Open AccessOptimal Sensor Placement and Fault Diagnosis Model of PV Array of Photovoltaic Power Stations Based on XGBoost
Author(s) -
Hai Wang,
Fuchun Sun
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/661/1/012025
Subject(s) - photovoltaic system , fault (geology) , power (physics) , process (computing) , computer science , maximum power point tracking , fault detection and isolation , reliability engineering , real time computing , engineering , electronic engineering , electrical engineering , artificial intelligence , voltage , inverter , actuator , physics , quantum mechanics , seismology , geology , operating system
As an important part of photovoltaic power stations, daily monitoring and maintenance of photovoltaic array are quite necessary. In order to be able to accurately locate the faulty module and diagnose fault types. The fault diagnosis model which based on XGBoost optimized by GridSearchCV on optimal sensor placement is proposed in this article. First, the change laws of external electrical characteristics of photovoltaic modules under the control of MPPT technology are analyzed in different fault states. On this basis, the parameters which can locate the faulty PV modules and the input of the XGBoost PV fault diagnosis model are obtained. Finally, during the process of the simulation and experiment, the failure data measured by the multi-sensor method can be used as a positioning quantity to locate the faulty module. The comparison results with the other three algorithms (LR, RF, XGBoost) prove that the performance of GS-XGBoost algorithm has great advantages in judging PV fault types (short circuit, open circuit, aging).