Order Publications

Reference: ISH2017_319

ISH Collection

Research on transformer fault early warning based on infrared image recognition



In most cases, the temperature of the power transformer will change when it is in fault, so the infrared thermal imaging technology is more and more widely used in fault diagnosis of power equipment. Nowadays, the infrared thermography is existed with artificial intervention treatment to obtain the temperature information of key components such as outlet casing, and the evaluation standards are relatively fixed and not up to the fault warning effect. To improve the current method, this paper employs infrared images recognition to reduce artificial intervention, the infrared images are recognized automatically. Characteristic data of infrared images using Support Vector Machine is used to achieve the purpose of early warning of transformer fault. At the beginning of the paper, Infrared image recognition using normalized cross correlation is employed to extract the region of interest in the image without artificial intervention, then searching the highest temperature and the lowest temperature points of the components, calculating the average, variance and other characteristics of the infrared images. The condition of the components of the transformer can be divided into normal, abnormal, poor and fault. Then, using the characteristics data of different condition of infrared images, the support vector machine is used to obtain the parameters of the model. Once the parameters of the model are obtained, test data can be used to identify the model and the accuracy of detection algorithm. Finally, the images are handled using the above model. Thus, the condition of the components of transformer can be assessed and the maintenance strategy can be put out. The proposed method has been applied successfully. The results show that the method not only greatly reduces the workload of the operator, but also has the effect of transformer failure warning, and has a good engineering application value.

File Size: 945,3 KB

Pages NB: 5

Year: 2017

FREE free
Non member
FREE free

Your AccountYour Account

Password forgotten


may download free of charge all publications including most recent ones

Non Members

may download free of charge publications over three years old and purchase all publications

For any questions in connection with the on-line library, contact:


Publications and Edition Manager