摘要
针对传统依靠I-V特性的光伏故障离线诊断方法建模复杂、成本高,而在线人工智能检测方法可解释性差、对数据集敏感等问题,提出一种新的融合热斑时序演变特征的光伏阵列故障检测方法(HSTF),利用真实光伏平台进行遮挡实验,以获取不同热斑故障程度下的组件I-V输出特性曲线,并建立等效电路模型以模拟不同热斑程度下光伏阵列输出特征,共同构建故障诊断方法的输入特征向量。之后构建LightGBM-DBO模型对获取数据集进行训练,建立融合热斑时序演变特征的光伏阵列故障检测模型。利用光伏平台遮挡实验验证模型性能效果,将该方法与传统神经网络算法、决策树算法等其他检测算法进行对比,验证方法的准确性与可靠性。
Traditional offline photovoltaic fault diagnosis methods that rely on I-V characteristics suffer from problems such as complex modeling and high costs,while online artificial intelligence detection methods are plagued by poor interpretability and sensitivity to datasets.Aiming at these problems and shortcomings,a new photovoltaic array fault detection method(heat spot timing feature,HSTF)based on the time series evolution characteristics of hot spots is proposed.The real photovoltaic platform is used to carry out the occlusion experiment to obtain the I-V output characteristic curve of the component under different hot spot fault degrees.The equivalent circuit model is established to simulate the output characteristics of the photovoltaic array under different hot spot degrees,and the input feature vector of the fault diagnosis method is constructed together.Then,the LightGBM-DBO model is constructed to train the obtained data set,and a photovoltaic array fault detection model integrating the temporal evolution characteristics of hot spots is established.The photovoltaic platform occlusion experiment is used to verify the performance of the model.The method is compared with other detection algorithms such as traditional neural network algorithm and decision tree algorithm to verify the accuracy and reliability of the method.
作者
韩明轩
董红召
佘翊妮
陈炜烽
Han Mingxuan;Dong Hongzhao;She Yini;Chen Weifeng(College of Mechanical Engineering,Zhejiang University of Technology,Hangzhou 310014,China;Hangzhou Jiahe Electric Co.,Ltd.,Hangzhou 310014,China)
出处
《太阳能学报》
北大核心
2025年第8期269-277,共9页
Acta Energiae Solaris Sinica
基金
杭州市重大科技创新项目(2022AIZD0163)。
关键词
太阳电池
故障检测
电流电压特性
特征提取
热斑
solar cells
fault detection
current voltage characteristics
feature extraction
hot spot
