基于改进深度强化学习算法的风电场功率优化控制

打开文本图片集

中图分类号：TP18；TM614 文献标识码：A 文章编号：2096-4706（2025）21-0044-09

Power Optimization Control of Wind Farms Based on Improved Deep ReinforcementLearning Algorithm

PENG Jian'，LIUDongwen² （1.Hunan TraditionalChineseMedical College，Zhuzhou 412012，China; 2.GuangdongNanfengElectric AutomationCo.，Ltd.，Meizhou 514523，Chinal

Abstract：To address the problem of the decrease inoverall power generation in wind farms caused by wake effects and thediffculty inaccuratelyestablishing physical modelsofwindfarms，aDRLcontrol methodthatsimultaneouslyconsiders wakesteeringcontrolandaxialinductioncontrolisproposed.Thismethodusestheyawangleandthrustcoeffcientof wind turbinesas variables toimplementcoordinatedcontrolof thewake，mitigatingtheimpactofwindturbine wakesandensuring the maximizationofoverallpower generation inwindfarms.Toimprovetheeficiencyof trainingsampleutilization，animproved DRLalgorithmis proposed inthispaper， which adopts a prioritized experiencereplay strategyon theunimproved DRLalgorithm to asigndiffrentsamplingprioritiesaccoding todifferencesintheimportanceofexperiences.SimulationontheWFSim wind farmplatformdemonstrates thatcomparedwithothermethods，theproposedcontrolstrategycansignificantlyincreasetheactive poweroutputof wind farms.Compared with the unimproved DRLalgorithm，the proposed improved DRLalgorithmcan avoid falling into local optima and enhance training effectiveness.

Keywords： wind farm; wake effect; Reinforcement Learning; yaw control; induction control

0 引言

风能作为一种清洁、可再生的能源，其大规模开发与利用主要依托于风电场的建设与运营。（剩余12592字）