Fault Diagnosis of Photovoltaic String Based on Graph Neural Network
The analysis of photovoltaic string operation status is the foundation of modeling and fault diagnosis,and fault diagnosis is an important means to ensure the safe and economic operation of photovoltaic systems.This article proposes a fault diagnosis model for adjacent photovoltaic strings based on graph neural networks,which monitors the operating status of photovoltaic systems by comparing the current and voltage data of photovoltaic strings during the historical window period.This method does not require the addition of new sensor equipment,and current and voltage data are directly obtained through the inverter.Six typical photovoltaic strings are selected for a certain photovoltaic power station,and common fault operation states and normal operation states are simulated.The graph neural network model is trained and tested based on simulation data.The test results show that although there are significant differences in the number,direction,and position of photovoltaic modules among the six typical strings,the trained graph neural network model can monitor and diagnose five common faults in real time.The results of comparative testing show that in the absence of weather data,the diagnostic accuracy of the graph neural network model is higher than the two commonly used diagnostic methods with higher accuracy.In addition,the graph neural network model can be extended to untrained photovoltaic strings and can maintain high diagnostic accuracy when multiple photovoltaic strings fail simultaneously.
国家科技期刊平台
NSTL
万方数据
