Application—Infrared Thermal Imaging and Intelligent Algorithm for Fault Detection of Power Equipment

Infrared Thermal Imaging and Intelligent Algorithm for Fault Detection of Power Equipment

   The substations and transmission lines in the power system are important hubs connecting power plants and users, and their safe and stable operation is essential. Once the power equipment fails, the safety of the power system and the stability of power supply will be greatly affected. Power equipment is prone to failure due to external environmental influences such as climate factors for a long time. Therefore, it is necessary to conduct regular inspection and maintenance of power equipment to ensure the normal operation of the power supply system. According to relevant statistics, up to 90% of power system accidents are caused by power equipment failure, and more than 50% of the faulty equipment will have abnormal fever symptoms in the early stage. The principle of infrared temperature measurement is that the detector detects and receives the infrared radiation energy emitted by the measured target, converts the received infrared radiation energy into the corresponding electrical signal, and then obtains the surface of the object through a special electrical signal processing system. Temperature distribution status. The thermal failure of power equipment is determined by factors such as the type of power equipment, the heating position and the degree of heat generation, and its temperature distribution is also different. Therefore, infrared thermal imaging technology is very suitable for thermal fault detection of power equipment. Analyzing the temperature distribution information on the surface of power equipment can detect potential hazards and faults in power equipment, and make a quantitative judgment on the severity of the fault.

   At present, the most important form of inspection in the power system is manual inspection, manual on-site diagnosis or collecting information for subsequent analysis. Manual inspection has a large workload and high management costs. Skill training is required for technical personnel, and information collection and fault analysis need to be completed manually. However, the domestic power system is widely distributed and the environment in some areas is harsh, which increases the cost and difficulty of inspections, and manual inspections become extremely complicated. If the inspection is not timely, once the power equipment fails, it will cause serious accidents. Nowadays, the fusion of convolutional neural networks, sensors and information technology has been extensively studied and applied to power system inspections, which reduces the cost and difficulty of inspections to a certain extent, in order to find safety hazards in time, eliminate faults and deal with emergencies The situation laid the foundation.

   Generally, the intensity of infrared radiation emitted from the surface of an object will attenuate in the air. Therefore, the infrared temperature measurement result of the measured target is often lower than the actual temperature of the measured target. The farther the distance, the greater the actual difference. Therefore, it is necessary to correct the infrared temperature measurement result of the measured target. The BP neural network algorithm is widely used in various scenarios, and when exploring the influencing parameters of infrared temperature measurement, the BP neural network algorithm has strong adaptability and high accuracy compared with linear interpolation, multiple linear regression, and nonlinear mapping. The temperature correction module will use BP neural network to perform temperature correction on the infrared temperature measurement results.

Figure 1. Infrared thermal imaging diagram of power equipment.

   The research designed a thermal fault detection method for power equipment for Shandong Electric Power, combining infrared thermal imaging technology with deep learning, reading video streams, deep learning power equipment detection, BP neural network temperature correction, and data visualization Fusion. The power equipment thermal fault detection architecture based on embedded deep learning has three main levels.

   The bottom layer is data reading. The infrared camera outputs a video stream in MPEG-4 format via Ethernet, and decodes the infrared camera video stream into frames and transmits it to the next layer.

   The middle layer is the data processing layer, which mainly performs thermal fault diagnosis from the infrared camera video stream obtained from the upper layer. A corresponding deep learning framework is built for the identification of power equipment detection algorithms in the data processing layer, and the relevant network model deployed is a model that has been trained. This layer only performs detection tasks and does not perform network model training tasks. This layer is mainly responsible for real-time power equipment detection and equipment positioning; BP neural network is deployed to perform temperature correction on infrared temperature measurement results. The corrected temperature will determine whether the device has abnormal fever symptoms through a priori knowledge base.

   The top layer is the data service layer, which visually displays the data processing results of the middle layer, and the final detection results can be presented in a more intuitive way.

Figure 2. Infrared thermal imaging diagram of power equipment.

   The thermal fault diagnosis method of power equipment based on embedded deep learning is mainly divided into three tasks: detection and location of power equipment, temperature extraction of target equipment, and thermal fault diagnosis of target equipment. The detection process of power equipment thermal fault detection is:

(1) Read the infrared thermal imaging video stream from the infrared thermal imaging camera and decode it into frames.

(2) The power equipment detection algorithm detects whether there are power equipment in each frame of image and locates it.

(3) According to the positioning information obtained in the previous step, obtain infrared temperature measurement and laser ranging data from the infrared thermal imager.

(4) According to the infrared temperature measurement and laser ranging data, the corrected temperature is obtained through the temperature correction module.

(5) Finally, use a priori knowledge base to perform thermal fault diagnosis on the corrected temperature, and obtain thermal fault detection results.