A weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment

Abstract

The present invention discloses a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, which belongs to the field of insulation status diagnosis of power equipment. The method fuses the information of ultrasonic images and ultraviolet images through adaptive weights to obtain a local discharge position information map, and collaboratively diagnoses the external insulation status of power equipment. The main steps of the method are to obtain local discharge fusion information through BP neural network model training, and then calculate the weight of collaborative diagnosis in combination with the Fisher discriminant criterion, and finally calculate and complete the fusion to obtain the local discharge information result. Compared with the existing technology, the collaborative diagnosis method of the present invention utilizes the technical advantages of ultrasonic detection and ultraviolet detection, makes up for the shortcomings of a single method, further improves the detection accuracy, and has high detection reliability.

Description

Technical Field

[0001] The present invention belongs to the field of insulation status diagnosis of power equipment, and in particular, relates to a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment. Background Art

[0002] With the increase in the voltage level of the power system and the increase in the number of connected devices, the operational reliability of equipment and systems is becoming increasingly important. How to conduct online monitoring or status assessment more accurately, efficiently and quickly is the core goal of equipment operation and maintenance work, and is also the key to ensuring power supply reliability.

[0003] Abnormal corona discharge of equipment or line external insulation is usually caused by potential insulation defects such as dirty insulator surface, insulation cracking, poor electrical contact, and abnormal local electric field. It is not only an important cause of accelerated insulation aging, but also an early sign of equipment external insulation failure under overvoltage. Therefore, effective detection of abnormal corona discharge is an important part of equipment and line fault prevention. Corona discharge is often accompanied by a series of sound, light and other phenomena, so ultrasonic detection and ultraviolet detection methods are produced accordingly. In order to improve the efficiency and intuitiveness of inspections, solar-blind ultraviolet imaging technology and acoustic imaging technology are also used in the visual detection of equipment and line external insulation. Ultraviolet imaging technology collects ultraviolet photons discharged in the solar-blind band through ultraviolet optical lenses, and finally uses photoelectric arrays for grayscale imaging, which is superimposed with the visible light field of view to obtain the location of the discharge source. Although this method has a solid theoretical basis, and studies have quantified the relationship between the discharge amount and the number of ultraviolet photons, it is limited by the principle of optical imaging and has certain limitations in practical applications. For example, when the discharge source is blocked or exists outside the visual bright field, ultraviolet photons are difficult to be effectively captured, resulting in a detection blind spot or positioning failure. In recent years, acoustic imaging technology has been introduced into the field of ultrasonic positioning due to its strong focusing ability, high resolution and wide-band sensitivity. Acoustic signals have strong penetration ability in solid media and are less affected by field of view obstruction, which effectively makes up for the shortcomings of ultraviolet imaging detection. However, ultrasonic imaging is easily affected by environmental noise, and the positioning accuracy will be significantly reduced in complex environments.

[0004] Research shows that the joint detection system can use the advantages of multiple monitoring methods to achieve complementarity and overcome the shortcomings of a single detection method. At the same time, with the application of image information fusion technology in medical detection, military satellites and infrared remote sensing, the development of multi-image joint detection technology has become more mature. Summary of the invention

[0005] The purpose of the present invention is to overcome the above-mentioned shortcomings, coordinate the technical advantages of ultrasonic and ultraviolet detection methods, and provide an acoustic-optical coordinated diagnostic algorithm to locate external insulation faults of power equipment more accurately and with stronger anti-interference ability.

[0006] The present invention is achieved through the following technical scheme: a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, which integrates the technical advantages of ultrasonic detection and ultraviolet detection, adjusts the weight by the environmental signal-to-noise ratio, and realizes acoustic-optical collaborative detection.

[0007] In the weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, information on reaction intensity in ultrasonic images and ultraviolet images is extracted and fused as a foreground image; the background image is a visible light image taken by a visible light camera; the fusion weight of the ultrasonic image and the ultraviolet image changes with the environmental signal-to-noise ratio, and the position information Y of the local discharge in the field of view is obtained. fu , which is convenient for visually locating the location of insulation degradation. What changes with the signal-to-noise ratio is the proportion of the foreground image synthesized by the feature quantities of the ultrasonic image and the ultraviolet image.

