TY - GEN
T1 - Transfer Learning for Identifying Impedance Estimation in Voltage Source Inverters
AU - Zhang, Mengfan
AU - Wang, Xiongfei
AU - Yang, Dongsheng
AU - Cui, Zihao
AU - Christensen, Mads Grasboll
N1 - Publisher Copyright:
© 2020 IEEE.
Copyright:
Copyright 2020 Elsevier B.V., All rights reserved.
PY - 2020/10
Y1 - 2020/10
N2 - The black-box impedance model of the voltage source inverters (VSIs) can be directly identified at the converter terminal without access to its internal control details, which greatly facilitate the converter-grid interactions. However, since the converter is inherently a nonlinear system, the measured converter impedance model will change along with the operating point. As the limited data amount in practical industrial applications, the existing impedance identification method cannot capture this operating point-dependent feature of the impedance model. In this paper, the model-based transfer learning method is employed to generate the operating-point dependent impedance model. This method can significantly reduce the required data amount used in model training so that the machine learning-based method could be applied for the practical industrial application. The comparison results confirm the accuracy of the impedance model obtained by this data-driven impedance identification method.
AB - The black-box impedance model of the voltage source inverters (VSIs) can be directly identified at the converter terminal without access to its internal control details, which greatly facilitate the converter-grid interactions. However, since the converter is inherently a nonlinear system, the measured converter impedance model will change along with the operating point. As the limited data amount in practical industrial applications, the existing impedance identification method cannot capture this operating point-dependent feature of the impedance model. In this paper, the model-based transfer learning method is employed to generate the operating-point dependent impedance model. This method can significantly reduce the required data amount used in model training so that the machine learning-based method could be applied for the practical industrial application. The comparison results confirm the accuracy of the impedance model obtained by this data-driven impedance identification method.
KW - Deep neural network
KW - impedance identification
KW - operating point variation
KW - transfer learning
KW - voltage source inverter
UR - http://www.scopus.com/inward/record.url?scp=85097160224&partnerID=8YFLogxK
U2 - 10.1109/ECCE44975.2020.9236090
DO - 10.1109/ECCE44975.2020.9236090
M3 - Article in proceeding
AN - SCOPUS:85097160224
T3 - ECCE 2020 - IEEE Energy Conversion Congress and Exposition
SP - 6170
EP - 6174
BT - ECCE 2020 - IEEE Energy Conversion Congress and Exposition
PB - IEEE
T2 - 12th Annual IEEE Energy Conversion Congress and Exposition, ECCE 2020
Y2 - 11 October 2020 through 15 October 2020
ER -