能
电气工程
作者
Yu Tu,Kai Zhang,Yaoran Chen,Zhaolong Han,Yong Cao,Dai Zhou
标识
DOI:10.1115/omae2024-125060
摘要
Abstract In wind farms, wake effects among turbines hinder optimal power generation, making collective yaw control crucial for efficiency. The accurate calculation of power generation for large-scale wind farm is time consuming and computationally expensive. To address this challenge, Surrogate-Based Optimization (SBO) emerges as an effective tool in the yaw optimization. In the design of experiment, wake effects are simulated by Gauss-Curl Hybrid (GCH) model in FLORIS. Based on the limited dataset, SBO utilizes surrogate models to approximate the relationship between yaw angles and total power in the entire design space, enabling a more efficient approach to yaw optimization. The selection of optimization algorithms within the SBO framework becomes pivotal, with the recognition that gradient-free methods excel in scenarios with multiple local minima, while gradient-based methods offer advantages in high-dimensional problems inherent in large-scale wind farms. This paper compares three optimization algorithms — Sequential Least Squares Quadratic Programming (SLSQP), Differential Evolution (DE), and Particle Swarm Optimization (PSO) — in the context of wind farm yaw optimization. Three yaw control case studies are presented, offering a comprehensive comparison of the strengths and limitations of the optimization algorithms. Results show that the gradient-free methods, balancing computational cost and accuracy, stand out as the preferred choice for wind farm yaw optimization. Despite the low-fidelity engineering wake model used in the current study, the SBO framework is also compatible with the higher-fidelity wind turbine models. This capability lays the foundation for optimizing yaw control under more realistic conditions.
科研通智能强力驱动
Strongly Powered by AbleSci AI