[0008] In the weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, ultrasonic images, ultraviolet images, and visible light images taken by a visible light camera are processed and feature extracted to obtain ultrasonic sound intensity U S , ultraviolet photon number vn image features and visible light image, the extracted ultrasonic sound intensity U S 、UV photon number v n They are used to obtain the local discharge fusion information Y of the output ultrasound and ultraviolet respectively. ul , Y uv After the adaptive weight adjustment determined by the environmental noise, the local discharge location information Y is obtained. fu .

[0009] It should be noted that the feature extraction of ultrasonic images, ultraviolet images and visible light images is to obtain the image feature ultrasonic sound intensity U S , image features, ultraviolet photon counts and visible light images; further explanation, after the ultrasonic imaging camera is positioned, different color areas will be displayed on its screen according to the strength of the discharge signal. Its essence is actually a pseudo-color image converted from the sound intensity matrix, and the extracted target is the sound intensity distribution matrix; the ultraviolet image reflects the number of ultraviolet photons sensed by the ultraviolet lens, and the ultraviolet camera will record the number of photons sensed at different positions in the space, which is the number of ultraviolet photons; the visible light image is the photo taken by the visible light camera and is used as the background image.

[0010] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment uses BP neural network to train the ultrasonic image and ultraviolet image obtained by detection in advance, and the ultrasonic and ultraviolet partial discharge fusion information Y output by the BP neural network is obtained. ul , Y uv Conduct external insulation status assessment of power equipment.

[0011] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment constructs two BP neural networks, and inputs samples of ultraviolet detection and ultrasonic detection respectively to evaluate the state of external insulation of power equipment, and obtains ultrasonic and ultraviolet partial discharge fusion information Y ul , Y uv , in preparation for subsequent testing.

[0012] In the weight-adaptive acoustic-optic collaborative diagnosis method for external insulation of power equipment, when constructing a BP neural network, the specific steps of BP neural network training are as follows: first, the hyperparameters of the BP neural network are set, and the number of neurons in the input layer, hidden layer and output layer of the BP neural network is determined according to the object to be detected, and its training accuracy is determined; secondly, the training samples are input into the BP neural network, feedforward propagation is performed, and the training error is calculated; then, the adamax optimization algorithm is used to perform a feedback process to reduce the training error and complete a single BP neural network training; finally, the above operations are repeated until the set number of training times is reached or the training error meets the training accuracy, and the training is terminated.

[0013] The input parameters in the present invention are the number of ultraviolet photons vn, the ultrasonic sound intensity U S , environmental signal-to-noise ratio SNR; the number of hidden layers in the hyperparameters is 1, each layer has 10 neurons, and the training accuracy is 1×10 -5 .

[0014] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, the input training samples are obtained by collecting needle-plate discharge model. The specific sample acquisition simulation conditions are set according to the situation of the object to be detected, and the basic rules are: the generation of partial discharge signal is simulated by the partial discharge model, and the acquisition settings of ultraviolet image are limited during the acquisition process, such as acquisition frame rate and recording time, and the shooting distance range is limited; the acquisition of ultrasonic image also limits the acquisition distance range, and the position of the object is taken as the normal when shooting, and the angle formed by the line connecting the camera and the object and the normal is selected for shooting, and a group of shots are taken at fixed intervals. The receiving frequency of the ultrasonic sensor for collecting partial discharge signals is set to 20-60kHz, and different signal-to-noise ratios are simulated by adding Gaussian white noise.

