基于Faster R-CNN的光伏发电容量估计方法研究

Research on Photovoltaic Capacity Estimation Methods Based on Faster R-CNN

中文关键词:光伏组件  目标检测  容量估计  智能识别

keywords:photovoltaic module  object detection  capacity estimation  intelligent recognition

中文摘要:全球对可再生能源需求的增长，使光伏电站在现代能源系统中占据重要地位，其快速发展也使运维管理压力日益严峻，而光伏发电容量估计正是智能化运维的基础。本文提出一种基于快速区域卷积神经网络（fast region-based convolutional neural networks，Faster R-CNN）的光伏发电容量估计方法。首先，对不同种类光伏组件的容量及影响进行分析；其次，采用基于视觉几何组16层网络（visual geometry group 16-layer network，VGG16）为主干的Faster R-CNN目标检测模型，对不同类型的光伏组件进行识别、分类和统计；最后，利用容量估计公式对检测结果的容量进行估计，从而实现光伏电站整体发电容量的估计。实验结果表明：所提算法能够有效区分不同种类的光伏组件，并能通过容量估计公式来计算光伏电站的整体容量，对光伏电站的智能化运维具有较大的指导意义。

Abstract:The global growth in demand for renewable energy has made photovoltaic （PV） power plants an important part of modern energy systems. Their rapid development has also brought mounting pressure on operation and maintenance （O&M） management，and the estimation of PV power generation capacity is the foundation of intelligent O&M. This paper proposes a PV power generation capacity estimation method based on faster region-based convolutional neural networks （Faster R-CNN）. First，the capacities and influencing factors of different types of PV modules are analyzed. Second，a Faster R-CNN target detection model with the visual geometry group 16-layer network （VGG16） as the backbone is adopted to identify，classify and count different types of PV modules. Finally，the capacity corresponding to the detection results is estimated using a capacity estimation formula，thereby realizing the estimation of the overall power generation capacity of PV power plants. Experimental results show that the proposed algorithm can effectively distinguish different types of PV modules and calculate the overall capacity of PV power plants via the capacity estimation formula，which has important guiding significance for the intelligent O&M of PV power plants.