Model-Free Optimization Scheme for Efficiency Improvement of Wind Farm Using Decentralized Reinforcement Learning

Xu, Zhiwei; Geng, Hua; Chu, Bing; Qian, Menghao; Tan, Ni

doi:10.1016/j.ifacol.2020.12.767

Cited by 11 publications

(7 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Many wind farm control methods based on reinforcement learning (RL) use the Q-learning algorithm [64]. Additionally, distributed Q-learning is developed for optimizing farmlevel power production [65], with strategies to avoid abrupt changes in control variables. Further research has proposed distributed RL algorithms for increasing power generation through yaw angle control [49], and concepts like gradient approximation and incremental comparison in RL for optimal control actions [66].…”

Section: Ref Yearmentioning

confidence: 99%

Review on the Application of Artificial Intelligence Methods in the Control and Design of Offshore Wind Power Systems

Song,

Shen,

Huang

et al. 2024

JMSE

View full text Add to dashboard Cite

As global energy crises and climate change intensify, offshore wind energy, as a renewable energy source, is given more attention globally. The wind power generation system is fundamental in harnessing offshore wind energy, where the control and design significantly influence the power production performance and the production cost. As the scale of the wind power generation system expands, traditional methods are time-consuming and struggle to keep pace with the rapid development in wind power generation systems. In recent years, artificial intelligence technology has significantly increased in the research field of control and design of offshore wind power systems. In this paper, 135 highly relevant publications from mainstream databases are reviewed and systematically analyzed. On this basis, control problems for offshore wind power systems focus on wind turbine control and wind farm wake control, and design problems focus on wind turbine selection, layout optimization, and collection system design. For each field, the application of artificial intelligence technologies such as fuzzy logic, heuristic algorithms, deep learning, and reinforcement learning is comprehensively analyzed from the perspective of performing optimization. Finally, this report summarizes the status of current development in artificial intelligence technology concerning the control and design research of offshore wind power systems, and proposes potential future research trends and opportunities.

show abstract

Section: Ref Yearmentioning

confidence: 99%

Review on the Application of Artificial Intelligence Methods in the Control and Design of Offshore Wind Power Systems

Song,

Shen,

Huang

et al. 2024

JMSE

View full text Add to dashboard Cite

show abstract

“…Many RL-based wind farm control methods are based on the Q-learning algorithm [131]. A decentralized Q-learning algorithm was developed in [132] for optimizing farm-level power production. This scheme has the ability to avoid sharp changes in control variables.…”

Section: Power Generation Maximization Via Rlmentioning

confidence: 99%

Wind farm control technologies: from classical control to reinforcement learning

2022

View full text Add to dashboard Cite

Wind power plays a vital role in the global effort towards net zero. The recent figure shows that 93GW new wind capacity was installed worldwide in 2020, leading to a 53% year-on-year increase. Control system is the core in wind farm operations and has essential influences on the farm’s power capture efficiency, economic profitability, and operation & maintenance cost. However, wind farms’ inherent system complexities and the aerodynamic interactions among wind turbines bring significant barriers to control systems design. The wind industry has recognized that new technologies are needed to handle wind farm control tasks, especially for large-scale offshore wind farms. This paper provides a comprehensive review of the development and most recent advances of wind farm control technologies. This covers the introduction of fundamental aspects in wind farm control in terms of system modelling, main challenges, and control objectives. Existing wind farm control methods for different purposes, including layout optimization, power generation maximization, fatigue loads minimization, and power reference tracking, are investigated. Moreover, a detailed discussion regarding the differences and connections among model-based, model-free and data-driven wind farm approaches is presented. In addition, highlights are made on the state-of-the-art wind farm control technologies based on reinforcement learning - a booming machine learning technique that has drawn worldwide attention. Future challenges and research avenues in wind farm control are also analysed.

show abstract

“…In the last decade significant progress has been achieved by RL methods for many sequential decision-making problems with successful applications in various fields including games, robotics or biology. By formulating farm power output maximization as a distributed optimization problem, decentralized multi-agent RL solutions have led to significant increases in total power production on experiments with both static [1,4,5] and dynamic wind farms simulators [6,7]. In particular, decentralized learning approaches bypass the exponential dependency of the search space on the number of turbines, and promise to be more tractable by modeling every turbine as an agent following a local learning algorithm [5,[7][8][9]].…”

Section: Introductionmentioning

confidence: 99%

“…By formulating farm power output maximization as a distributed optimization problem, decentralized multi-agent RL solutions have led to significant increases in total power production on experiments with both static [1,4,5] and dynamic wind farms simulators [6,7]. In particular, decentralized learning approaches bypass the exponential dependency of the search space on the number of turbines, and promise to be more tractable by modeling every turbine as an agent following a local learning algorithm [5,[7][8][9]]. Yet new evaluations on dynamic simulations suggest that convergence time remains too long for online learning on real wind farms [7,10].…”

Section: Introductionmentioning

confidence: 99%

Towards fine tuning wake steering policies in the field: an imitation-based approach

Bizon Monroc,

Bušić,

Dubuc

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Yaw misalignment strategies can increase the power output of wind farms by mitigating wake effects, but finding optimal yaws requires overcoming both modeling errors and the growing complexity of the problem as the size of the farm grows. Recent works have therefore proposed decentralized multi-agent reinforcement learning (MARL) as a model-free, data-based alternative to learn online. These solutions have led to significant increases in total power production on experiments with both static and dynamic wind farms simulators. Yet experiments in dynamic simulations suggest that convergence time remains too long for online learning on real wind farms. As an improvement, baseline policies obtained by optimizing offline through steady-state models can be fed as inputs to an online reinforcement learning algorithm. This method however does not guarantee a smooth transfer of the policies to the real wind farm. This is aggravated when using function approximation approaches such as multi-layer neural networks to estimate policies and value functions. We propose an imitation approach, where learning a policy is first considered a supervised learning problem by deriving references from steady-state wind farm models, and then as an online reinforcement learning task for adaptation in the field. This approach leads to significant increases in the amount of energy produced over a lookup table (LUT) baseline on experiments done with the mid-fidelity dynamic simulator FAST.Farm under both static and varying wind conditions.

show abstract

Model-Free Optimization Scheme for Efficiency Improvement of Wind Farm Using Decentralized Reinforcement Learning

Cited by 11 publications

References 17 publications

Review on the Application of Artificial Intelligence Methods in the Control and Design of Offshore Wind Power Systems

Review on the Application of Artificial Intelligence Methods in the Control and Design of Offshore Wind Power Systems

Wind farm control technologies: from classical control to reinforcement learning

Towards fine tuning wake steering policies in the field: an imitation-based approach

Contact Info

Product

Resources

About