[0015] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, after inputting training samples to complete the training of the BP neural network, when applying the method, takes visible light images, ultrasonic images and ultraviolet videos of the object to be detected; and at the same time detects the signal-to-noise ratio (SNR) of the environment;

[0016] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment processes ultrasonic and ultraviolet detection to obtain an image of the object to be detected, and extracts the ultrasonic sound intensity U S 、UV photon number v n etc. image features;

[0017] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment uses the extracted image feature ultrasonic sound intensity U S 、UV photon number v n The environmental signal-to-noise ratio (SNR) is input into the trained BP neural network to obtain the output partial discharge fusion information Y ul , Y uv ;

[0018] The weight of the adaptive weighted acoustic-optical coordinated diagnosis method for external insulation of power equipment changes with the environmental signal-to-noise ratio, further affecting the proportion of the two monitoring methods in the final image fusion. The weight calculation is completed through the Fisher criterion, and the calculation process is:

[0019] ①Suppose there are n samples in the data set belonging to p categories, each category contains n i samples, and the between-class variance is The intra-class variance is Then the calculation expression of feature k in the data set is:

[0020]

[0021]

[0022] where x (k) is a sample with feature k; is the average value of class i; m (k) is the average value of all samples of feature k, x represents the sample, there are n samples in total, n i Indicates the number of samples in each category;

[0023] ② J of characteristic k F The values ​​are: J F The value represents the separability of the feature, J F The larger the value, the more separable the data is;

[0024] ③ Calculate J under different signal-to-noise ratios F (k) value, the weights under different signal-to-noise ratios are:

[0025]

[0026] Where i = 1, 2 represents ultrasound image and UV image respectively; Refers to the weight of image i under the signal-to-noise ratio SNR; J F (i, SNR) refers to the J of image i under the signal-to-noise ratio SNR F value.

[0027] The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment combines the output ultrasonic and ultraviolet partial discharge fusion information Y ul , Y uv Respectively with weight Multiply them to get the partial discharge collaborative diagnosis location information Y fu , Then the partial discharge collaborative diagnosis position information Y fu Normalization is performed, and then pseudo-colorization is performed to convert different colors corresponding to the numerical intervals into a color map, thereby completing pseudo-colorization; the pseudo-colorized image is used as the foreground image and fused with the visible light image through a fusion method based on the α component to obtain the final partial discharge collaborative diagnosis position information map.

[0028] Through the above-mentioned design scheme, the present invention can bring the following beneficial effects: the present invention proposes a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, constructs a joint diagnosis system, and provides a more accurate means for evaluating the status of external insulation of power equipment; collaborative diagnosis of ultrasonic and ultraviolet detection methods utilizes the advantages of the two methods to achieve complementarity and make up for the shortcomings of a single detection method; adaptively adjusts the weight of information fusion as the environmental signal-to-noise ratio changes, improves the applicability and accuracy of the collaborative diagnosis method, gives full play to the characteristics of each algorithm, and has higher detection capabilities and detection accuracy. In summary, it is possible to achieve a visual evaluation of the external insulation status of power equipment and accurately locate the fault location, facilitate intelligent operation and maintenance, and improve the operation and maintenance level of power equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present application. The exemplary embodiments of the present invention and their descriptions are used to understand the present invention and do not constitute improper limitations of the present invention. In the drawings:

[0030] Figure 1 It is a schematic diagram of BP neural network model training of the present invention.

[0031] Figure 2 It is a schematic diagram of the weight-adaptive external insulation collaborative diagnosis algorithm of the present invention.

[0032] Figure 3 This is a diagram of the imaging result of the acoustic-optical collaborative detection of the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical scheme in the present invention is clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the present invention is not limited to the following embodiments, and the specific implementation method can be determined according to the technical scheme of the present invention and the actual situation. In order to avoid confusing the essence of the present invention, the known methods, processes and procedures are not described in detail.

[0034] like Figure 1 , Figure 2 and Figure 3 As shown, a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment includes BP neural network training, detection of objects to be detected, image processing and feature extraction, weight calculation, and acquisition of partial discharge collaborative diagnosis position information map.

