Research On Automatic Diagnosis Algorithm For Thermal Faults Of Power Equipment Based On Infrared Images

As a significant node in the power transmission and distribution network,the malfunction of power equipment will directly cause the catastrophic destruction of the power system,and even cause potential life threat.Therefore,for the purpose of guaranteeing the security and stability of the running state of power equipment,it is very critical to carry out regular testing and maintenance.With the advantages of non-invasive,safe and relatively low cost,infrared thermal imaging detection technology has progressively become an effective method to check the operation status of power equipment.At present,the work of using infrared thermal image to evaluate the running state of power equipment can only be done by experienced technicians,resulting in the whole analysis process of infrared thermal image timeconsuming,laborious and high cost.In order to break these limitations,an unsupervised thermal malfunction diagnosis measure for power equipment is proposed in this study,which regards infrared thermal images of power equipment as the research subject.As the most important part of thermal fault diagnosis,various algorithms in thermal image processing of power equipment are studied in this paper.Since the noise generated during collection and formation will greatly reduce the readability of the digital image,this paper firstly improves the traditional wavelet threshold function and combines the bilateral filter to eliminate the mixed noise in the infrared image of power equipment.The results indicate that the presented algorithm can achieve a higher level of denoising.Secondly,due to the low efficiency and lack of spatial structure information of traditional FCM algorithm,an FCM algorithm based on cluster center initialization and weighted membership function is set up.The results indicate that the ameliorated algorithm performs better in segmentation tasks.Thirdly,in consideration of the issue that the effectiveness of infrared thermal images of power equipment is easily affected by factors such as shooting angle and distance,an algorithm is presented to unify Harris corner features and Hu invariant moments,and convert corner features into feature vectors.The results indicate that feature extraction using Harris moment invariants has the advantages of short time and high stability.Then the model based on support vector machine is set up to classify and analyze the infrared thermal image of power equipment.The results indicate that the identification precision of this method is high enough to supply the practical demand to a certain extent.Finally,according to the infrared temperature measurement technology and the causes and types of malfunctions of power equipment,a malfunction diagnosis simulation system based on the relative temperature difference judgment method is constructed,which realizes the identification of infrared thermal images of power equipment and the unsupervised diagnosis of thermal faults.