基于红外图像的电力变压器油位自动检测方法

Automatic Detection of Oil Level of Electric Power Transformers Using Infrared Image

为了自动准确地检测出变压器油位,提出了一种基于红外图像的检测方法。先对获得的变压器油枕红外图像进行边缘检测,得到油枕边缘点。在此基础上,采取迭代椭圆拟合的方法得到油枕的位置和形状,即在每次迭代中根据所保留的边缘点进行椭圆拟合,并删去距离椭圆较远的边缘点,直至相邻两次迭代得到的椭圆的差别满足要求为止。在油枕内取多条测温线,根据有油与无油部分存在温度差的特点,微分求出各测温线上灰度变化最大的分界点,迭代采用直线拟合并剔除偏差较大的点,得到油位分界线。最后计算椭圆中有油与无油部分的面积,将面积归算至油枕端面圆形中得到最终的油位值。实验证明所提出的方法是可行的,能够自动检测出变压器油位,可应用于无人值守变电站电力设备红外巡检与预警系统。

In order to detect the oil level of electrical power transformers in the substations accurately and automatically,a method based on the infrared image was put forward. The edge points of the oil conservator of the electrical power transformer were identified by the edge detection process on the infrared image of the transformer oil conservator. An iterative elliptic fitting approach was used to obtain the location and the shape of the oil conservator. The points far away from the obtained ellipse edge were eliminated in each iteration step until the difference of the ellipse between two adjacent iterations was quite small and met the requirements of the convergent precision. Several temperature measurement lines were implanted inside the oil conservator. Taking the derivatives of the temperature respect to the position for each temperature measurement line,the point with the maximum temperature changing ratio was regarded as the boundary point of the oil surface. Based on the obtained boundary points and by a least squares approach,the oil surface could be determined. Finally,the oil level was obtained by normalizing the area of the oil part and the oil-free part within the ellipse into the circle shape.The experimental results show that the proposed method is feasible and it can detect the oil level of transformer automatically. Therefore,it can be used in the infrared monitoring and warning system in unmanned substations.