摘要
针对传统的光伏阵列故障诊断方法准确率低、模型性能差以及光伏I-V曲线数据利用率低的问题,提出基于HPOCatBoost的光伏阵列故障诊断模型。首先,利用光伏阵列模型深入研究短路、开路、老化、阴影遮挡和环境因素(温度、太阳辐照度)对I-V曲线变化的影响,并对其输出特性和故障成因进行系统分析;其次,通过Ordered TS编码来解决CatBoost中目标泄露导致预测偏移的问题,提高诊断模型的泛化能力;最后,CatBoost模型的性能受部分超参数的影响,故提出采用猎人猎物算法(HPO)对模型的关键超参数(树的数量、树的深度和学习率等)进行优化,进一步提升其在故障诊断上的性能表现,并对运行结果和实际光伏平台实验数据进行分析。实验结果表明,该模型的诊断准确率为99.5%,且相较于优化前的CatBoost模型,模型整体的准确率提高3.4%。
Aiming at the problems of low accuracy,poor model performance and low utilization of photovoltaic(PV)I-V curve data in traditional PV array fault diagnosis methods,this study proposes a PV array fault diagnosis model based on HPO-CatBoost.Firstly,the PV array model is used to deeply study the effects of short circuit,open circuit,aging,shading and environmental factors(temperature,irradiance)on the changes of I-V curves and systematically analyze their output characteristics and fault causes.Secondly,the problem of prediction bias due to target leakage in CatBoost is solved by Ordered TS coding to improve the generalization ability of the diagnostic model.Finally,the performance of CatBoost model is affected by some hyperparameters,so it is proposed to use hunter-prey optimizer(HPO)to optimize the key hyperparameters of the model(number of trees,tree depth and learning rate,etc.)to further improve its performance in fault diagnosis,and analyze the operation results and the experimental data of the actual PV platform.The experimental results show that the diagnostic accuracy of the model is 99.5%,and the overall accuracy of the model is improved by 3.4%compared to the CatBoost model before optimization..
作者
彭自然
许怀顺
肖伸平
肖满生
Peng Ziran;Xu Huaishun;Xiao Shenping;Xiao Mansheng(School of Electrical and Information Engineering,Hunan University of Technology,Zhuzhou 412007,China;Hunan Provincial Key Laboratory of Electric Drive Control and Intelligent Equipment,Zhuzhou 412007,China;School of Computer Science,Hunan University of Technology,Zhuzhou 412007,China)
出处
《太阳能学报》
北大核心
2025年第7期663-673,共11页
Acta Energiae Solaris Sinica
基金
国家重点研发计划(2019YFE0122600)
湖南省教育厅重点科研项目(22A0423)
湖南省自科科学基金(2023JJ60267,2022JJ50073)。
