Reinforcement Learning-Based Adaptive Control for Floating Wind Turbines

Project Description:

This Research Project is part of the EPSRC CDT in Offshore Wind Energy Sustainability and Resilience’s Innovations in Offshore Floating Wind Energy Systems Cluster.

Wind energy is a vital component of renewable energy sources, and floating offshore wind turbines (FOWTs) are gaining attention due to their potential for efficient power generation in deep waters. However, FOWTs are subjected to complex wind-wave environments, leading to challenges in maintaining power generation efficiency and structural integrity. This research proposal aims to develop an innovative reinforcement learning (RL)-based adaptive control strategy for wind turbine pitch control in FOWTs. The proposed approach will focus on simultaneously achieving power regulation and load mitigation without relying on accurate analytical models. The control strategy will utilize the Incremental Dual Heuristic Programming (IDHP) algorithm within a critic-actor RL framework, enabling real-time adaptation to changing environmental conditions and improving FOWT performance.

Methodology
Literature Review: Conduct a comprehensive review of existing wind turbine control strategies, reinforcement learning applications in wind turbine control [1-4], and floating offshore wind turbine challenges to establish a solid foundation for the research.

Control Algorithm Design: Develop an RL-based control algorithm that combines Incremental Dual Heuristic Programming (IDHP) [1] with a critic-actor structure to achieve power regulation and load mitigation in FOWTs. The algorithm will be designed to adapt in real-time based on online measurements and without the need for accurate analytical models, by adapting an incremental model [2,4].

Simulation and Validation: Utilize the FAST simulator, developed by NREL [5], to simulate the behaviour of FOWTs under various wind-wave conditions. The 5-MW NREL wind turbine on a spart floating platform (OC3-Hywind) is the chosen model. This model is available within FAST. Data will be generated in MATLAB using extensive simulation of the FOWT under different sea/wind states, the TurbSim software will be utilized to generate artificial turbulent wind data, and Jonswap/Pierson-Moskowitz spectrum code will be used to generate wave states. Implement the proposed RL-based control algorithm and assess its performance in reducing power fluctuations and mitigating loads and motions.

Performance Evaluation: Analyse and compare the performance of the proposed RL-based control strategy with existing control methods [2,3], considering power generation efficiency, structural integrity, load mitigation, and overall FOWT stability. The benchmark controller to compare against is the Reference Open-Source Controller (ROSCO) for Floating wind turbines, developed by NREL.

Training & Skills

The student will receive in-house training that is essential to mastering the technical aspects of the project. This includes guidance from supervisors Dr. M. Abdelrahman and Prof. J. Gilbert for FOWT modelling and control systems using MATLAB/Simulink. Additionally, mentoring by Dr. A. Moustafa will provide expertise in advanced deep learning and explainable AI techniques, ensuring the student develops the necessary skills to successfully execute the project.

References & Further Reading

[1]Zhou, Y., van Kampen, E.J. and Chu, Q.P., 2018.Incremental model based online dual heuristicprogramming for nonlinear adaptive control. ControlEngineering Practice, 73, pp.13-25.

[2]Jingjie Xie, Hongyang Dong, Xiaowei Zhao,
Data-driven torque and pitch control of wind turbinesvia reinforcement learning, Renewable Energy,Volume 215, 2023,

[3]J. Xie, H. Dong and X. Zhao, “Power Regulation andLoad Mitigation of Floating Wind Turbines viaReinforcement Learning,” in IEEE Transactions onAutomation Science and Engineering, doi:10.1109/TASE.2023.3295576.

[4]Xuerui Wang, Sihao Sun, Incremental fault-tolerant control for a hybrid quad-plane UAVsubjected to a complete rotor loss, bAerospaceScience and Technology, Volume 125,2022,

[5]Nrel OpenFAST Simulator.https://www.nrel.gov/wind/nwtc/openfast.html ,accessed on 24/08/2023.