摘要
为提升光伏阵列故障诊断的准确率,提出一种基于改进长鼻浣熊优化算法优化极限学习机的光伏阵列故障诊断方法。首先,分析阵列中光伏组件在发生故障时的输出特性,选择合适的故障特征;其次,针对极限学习机在光伏阵列故障分类时初始权值和阈值的随机性问题,采用长鼻浣熊优化算法求解最优的初始权重和阈值;进一步地,针对长鼻浣熊算法初始参数的随机性和全局搜索能力的局限性问题,通过Circle混沌映射、莱维飞行和动态折射反向学习对该算法进行优化,提高寻优精度和速度;最后,结合光伏阵列故障实验数据,验证故障诊断模型的分类效果。结果表明,对于训练集和测试集数据,该诊断模型提高了故障分类精度,诊断率分别达到100%和98.33%,优于传统极限学习机、BP神经网络、支持向量机和卷积神经网络故障诊断的准确率。
In this article,an improved Coati Optimization Algorithm is proposed to optimize the diagnosis method of Extreme Learning Machine to achieve accurate classification of PV array failures.Firstly,the output characteristics of PV panels failing in the array are analyzed and appropriate fault characteristics are selected.Secondly,Coati Optimization Algorithm is utilized to solve the optimal initial weights and thresholds by exploiting the stochasticity of the initial weights and thresholds of Extreme Learning Machine for PV array fault classification.Furthermore,both the randomness of the initial parameters and the limitation of global search ability of Coati Optimization Algorithm are optimized by circle chaotic mapping,levy flight and dynamic refractive inverse learning,which greatly reduces the training time and improves the optimization search accuracy of the algorithm.Finally,the classification performance of the failure diagnostic model will be validated using experimental failure data of PV array,the corresponding results indicate that the diagnostic model improves the fault classification accuracy to 100%and 98.33%for the training set and test set,respectively,which is superior to the traditional extreme learning machine,BP neural network,support vector machine and convolutional neural networks.
作者
张健
赵咪
黄毅
李景云
Zhang Jian;Zhao Mi;Huang Yi;Li Jingyun(School of Mechanical and Electrical Engineering,Shihezi University,Shihezi 832003,China;Xinjiang Tianfu Jinyang New Energy Co.,Ltd,Shihezi 832000,China)
出处
《太阳能学报》
北大核心
2025年第1期632-640,共9页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(62363030)
石河子大学国际科技合作推进计划项目(GJHZ202108)。
关键词
光伏组件
故障分析
特征选择
监督学习
极限学习机
改进长鼻浣熊优化算法
PV panels
failure analysis
feature selection
supervised learning
extreme learning machine
improved coati optimization algorithm
