Dynamic Optimal Power Flow Method Based on Reinforcement Learning for Offshore Wind Farms Considering Multiple Points of Common Coupling  

作　　者：Yang Fu Zixu Ren Shurong Wei Lingling Huang Fangxing Li Yang Liu 

机构地区：[1]Engineering Research Center of Offshore Wind Technology Ministry of Education(Shanghai University of Electric Power),Shanghai 200090,China [2]Department of Electrical Power Engineering,Shanghai University of Electric Power,Shanghai 200090,China [3]Department of Electrical Engineering and Computer Science,The University of Tennessee,Knoxville,TN 37996,USA

出　　处：《Journal of Modern Power Systems and Clean Energy》2024年第6期1749-1759,共11页现代电力系统与清洁能源学报(英文)

基　　金：This work was supported in part by the National Natural Science Foundation of China(No.52377063);the Shanghai Action Plan for Science,Technology and Innovation(No.22dz1206100);the Major Natural Science Project of Shanghai Municipal Education Commission(No.2021-01-07-00-07-E00122);the Program for Professor of Special Appointment(Eastern Scholar)at Shanghai Institutions of Higher Learning(No.TP2020066).

摘　　要：The widespread adoption of renewable energy sources presents significant challenges for power system dis-patching.This paper proposes a dynamic optimal power flow(DOPF)method based on reinforcement learning(RL)to ad-dress the dispatching challenges.The proposed method consid-ers a scenario where large-scale offshore wind farms are inter-connected and have access to an onshore power grid through multiple points of common coupling(PCCs).First,the opera-tional area model of the offshore power grid at the PCCs is es-tablished by combining the prediction results and the transmis-sion capacity limit of the offshore power grid.Built upon this,a dynamic optimization model of the power system and its RL en-vironment are constructed with the consideration of offshore power dispatching constraints.Then,an improved algorithm based on the conditional generative adversarial network(CGAN)and the soft actor-critic(SAC)algorithm is proposed.By analyzing an improved IEEE 118-node system,the proposed method proves to have the advantage of economy over a longer timescale.The resulting strategy satisfies power system opera-tion constraints,effectively addressing the constraint problem of action space of RL,and it has the added benefit of faster so-lution speeds.

关 键 词：Offshore power grid optimal scheduling dy-namic optimal power flow(DOPF) reinforcement learning(RL) renewable energy 

分 类 号：TM614[电气工程—电力系统及自动化] TP181[自动化与计算机技术—控制理论与控制工程]