[0035] When this embodiment is applied:

[0036] First, the training of BP neural network for ultrasound and ultraviolet detection needs to be completed respectively. During the training, the number of ultraviolet photons vn of the ultraviolet detection sample and the ultrasonic sound intensity U of the ultrasonic detection sample need to be extracted. S , and record the corresponding environmental signal-to-noise ratio at the same time, and input the BP neural network with determined parameters to obtain the partial discharge fusion information;

[0037] Secondly, the input training samples are obtained through the needle-plate discharge model; the specific sample acquisition simulation conditions are set according to the situation of the object to be detected, and the basic rules are: the generation of local discharge signals is simulated through the local discharge model, and the acquisition frame rate and recording time of the ultraviolet image are limited during the acquisition process, and the shooting distance range is also limited; the acquisition of ultrasonic images is also limited to the acquisition distance range. When shooting, the position of the object is taken as the normal, and the angle formed by the line connecting the camera and the object and the normal is selected for shooting. A group of shots are taken at fixed intervals. The receiving frequency of the ultrasonic sensor for collecting local discharge signals is set to 20-60kHz, and different signal-to-noise ratios are simulated by adding Gaussian white noise;

[0038] Secondly, after completing the sample training, shoot the visible light background, ultraviolet image video, and ultrasonic detection image of the object to be detected, extract the corresponding ultraviolet photon number vn, and ultrasonic detection sound intensity U S The characteristic information, together with the environmental signal-to-noise ratio, is input into the trained BP neural network to obtain the partial discharge fusion information, Y uv and Y ul ;

[0039] Then, the weight values ​​of the two detection methods under different signal-to-noise ratios are calculated by the Fisher criterion, and the obtained The calculation process is:

[0040] Suppose there are n samples in the data set belonging to p categories, and each category contains ni samples, and the between-class variance is The intra-class variance is Then the calculation expression of feature k in the data set is:

[0041]

[0042]

[0043] where x (k) is the feature k of each sample; is the average value of class i; m (k) is the average value of all samples of feature k, x represents the sample, there are n samples in total, n i Indicates the number of samples in each category;

[0044] J of characteristic k F The values ​​are:

[0045] Calculate J for different signal-to-noise ratios F (k) value, the weights under different signal-to-noise ratios are:

[0046]

[0047] Where i = 1, 2 represents ultrasound and UV images, respectively; Refers to the weight of image i under the signal-to-noise ratio SNR; J F (i, SNR) refers to the J of image i under the signal-to-noise ratio SNR F value;

[0048] Finally, the local discharge position information of the acoustic-optical collaborative detection is obtained by multiplying the local discharge fusion information with the weight to obtain the local discharge collaborative diagnosis position information Y fu , Then the partial discharge collaborative diagnosis position information Y fu After normalization, it is further pseudo-colorized. The intensity of the local discharge fusion information is divided into intervals and matched with different colors. The local discharge fusion information intensity matrix is ​​converted into a color intensity matrix, thereby completing the pseudo-colorization. The pseudo-colorized image is used as the foreground image and fused with the visible light image through a fusion method based on the α component to obtain the final local discharge collaborative diagnosis location information map.

[0049] In summary, the present invention relates to a weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment. The method fuses the information of ultrasonic images and ultraviolet images through adaptive weights to obtain a local discharge position information map, and collaboratively diagnoses the external insulation status of power equipment. The main steps of the method are to obtain local discharge fusion information through BP neural network model training, and then calculate the weight of collaborative diagnosis in combination with the Fisher discriminant criterion, and finally calculate and complete the fusion to obtain the local discharge information result. Compared with the existing technology, the collaborative diagnosis method of the present invention utilizes the technical advantages of ultrasonic detection and ultraviolet detection, makes up for the shortcomings of a single method, further improves the detection accuracy, and has high detection reliability.

