基于猎人-猎物优化支持向量机的光伏阵列故障识别研究
Research on Fault Diagnosis of Photovoltaic Array Based on Hunter-prey Optimization Support Vector Machine
周恒 1肖文波 1李勇波 1吴华明1
作者信息
- 1. 南昌航空大学 测试与光电工程学院,南昌 330063
- 折叠
摘要
为提升光伏电站的运行效率,对光伏阵列的故障识别方法展开了深入研究,并建立基于猎人-猎物优化支持向量机(HPO-SVM)的故障识别方法.该方法以光伏阵列模型的 5个参数作为故障特征向量对光伏阵列的故障进行识别,基于仿真数据和实验数据评估该方法的有效性和可靠性.在基于仿真数据进行故障识别时,HPO-SVM的识别准确率达到了99.555 6%,十折交叉验证的平均准确率为 98.264 1%.与支持向量机(SVM)相比,HPO-SVM的准确率分别提高了 8.444 5%和 8.608 6%.在基于实验数据进行故障识别时,HPO-SVM的识别准确率达到了 98.064 5%,十折交叉验证的平均准确率为94.299 5%.相较于SVM,HPO-SVM的准确率分别提高了 7.741 9%和 4.570 2%.结果表明,HPO-SVM对光伏阵列的故障识别具有较高的准确度、可靠性,并具有良好的泛化性能.
Abstract
To enhance the operational efficiency of photovoltaic power plants,extensive research has been conducted on fault detection methods for photovoltaic arrays,and a fault detection method based on hunter-prey optimization support vector machine(HPO-SVM)has been established.This method utilizes five parameters of the photovoltaic array model as fault feature vectors for fault detection,and evaluates the effectiveness and reliability of the method based on simulation data and experimental data.When conducting fault detection based on simulation data,the recognition accuracy of HPO-SVM reached 99.555 6%,with an average accuracy of 98.264 1%in 10-fold cross-validation.Compared to support vector machine(SVM),the accuracy of HPO-SVM improved by 8.444 5%and 8.608 6%respectively.Similarly,on experimental data,the HPO-SVM achieves a recognition accuracy of 98.064 5%,with an average accuracy of 94.299 5%in 10-fold cross-validation,surpassing the SVM method by 7.741 9%and 4.570 2%respectively.These results highlight the superior accuracy,reliability,and generalization performance of the HPO-SVM approach.
关键词
光伏阵列/故障识别/猎人-猎物优化/支持向量机/准确率Key words
photovoltaic array/fault identification/hunter-prey optimization/support vector machine/accuracy rate引用本文复制引用
基金项目
国家自然科学基金(12064027)
国家自然科学基金(62065014)
江西省教育厅科技项目(GJJ2204302)
江西省高层次高技能领军人才培养工程入选项目(2022-63号)
九江市浔城英才项目(2024)()
南昌航空大学校级教学改革项目(JY22077)
出版年
2024
NETL
NSTL
万方数据