{"title":"Parameter design of grid-tied inverter using reinforcement learning","authors":"Z. Wang, W. Wang","doi":"10.1049/icp.2021.2547","DOIUrl":null,"url":null,"abstract":"Grid-tied inverters have been widely adopted in distributed generation systems. The output current distortion and oscillation occur under certain conditions especially in the weak grid and multi-parallel inverters. To maintain the stability and ensure the dynamic and steady performance of current tracking, the controller parameters required iterative design and evaluation. Taking single phase inverter as an example, traditional design approaches are discussed first. Then, considering the difficulties in the parameter design process, deep reinforcement learning (DRL) is introduced to explore the optimal combination of controller parameters, by which the parameters of controllers can be automatically tuned. The realization of the DRL approach is discussed in detail and simulation validates the correctness of the attempt.","PeriodicalId":242596,"journal":{"name":"2021 Annual Meeting of CSEE Study Committee of HVDC and Power Electronics (HVDC 2021)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"1900-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"2021 Annual Meeting of CSEE Study Committee of HVDC and Power Electronics (HVDC 2021)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1049/icp.2021.2547","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Grid-tied inverters have been widely adopted in distributed generation systems. The output current distortion and oscillation occur under certain conditions especially in the weak grid and multi-parallel inverters. To maintain the stability and ensure the dynamic and steady performance of current tracking, the controller parameters required iterative design and evaluation. Taking single phase inverter as an example, traditional design approaches are discussed first. Then, considering the difficulties in the parameter design process, deep reinforcement learning (DRL) is introduced to explore the optimal combination of controller parameters, by which the parameters of controllers can be automatically tuned. The realization of the DRL approach is discussed in detail and simulation validates the correctness of the attempt.