上海交通大学学报 ›› 2022, Vol. 56 ›› Issue (6): 693-700.doi: 10.16183/j.cnki.jsjtu.2021.515

所属专题: 《上海交通大学学报》“新型电力系统与综合能源”专题(2022年1~6月)

• 新型电力系统与综合能源 •    下一篇

一种并网逆变器直流电容容值辨识方法

朱城昊1, 王晗1, 孙国歧2, 魏晓宾2, 王富文3, 蔡旭1()   

  1. 1.上海交通大学 电力传输与功率变换控制教育部重点实验室, 上海 200240
    2.山东德佑电气股份有限公司, 山东 淄博 255035
    3.陕西鲁能靖边风力发电有限责任公司,西安 718500
  • 收稿日期:2021-12-15 出版日期:2022-06-28 发布日期:2022-07-04
  • 通讯作者: 蔡旭 E-mail:xucai@sjtu.edu.cn
  • 作者简介:朱城昊(1994-),男,内蒙古自治区赤峰市人,硕士生,从事风电变流器故障诊断研究.
  • 基金资助:
    山东省重点研发计划(重大科技创新工程)(2019JZZY020804)

An Identification Method for DC-Link Capacitor Capacitance of Grid Connected Inverter

ZHU Chenghao1, WANG Han1, SUN Guoqi2, WEI Xiaobin2, WANG Fuwen3, CAI Xu1()   

  1. 1. Key Laboratory of Control of Power Transmission and Conversion of the Ministry of Education, Shanghai Jiao Tong University, Shanghai 200240
    2. Shandong Deyou Electric Co., Ltd., Zibo 255035, Shandong, China
    3. Shaanxi Luneng Jingbian Wind Power Co., Ltd., Xi’an 718500, China
  • Received:2021-12-15 Online:2022-06-28 Published:2022-07-04
  • Contact: CAI Xu E-mail:xucai@sjtu.edu.cn

摘要:

直流电容是并网逆变器中最易老化失效的器件之一,对其进行容值参数辨识可以及时发现并更换老化电容,有利于提高系统可靠性.提出一种基于预充电电路的并网逆变器直流电容容值辨识方法,通过分析预充电过程中电容容值与充电电流、电压的数值关系,结合历史运行数据,构建电容状态特征向量集合.利用该集合训练支持向量回归(SVR)模型,建立状态值与容值的回归预测关系,并采用粒子群优化(PSO)算法对模型进行优化,优化后的模型可用于直流电容容值辨识.仿真和实验结果表明,所提方法可对并网逆变器直流电容进行容值辨识,辨识误差小于0.95%.该方法无需增加硬件电路且不改变控制算法,具有一定的实用价值.

关键词: 并网逆变器, 直流电容, 容值辨识, 预充电电路, 支持向量回归

Abstract:

DC-link for the capacitor is one of the most vulnerable components of the grid connected converter, whose capacitance identification will help to improve the system reliability by finding and replacing the aging capacitor in time. An identification method for the DC-link capacitor capacitance of the grid connected inverter based on pre-charging circuit is proposed. By analyzing the relationship between the capacitance and the charging current, charging voltage during pre-charging process, and combining the historical operating data, the set of capacitance state feature vector is built. The support vector regression (SVR) model is trained and the regression prediction relationship between the state value and the capacitance is set. The model is optimized by using the particle swarm optimization (PSO) algorithm, which can be used for capacitance identification of the DC-link capacitor. Simulation and experiments results show that the proposed method can implement the accurate capacitance identification of the DC-link capacitor of the grid connected inverter, with an identification error of less than 0.95%. This method does not need to add hardware circuit and change the control algorithm, and has a certain practical value.

Key words: grid connected inverter, DC-link capacitor, capacitance identification, pre-charging circuit, support vector regression (SVR)

中图分类号: