基于MADRL算法的海上風電場功率與載荷聯合優化
趙偉康,張宇琪,唐淵
(湖南工業大學 交通與電氣工程學院,湖南 株洲 412007)
摘 要 :針對海上風電場尾流損失明顯和疲勞損耗分布不均勻導致風電場維護頻率高的問題,提出了一種基于多智能體深度強化學習 (MADRL) 的風電場控制策略。通過分析風機的發電功率與疲勞載荷, 建立發電量與疲勞損耗的衡量模型,明確控制變量與狀態變量 ;再根據風機之間的氣動耦合關系進行分組,構建MADRL優化控制框架,將全部風機之間的合作轉變為組內合作加組間合作模式。在WFSim風電場模型中采用MADRL算法進行多目標優化求解,結果表明,所提策略能在風況變化的情況下有效減輕尾 流效應帶來的影響,在提升風電場整體發電效率的同時平衡機組間的疲勞損耗。
關鍵詞 : 海上風電場 ;尾流效應 ;多智能體深度強化學習 ;疲勞損耗 ;發電效率
中圖分類號 :TM614 ;TM714 文獻標識碼 :A 文章編號 :1007-3175(2025)12-0009-07
Joint Optimization of Power and Load in Offshore Wind Farms Based on MADRL Algorithm
ZHAO Wei-kang, ZHANG Yu-qi, TANG Yuan
(School of Traffic and Electrical Engineering, Hunan University of Technology, Zhuzhou 412007, China)
Abstract: To address the high maintenance frequency of offshore wind farms caused by significant wake losses and uneven fatigue damage distribution, a wind farm control strategy based on multi-agent deep reinforcement learning (MADRL) is proposed. By analyzing the relationship between turbine power generation and fatigue loads, a measurement model linking power output and fatigue damage is established to define control variables and state variables. Wind turbines are grouped based on aerodynamic coupling relationships, establishing a MADRL optimization control framework that transforms inter-turbine cooperation into a hybrid model of intra-group and inter-group collaboration. Multi-objective optimization using the MADRL algorithm is performed within the WFSim wind farm model. Results demonstrate that the proposed strategy effectively mitigates wake effects under varying wind conditions, simultaneously enhancing overall power generation efficiency while balancing fatigue losses across turbines.
Key words: offshore wind farm; wake effect; multi-agent deep reinforcement learning; fatigue loss; power generation efficiency
參考文獻
[1] Global Wind Energy Council.Global offshore wind report 2025[R].Brussels :GWEC,2025.
[2] 陳婕 .考慮尾流效應的風電場多目標分層隨機控制 [D]. 北京 :華北電力大學,2023.
[3] 魏賞賞,李智寒,陳一凱,等 . 基于狀態擴張輸入輸出動態模態分解的風力機尾流降階模型 [J]. 太陽能學 報,2024,45(10) :580-587.
[4] 樊濤,李剛,馬晨陽,等 . 考慮經濟性和舒適性的海上風電運維規劃優化方法 [J]. 船舶工程,2025, 47(4) :160-169.
[5] MUNOZ-PALOMEQUE E, SIERRA-GARCIAJ E, SANTOS M. Wind turbine maximum power pointtracking control based on unsupervised neural networks[J].Journal of Computational Design and Engineering,2023,10(1) :108-121.
[6] 蔡瑋,胡陽,劉吉臻 . 計及尾流的風電場智能等效建模及偏航優化控制 [J]. 動力工程學報,2024, 44(7) :1051-1059.
[7] 王冠朝,霍雨翀,李群,等 . 基于深度強化學習與 改進Jensen模型的風電場功率優化 [J]. 中國電力, 2025,58(4) :78-89.
[8] 褚夢珂,劉春,張勇,等 . 基于多智能體深度強化學習算法的風電場偏航控制策略 [J]. 上海電力大學學報,2025,41(4) :310-317.
[9] YAO Q, MA B, ZHAO T, et al.Optimized active power dispatching of wind farms considering data-driven fatigue load suppression[J].IEEE Transactions on Sustainable Energy,2022, 14(1) :371-380.
[10] 葛暢,閻潔,劉永前,等 . 海上風電場運行控制維護關鍵技術綜述 [J]. 中國電機工程學報,2022, 42(12) :4278-4291.
[11] 蔡瑋,胡陽,劉吉臻 . 基于尾流判定的風電場偏航角分群優化控制 [J]. 動力工程學報,2024,44(8) : 1234-1243.
[12] 張勇,劉春,褚夢珂,等 . 計及尾流的改進深度確定性策略梯度風電場功率優化控制策略 [J]. 電力系統及 其自動化學報,2025,37(2) :68-77.
[13] 陳皓勇,席松濤 . 海上風電成本構成及價格機制 [J]. 風能,2022(1) :12-15.
[14] BOERSMA S, DOEKEMEIJER B, VALI M, et al.A control-oriented dynamic wind farm model: WFSim[J].Wind Energy Science,2018,3(1) :75-95.
[15] HE R, YANG H, LU L.Optimal yaw strategy and fatigue analysis of wind turbines under the combined effects of wake and yaw control[J].Applied Energy,2023,337 :120878.
[16] 許曉玥 . 考慮風機疲勞的風電場功率優化策略 [D]. 長沙 :湖南大學,2021.
[17] WANG X, ZHOU J, QIN B, et al.Coordinated power smoothing control strategy of multi-wind turbines and energy storage systems in wind farm based on MADRL[J].IEEE Transactions on Sustainable Energy,2023,15(1) :368-380.
[18] JONKMAN J, BUTTERFIELD S, MUSIAL W, et al. Definition of a 5-MW reference wind turbine for offshore system development[R].Golden :National Renewable Energy Laboratory(NREL),2009.
[19] 閆曄 . 考慮風電不確定性的分時電價研究 [D]. 西安 : 西安理工大學,2020.