Claims

1. A weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of power equipment, characterized in that: The method comprises the following steps: S1. Evaluate the external insulation status of power equipment through ultrasonic testing and ultraviolet testing, obtain ultrasonic images and ultraviolet images; and simultaneously detect the signal-to-noise ratio (SNR) of the environment; S2, performing weighted fusion on the ultrasonic image and the ultraviolet image obtained in step S1 to obtain partial discharge fusion information, which is used as a foreground image after fusion, and the visible light image taken by the visible light camera is used as a background image, and the foreground image and the background image are subjected to partial discharge detection image fusion to obtain a partial discharge position information map, thereby evaluating the external insulation state of the power equipment; Among them, the fusion weight of the ultrasound image and the ultraviolet image in step S2 changes with the different environmental signal-to-noise ratios; the calculation of the fusion weight of the ultrasound image and the ultraviolet image is achieved through the Fisher criterion, and the specific process is: ①Suppose there are n samples in the data set belonging to p categories, each category contains n i samples, and the between-class variance is The intra-class variance is Then the calculation expression of feature k in the data set is: where x (k) is a sample with feature k; is the average value of class i; m (k) is the average value of all samples of feature k, x represents the sample, there are n samples in total, n i Indicates the number of samples in each category; ② J of characteristic k F The values ​​are: ③ Calculate J under different signal-to-noise ratios F (k) value, the weights under different signal-to-noise ratios are: Where i = 1, 2 represents ultrasound image and UV image respectively; Refers to the weight of image i under the signal-to-noise ratio SNR; J F (i, SNR) refers to the J of image i under the signal-to-noise ratio SNR F value.

2. The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of electric power equipment according to claim 1 is characterized in that: In step S2, the ultrasonic image and the ultraviolet image are weightedly fused to obtain partial discharge fusion information, which is used as the foreground image after fusion, and the visible light image taken by the visible light camera is used as the background image. The foreground image and the background image are subjected to partial discharge detection image fusion to obtain the partial discharge position information map. The specific process includes: ① Construct two BP neural networks, which are used to input ultrasonic detection and ultraviolet detection samples respectively, and output the partial discharge fusion information Y of ultrasonic and ultraviolet. ul , Y uv ; ②After image processing and feature extraction, the ultrasonic image, ultraviolet image and visible light image taken by the visible light camera are used to obtain the ultrasonic sound intensity U S 、UV photon number v n and visible light image, the extracted ultrasonic sound intensity U S and the number of ultraviolet photons v n The environmental signal-to-noise ratio is input into the corresponding trained BP neural network respectively, and the ultrasonic and ultraviolet partial discharge fusion information Y is output respectively. ul , Y uv ; ③ The output ultrasound and ultraviolet partial discharge fusion information Y ul , Y uv Respectively with weight Multiply them to get the partial discharge collaborative diagnosis location information Y fu , Then the partial discharge collaborative diagnosis position information Y fu Normalization processing is further performed after normalization processing. The intervals are divided according to the intensity of the local discharge fusion information and different colors are matched. The local discharge fusion information intensity matrix is ​​converted into a color intensity matrix, thereby completing the pseudo-colorization; the pseudo-colorized image is used as the foreground image and fused with the visible light image through a fusion method based on the α component to obtain the final local discharge collaborative diagnosis location information map.

3. The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of electric power equipment according to claim 2 is characterized in that: When constructing a BP neural network, the specific steps of BP neural network training are: ① Set the hyperparameters of the BP neural network, determine the number of neurons in the input layer, hidden layer and output layer of the neural network according to the object to be detected, and determine its training accuracy; ② Input the training samples into the BP neural network, perform feedforward propagation, and calculate the training error; ③Use the adamax optimization algorithm to conduct feedback process, reduce training error, and complete single BP neural network training; ④ Repeat steps ② and ③ until the set number of training times is reached or the training error meets the training accuracy, and then the training ends.

4. The weight-adaptive acoustic-optical collaborative diagnosis method for external insulation of electric power equipment according to claim 3 is characterized in that: The input training samples are obtained through the needle-plate discharge model; the specific sample acquisition simulation conditions are set according to the situation of the object to be detected, and the basic rules are: the generation of local discharge signals is simulated through the local discharge model, and the acquisition frame rate and recording time of the ultraviolet image are limited during the acquisition process, and the shooting distance range is also limited; the acquisition of ultrasonic images also limits the acquisition distance range. When shooting, the position of the object is taken as the normal, and the angle formed by the line connecting the camera and the object and the normal is selected for shooting. A group of shots are taken at fixed angles. The receiving frequency of the ultrasonic sensor for collecting local discharge signals is set to 20-60kHz, and different signal-to-noise ratios are simulated by adding Gaussian white noise.

Images