Research On Transformer Oil Level Detection Method Based On Video Image

The height of the transformer oil level is an important indicator in the production process of the power plant.If the indicator is too high or too low,it is not conducive to the safe and stable operation of the transformer.It will not only affect the service life of the transformer itself,but also may damage other power equipment and even threaten the production safety of the entire power plant.Therefore,it is very important to detect the height of the transformer oil level.This paper studies how to quickly and accurately detect the height of the oil level of the transformer.The inspection robot is used to capture the video image data of the oil level of the transformer bushing,and the image is reconstructed and processed to improve the clarity of the image.To solve the problem of line occlusion,the rain removal algorithm is used to remove the influence of rain lines on the image.And the maximum inter-class variance method is used to segment the oil level area,and the area projection theorem is used to calculate the oil level height,so as to realize the detection of the transformer oil level height.The main research content of this paper is divided into the following three parts:（1）Aiming at the problems of low resolution and blur of transformer oil level images,an image super-resolution reconstruction algorithm based on residual sub-pixel convolutional neural network is designed.First,select the residual network with the batch normalization layer removed to improve the convolutional neural network super-resolution algorithm;then re-extract and filter the multi-layer feature maps to form a global feature multiplexing module to enhance the detailed information of the oil level image;the obtained feature image is input into the sub-pixel convolutional network to obtain a high-definition transformer oil level image.（2）Aiming at the situation that the image collected in rainy days has rain lines blocking the picture,a single image rain removal algorithm based on HSI space transformation and improved dense connection network is designed.Firstly,it is clarified that the rain lines in the video image mainly presents the brightness information through analysis;then the collected image is transformed from the RGB space to the HSI space,and the brightness information is selected from it;finally,the improved dense connection network is used to learn the characteristics of brightness information,so as to obtain the transformer oil level image without the influence of rain line.（3）For the identification of transformer oil level and liquid level,a method for identifying oil level and liquid level based on the maximum inter-class variance method an determine d the area projection theorem is designed.First,determine the region of interest from the collected image;then use the maximum inter-class variance method with adaptive threshold under different lighting conditions to extract the oil level area in the image;finally,use the area projection theorem to convert the captured transformer bushing glass elliptical window reduction into a circle,so as to calculate the specific height of the transformer oil level.The transformer oil level detection method based on video images proposed in this paper can quickly and accurately identify the height of the oil level of the transformer bushing.By detecting the oil level of the transformer,it can be determined whether the oil level is within the specified qualified range,so as to ensure the safe and stable operation of the transformer.At the same time,this method makes the process of transformer oil level detection more concise and intelligent.