文档名：基于机器学习的带被动阻尼直流微电网系统的稳定性检测
摘要：直流微电网中恒功率负荷(CPL)具有负阻尼特性,该特性会降低系统稳定性.为此,通过在滤波器上添加被动阻尼来增强直流微电网系统的稳定性,并提出一种基于机器学习的方法来检测带被动阻尼直流微电网系统的稳定性.首先,建立带被动阻尼直流微电网系统的小信号模型,以此来确定影响系统稳定性的参数.其次,以所选系统参数为变量建立仿真场景,以此来获取用于机器学习算法训练的数据集.再次,提出一种基于轻量型梯度提升机(LGBM)的直流微电网稳定性检测模型,并采用沙普利加解释法(SHAP)分析所选参数对LGBM预测结果和直流微电网系统稳定性的影响.最后,通过仿真和硬件在环实验验证所提方法的有效性和优越性.
Abstract：InDCmicrogrids,thenegativeimpedancecharacteristicsofconstantpowerloads(CPL)canreducesystemdamping,leadingtosystemcollapse.ToenhancethestabilityofDCmicrogridsystems,researchersproposevariousactivedampingadditionmethods,includingvirtualactiveresistancecompensation,synchronousbuckcircuit,andoutputcurrentfeedforward,whichrequirecomplexcontrolalgorithmsandsignalprocessing.Inaddition,thesemethodsmayreduceloadperformanceandaffectthesystem'sdynamicresponse.Comparedwithactivedampers,passivedampingonlyrequiressimplepassivecomponentswithoutcomplexcontrolalgorithmsandsignalprocessingtechnology.Accordingly,theimplementationissimple,andthecostislow.ToanalyzethestabilityoftheDCmicrogridafteraddingdampers,traditionalmethodsbasedonthemechanismmodelignorethestructureinfluence,andthemodelerrorgraduallyincreaseswiththeincreaseofsystemdimensionandcomplexity.Thus,thispaperproposesamachinelearning-basedmethodtodetectthestabilityofDCmicrogridswithpassivedamping.Thismethodcaneffectivelyimprovetheerrorproblemofthemechanismmodel,comprehensivelyconsidertherelationshipbetweenlocalandglobalstability,andrealizetherapidandaccurateonlinedetectionofthepassivedampedDCmicrogridstablestate.Firstly,thispaperestablishesanimpedancemodelfortheDCmicrogridsystemwithpassivedamping,usingtheRouth-Hurwitzcriteriontodeterminetheparametersthatinfluencesystemstability.Secondly,simulationscenariosareconductedbasedontheselectedsystemparameterstoobtainsampledatafortrainingthemachinelearningalgorithm.Astabilitydetectionmodelbasedonthelightgradientboostingmachine(LGBM)isintroducedforDCmicrogrids.TheinfluenceoftheselectedparametersontheLGBMpredictionandthestabilityoftheDCmicrogridsystemisfurtherexploredusingShapleyadditiveexplanations(SHAP).Finally,theeffectivenessandsuperiorityoftheproposedmethodarevalidatedthroughsimulationsandhardware-in-the-loopexperiments.TheresultsshowthattheCPLcriticalvalueisincreasedby31kW,15kW,and22kW,respectively,whenRCparallel,RLparallel,andRLCseriesdampersareaddedtothesingleenergystorageDCmicrogrid.TheRLseriesdamperonlyimprovestheCPLcriticalvalueby3kW.AfterestablishingsixsimulationscenarioswithRCparallel,RLparallel,andRLCseriesdampers,theproposedLGBMachievespredictionaccuracyofmorethan96.5%andafalsealarmrateoflessthan2.5%whenthereisuncertaintyinsourceandload.Comparedwithtraditionalmathematical,machinelearning,anddeeplearningmethods,theproposedmethodhasadvantagesinpredictionaccuracyandcomputationalspeed.Amongtheselectedsystemparameterfeatures,CPL,lineinductance,anddamperinductancearenegativelycorrelatedwiththestabilityoftheDCmicrogrid.Incontrast,linecapacitanceanddampercapacitancearepositivelycorrelatedwiththestabilityofDCmicrogrids.Thefollowingconclusionscanbedrawn.(1)RCparalleldampers,RLparalleldampers,andRLCseriesdamperssignificantlyenhancethestabilityoftheDCmicrogridsystemwithCPL.However,theRLseriesdamperdoesnotshowanidealimprovementeffect.(2)TheproposedLGBMmodelconsistentlyachievesthehighestpredictionaccuracyacrossmultiplescenariosandreducescomputationaltime.(3)TheSHAPmethodeffectivelyexplainstheimpactofeachinputfeatureonthestabilityoftheLGBMdetectionDCmicrogrid,providinginterpretabilitytotheLGBMmodel.
作者：刘笑 杨建 李力 董密 宋冬然Author：LiuXiao YangJian LiLi DongMi SongDongran
作者单位：中南大学自动化学院长沙410083
刊名：电工技术学报
Journal：TransactionsofChinaElectrotechnicalSociety
年，卷(期)：2024, 39(8)
分类号：TM712
关键词：直流微电网 稳定性检测 被动阻尼 轻量型梯度提升机 沙普利加解释
Keywords：DCmicrogrid stabilitydetection passivedamping lightgradientboostingmachine Shapleyadditiveexplanations
机标分类号：TM712TP391TM46
在线出版日期：2024年4月29日
基金项目：国家自然科学基金，湖南省自然科学基金资助项目基于机器学习的带被动阻尼直流微电网系统的稳定性检测[
期刊论文] 电工技术学报--2024, 39(8)刘笑 杨建 李力 董密 宋冬然直流微电网中恒功率负荷(CPL)具有负阻尼特性,该特性会降低系统稳定性.为此,通过在滤波器上添加被动阻尼来增强直流微电网系统的稳定性,并提出一种基于机器学习的方法来检测带被动阻尼直流微电网系统的稳定性.首先,建立...参考文献和引证文献
参考文献
引证文献
本文读者也读过
相似文献
相关博文
基于机器学习的带被动阻尼直流微电网系统的稳定性检测 Stability Detection of DC Microgrid Systems with Passive Damping Based on Machine Learning
基于机器学习的带被动阻尼直流微电网系统的稳定性检测.pdf
- 文件大小:
- 2.87 MB
- 下载次数:
- 60
-
高速下载
|
|
