Research On Temperature Monitoring Method Of Power Electronic Circuit Based On Infrared Thermal Image And Object Detection

With the widespread application of power electronic converters in all walks of life,users have also put forward higher requirements for their reliability.In order to improve the reliability of the thermal design of the converter,it is necessary to perform temperature detection on the converter during the design verification stage.The temperature measurement method based on infrared thermal imaging has the advantages of rich temperature information,wide coverage,safety and efficiency,etc.,and has been widely used in circuit temperature monitoring.However,due to the mechanism of infrared thermal imaging,the circuit component cannot be automatically positioned in the thermal image.This paper combines infrared thermal imaging and image recognition technology,and proposes a method to solve the location of circuit components in infrared thermal images,and then realizes the automatic detection of circuit faults based on temperature information.Visible light images have high resolution and clear images.Combining visible light images and infrared thermal images,using target detection algorithms in the field of deep learning can accurately determine the location and size of components and obtain comprehensive temperature information of components.This paper studies how to integrate the target detection algorithm with the recognition function of power electronic components,produced the industry’s first data set for power electronic component recognition,trained the model of the circuit component recognition algorithm,and introduced the average precision average and Io U.The training results are quantified and evaluated by other indicators,and the function of automatic recognition and positioning of circuit components in the visible light image is realized.Using the component recognition algorithm to output the component position information in the visible light picture,this paper uses the rigid body transformation to register the position frame to the infrared thermal image,completing the preliminary positioning of the component in the thermal image.Aiming at the problem that the rectangular position frame of the infrared thermal image contains the component part and the non-component part image,this paper further thresholds the partial image of the position frame to extract the area that only contains the component,so as to realize the more accurate extraction of the component temperature information.On this basis,this paper adopts the temperature sequence detection method to diagnose the faults of the components.The maximum temperature,the minimum temperature,the average temperature and the median temperature of the components are used as the temperature characteristic values,and the temperature characteristics of the circuit component to be tested and the reference temperature of the standard circuit are used.Comparing the calculations,the mismatch rate of the temperature sequence is obtained,and it is judged whether the component fails according to the mismatch rate of multiple temperature characteristic values.Then a method of feeding back temperature information to the circuit controller for temperature closed-loop control is proposed.In the experimental part,this article uses a three-phase bridge inverter circuit as an example to verify the above design.The experimental results prove the feasibility and accuracy of the circuit temperature monitoring system design in this article.This paper is an innovative interdisciplinary attempt between power electronics and image recognition.The innovations are:(1)the first data set for power electronic component recognition has been established,which is convenient for researchers to carry out subsequent algorithmic research.(2)use visual algorithms to automatically identify component types and obtain component temperature information(3)design a circuit fault diagnosis algorithm based on circuit operation principles,and use component temperature as feedback information for circuit control.