基于红外成像的变电站设备安全预警关键技术研究

发布时间：2018-02-01 04:54

本文关键词： 高压电气设备红外图像图像配准图像分割 SF6气体检测　出处：《西安科技大学》2014年博士论文　论文类型：学位论文

【摘要】：电力系统正向着智能电网方向发展，要求能够对变电站等电力设施进行在线监测和安全预警。但变电站故障检测还无法做到自动、及时地对设备故障报警。将基于红外成像的图像分析和视频分析技术，与变电站在线监测系统相结合，可达到事件检测实时报警，有助于提升变电站安全运行应急保障能力，具有重要的理论研究意义和工程价值。红外摄像机的远程温度监测系统还无法做到自动检测和及时的故障报警，提出了一种变电站电气设备红外遥视与安全预警系统，将红外与可见光摄像头安装在云台上，能够自动循环采集变电站的红外测温图像。针对所获得的遥视图像背景复杂、环境变化较大的问题，提出了对象分割的方法，将图像中的感兴趣设备进行初展开，也为系统实现数据库和故障诊断与预警专家系统奠定了良好基础。针对系统采集的变电站设备红外图像存在位置偏移、角度旋转和几何畸变等问题，首先采用基于相位相关和Harris角点匹配相结合的配准方法，对大部分偏差图像实现了图像配准；而对于个别难以配准的设备图像存在的失配问题，深入分析造成失配的原因，提出了改进的SIFT特征配准方法，根据相位相关法得出主运动方向以便减小搜索区域，从而既减小匹配搜索时间，也提高了匹配精度。实验表明，该方法能够成功配准所有测试图像，且能够满足系统实时性要求。变电站中变压器的油枕油位和变压器散热器是变电站中需要实时监测的重要设备，油位过低过高和散热片油路不通都会造成严重的安全隐患。针对红外油枕图像清晰度不高、油位分界线不明显的情况，提出了改进的分水岭分割算法，采用不同尺度的结构元素计算梯度放大图像中不同程度的弱边缘，同时采用极小值强制标定技术，有效地防止过分割。实验表明该算法能准确有效地分割出油枕边界和油位，精度满足系统要求。针对变压器散热器故障边界不清晰的问题，提出了一种改进的稳健模糊核聚类算法，通过引入核方法以及空间和灰度相似性测度增强算法的稳健性。实验表明该算法分割效果优于FCM、KFCM等模糊聚类分割方法，能较为清晰的分割出有故障的散热器片和位置。采用SF6气体绝缘的电力设备已大量投入电网运行，变电站内充气设备一旦发生漏气将造成严重的不良后果。为了实现变电站连续及时的无人监测，提出了一种自动检漏方法，利用SF6泄漏在红外视频中类似烟雾的动态特性，采用混合高斯背景建模的方法进行SF6气体的泄漏检测，并基于数学形态学方法去除斑点等噪声，找到疑烟区域，，然后在疑烟区域内分析和提取烟雾区别于其他运动物体的独特的动态特征，最终判断出泄漏区域。通过对多个红外气体成像检漏仪采集到的SF6视频进行泄漏检测，实验结果表明该方法能有效克服光线变化等背景扰动影响，较准确地实现泄漏点的自动监测和定位。
[Abstract]:The power system is developing in the direction of the intelligent power grid , which requires the online monitoring and safety early warning of the power facilities such as the transformer substation . However , the fault detection of the substation can not automatically and timely alarm the equipment failure . The infrared imaging - based image analysis and video analysis technology can be combined with the on - line monitoring system of the substation , so that the real - time alarm of the event detection can be achieved , and the safety operation emergency protection capability of the transformer substation can be realized , and the important theoretical research significance and engineering value are provided . The remote temperature monitoring system of infrared camera can not automatically detect and timely fault alarm , and puts forward an infrared remote vision and safety early warning system for electrical equipment of transformer substation . The infrared and visible light camera is mounted on the cloud platform , and the infrared temperature measuring image of the substation can be automatically circulated . Aiming at the problems of position deviation , angle rotation and geometric distortion of the infrared image of substation equipment acquired by the system , a registration method based on phase correlation and Harris corner matching is adopted to realize image registration for most of the deviation images . An improved watershed segmentation algorithm is proposed in this paper . It is proved that the algorithm can accurately and effectively separate the boundary and the oil level of the oil conservator and meet the requirements of the system . The experimental results show that the algorithm can effectively and effectively separate the boundary and the oil level of the oil conservator . In order to realize the continuous and timely unmanned monitoring of the substation , an automatic leak detection method is put forward . In order to realize the continuous and timely unmanned monitoring of the transformer substation , an automatic leak detection method is put forward . In order to realize the continuous and timely unmanned monitoring of the substation , the leakage detection of SF6 gas is carried out by using the method of Gaussian background modeling . The leakage detection is carried out by using the method of mathematical morphology to distinguish the smoke from other moving objects . The experimental results show that the method can effectively overcome the influence of background disturbance such as light change and the like , and realize the automatic monitoring and positioning of the leakage point more accurately .

【学位授予单位】：西安科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TD76

【参考文献】