基于多层次神经网络与多点位的核电站高压开关设备无线测温方法  

Wireless Temperature Measurement Method of High-Voltage Switchgear in Nuclear Power Plant Based on Multi-Level Neural Network And Multi-Point

作　　者：曹双华 翁黎明 陈锋 刘双 洪诗鑫 凌世情 CAO Shuang-hua;WENG II-ming;CHEN feng;LIU shuang;HONG Shi-xin;LING Shi-qing(Nuclear Power Operations Research(Shanghai)Co.,Ltd.,Shanghai 200125;Sanmen Nuclear Power Co.,Ltd.,Taizhou 317100;Cnnp Guodian Zhangzhou Energy Co.,Ltd.,Zhangzhou 363300)

机构地区：[1]核电运行研究(上海)有限公司,上海200125 [2]三门核电有限公司,台州317100 [3]中核国电漳州有限公司,漳州363300

出　　处：《环境技术》2024年第5期182-187,共6页Environmental Technology

摘　　要：光线条件、视角和拍摄距离等因素影响下,温感成像质量较差,且数据泄露或被篡改的风险较大,高压开关设备温度信息存在偏差或不稳定线性,导致无线测温过程的温度拟合增益系数较低。为此,提出基于多层次神经网络与多点位的核电站高压开关设备无线测温方法。通过放射性结构的拍摄点位布设方案,获取温感图像,并对其进行灰度阈值分割处理;采用领域平均法,结合多点位需求,对获取到的二值温感图像进行平滑处理,并构建多层次神经网络,将二值温感图像作为输入,通过卷积层以及激活函数,增强其非线性映射能力,分析与计算具有平滑期望值的无线测温结果即设备表面的温度值预测值。实验结果表明,该方法应用后,测温结果的温度拟合增益系数更高,具备更为理想的测温精度,为核电站高压开关设备的无线测温提供了准确、可靠的技术支持。Under the influence of factors such as lighting conditions,viewing angle,and shooting distance,the quality of temperature sensing imaging is poor,and there is a high risk of data leakage or tampering.The temperature information of high-voltage switchgear has deviations or unstable linearity,resulting in a low temperature fitting gain coefficient in the wireless temperature measurement process,.To this end,a wireless temperature measurement method for high-voltage switchgear in nuclear power plants based on multi-level neural networks and multipoint positioning is proposed.Obtain temperature sensing images through the placement of shooting points for radioactive structures,and perform grayscale threshold segmentation processing on them;Using domain averaging method and combining with multi-point requirements,the obtained binary temperature sensing image is smoothed,and a multi-level neural network is constructed.The binary temperature sensing image is used as input,and its nonlinear mapping ability is enhanced through convolutional layers and activation functions.The wireless temperature measurement results with smooth expected values,namely the predicted surface temperature values of the device,are analyzed and calculated.The experimental results show that after the application of this method,the temperature fitting gain coefficient of the temperature measurement results is higher,and it has more ideal temperature measurement accuracy,providing accurate and reliable technical support for wireless temperature measurement of high-voltage switchgear in nuclear power plants.

关 键 词：神经网络 核电站 开关设备 无线测温 温感图像 

分 类 号：TM721[电气工程—电力系统及自动化]