基于FasterR-CNN算法的电厂继电保护压板状态自动化检测方法  

Automated Detection Method for Power Plant Relay Protection Pressure Plate Status Based on Faster R-CNN Algorithm

作　　者：陈蒙 CHEN Meng(Xiexin Binhai Power Generation Co.,Ltd.,State Power Investment Corporation,Yancheng,Jiangsu 224500,China)

机构地区：[1]国家电投集团协鑫滨海发电有限公司,江苏盐城224500

出　　处：《自动化应用》2025年第4期105-107,111,共4页Automation Application

摘　　要：传统压板状态检测方法在面对结构复杂的继电保护压板时的检测效果并不理想。为此,基于FasterR-CNN算法,提出了一种新的电厂继电保护压板状态自动化检测方法。首先,根据压板状态条件逻辑,利用区域生成网络生成压板状态候选区域;然后,通过建立检索向量和压板状态特征矩阵的方式,检索压板状态特征,生成特征图;最后,将候选区域覆盖在由卷积神经网络生成的特征图上,使用FasterR-CNN算法识别压板状态,再通过融合处理输出检测结果。实验中,选择不同工况下的压板图像作为实验数据集,验证该方法在不同工况下的检测效果。实验结果表明,该方法的正确率平均值可达到99.57%,说明其检测效果良好。The traditional method of detecting the status of the pressure plate is not ideal when facing complex relay protection pressure plates.Therefore,based on the Faster R-CNN algorithm,this study proposes a new automatic detection method for the status of power plant relay protection pressure plates.Firstly,based on the logic of the pressure plate state condition,a region generation network is used to generate candidate regions for the pressure plate state.Then,by establishing retrieval vectors and pressure plate state feature matrices,the pressure plate state features are retrieved and a feature map is generated.Finally,the candidate regions are overlaid on the feature map generated by the convolutional neural network,and the Faster R-CNN algorithm is used to identify the pressure plate state.The detection results are then output through fusion processing.In the experiment,pressure plate images under different working conditions were selected as the experimental dataset to verify the detection performance of the method under different working conditions.The experiment results show that the average accuracy of this method can reach 99.57%,indicating that the detection effect of this method is good.

关 键 词：FasterR-CNN算法 电厂继电保护 压板状态 自动化检测 

分 类 号：TM133[电气工程—电工理论与新技术]