基于图神经网络的光伏组串故障诊断

Fault Diagnosis of Photovoltaic String Based on Graph Neural Network

光伏组串运行状态分析是建模和故障诊断的基础,而故障诊断是保障光伏系统安全经济运行的重要手段。提出基于图神经网络的邻近光伏组串故障诊断模型,通过比较历史窗口期内光伏组串的电流和电压数据来监测光伏系统的运行状态。该方法无需新增传感器设备,电流和电压数据直接通过逆变器获得。选取某光伏电站6组典型光伏组串,仿真模拟了常见故障运行状态和正常运行状态,并结合仿真数据对图神经网络模型进行了训练和测试。测试结果显示,尽管6个典型组串在光伏组件数量、方向、位置等方面存在较大差异,训练好的图神经网络模型仍然可对5种常见故障进行实时监测和诊断。对比测试的结果显示:在没有天气数据的情况下,图神经网络模型的诊断准确率高于当前精度较高的两种常见诊断方法;此外,图神经网络模型可推广到未经训练的光伏组串,并且能在多个光伏组串同时发生故障时保持较高的诊断准确率。

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.