光伏组件输出特性的异常分类研究  

Research on the Anomaly Classification of Output Characteristics of PV Module

作　　者：史新科 杨成佳[1] SHI Xinke;YANG Chengjia(School of Electrical Engineering and computer Science,Jilin Jianzhu University,Changchun 130118,China)

机构地区：[1]吉林建筑大学电气与计算机学院,长春市130118

出　　处：《建筑电气》2024年第5期62-68,共7页Building Electricity

基　　金：吉林省教育厅科学技术研究项目,项目名称:基于全要素数据融合的公共照明智能控制系统研究,项目编号:JJKH20240366KJ。

摘　　要：光伏组件异常直接影响到发电量与组件的使用寿命,及时发现光伏组件异常并消除异常,将直接提高光伏组件发电效率。为准确检测光伏组件的异常状态,以Matlab平台为基础,模拟6种光伏组件异常状态,并分析输出特性,提出以输出特性曲线的五特征法判断异常类型。用Python语言搭建概率神经网络,并用粒子群优化算法优化平滑因子。用粒子群优化概率神经网络(PSO-PNN)模型训练异常数据,结果表明:概率神经网络对数据集较敏感,在较大的数据集,模型分类准确率高,能有效检测光伏组件异常状态。Anomalies in photovoltaic module(PV module)will directly affect the power generation and the service life of PV module.Identifying and eliminating anomalies in PV module timely will directly improve the power generation efficiency of PV module.In order to accurately detect the abnormal conditions of PV module,6 types of abnormal conditions of PV module are simulated based on Matlab platform and the output characteristics are analyzed,and the five-feature method for output characteristic curves are proposed to judge the type of abnormal conditions.Python language is used to establish the probabilistic neural network and particle swarm optimization algorithm is adopted to optimize the smoothfactor.T Theparticle swarm optimization-probabilistic neural network(PSO-PNN)model is used to train abnormal data.The result shows that,the probabilistic neural network is sensitive to the dataset.For a large dataset,the model classification has a high accuracy,which can effectively detect abnormal conditions of PV module.

关 键 词：光伏组件 MATLAB仿真 异常状态检测 U-I特性曲线 五特征法 特征提取 PSO-PNN网络 平滑因子 

分 类 号：TM615[电气工程—电力系统及自动化]