Method for comprehensively searching hidden defects of operating equipment of transformer substation

Abstract

The invention discloses a comprehensive searching method for hidden defects of operation equipment of a transformer substation, which comprises the following steps of: firstly, establishing a primary equipment hidden defect library, taking all defective parts of primary equipment as observation objects, and carrying out machine vision and acoustic vibration acquisition; the method comprises the following steps: determining whether primary equipment is deviated or loosened, judging a part where a primary fault occurs according to the sound vibration signal correlation degree, judging whether a secondary signal is abnormal or not through secondary circuit auxiliary contact signal detection, and comprehensively judging an image pairing result, the sound vibration correlation degree and whether the secondary signal is abnormal or not. And determining hidden dangers and defect types of the equipment. The method provided by the invention can be used for distinguishing different hidden dangers and faults of the equipment and improving the detection accuracy.

Description

Technical Field

[0001] The present invention relates to the field of transformer substation defect detection, and in particular to a comprehensive method for detecting invisible defects of transformer substation operating equipment. Background Art

[0002] In the actual operation of substation equipment, a large number of auxiliary contacts of circuit breakers or disconnectors are involved in the secondary circuit. Abnormal operation of primary equipment (such as connecting rod not in place, loose screws causing slight displacement of travel switch, etc.) often leads to abnormal secondary circuit or triggers signal alarm. Primary equipment abnormalities are very hidden, and the installation locations of some auxiliary contacts are not easy to observe. In actual work, personnel are required to keep a safe distance from the operating equipment, which makes it more difficult to find defects. For this reason, we consider using machine vision methods to perform feature detection on primary equipment, combining acoustic and vibration feature comparison, and establishing a connection between the secondary circuit and the primary equipment, and studying a comprehensive method for finding invisible defects in substation operating equipment. Summary of the invention

[0003] The purpose of the present invention is to provide a comprehensive method for finding invisible defects of substation operating equipment, which can combine machine vision methods and acoustic vibration characteristics to promptly discover invisible defects of equipment that are not easy to find.

[0004] In order to achieve the above object, the present invention is implemented through the following technical solutions: A comprehensive method for finding invisible defects of substation operating equipment, comprising the steps of: Read the data in the historical defect database, determine and establish the image acquisition points and acoustic vibration acquisition points of the equipment of interest, and form a one-to-one or many-to-one mapping relationship of the primary equipment invisible defect library; Take all the parts of the primary equipment where defects have occurred as observation objects, collect images at the image collection points of the equipment of interest, compare and match the images of the primary equipment in normal operation, the images of the primary equipment in action or other starting conditions, and determine whether there is image deviation of the primary equipment based on the similarity; Acoustic vibration collection is performed at the acoustic vibration collection points of interest, and the collected sound and vibration characteristic waveforms are used as the parent sequence, and the sound and vibration characteristic waveforms of all primary equipment failures are used as subsequences. According to the correlation between the parent sequence and the subsequence, it is determined whether there are abnormal acoustic and vibration signals in the primary equipment; Number the secondary circuits related to the primary equipment, associate the secondary circuit auxiliary contacts with related signals to form an associated signal library, take the signal before the primary equipment is actuated as the basic set, and the signal after the primary equipment is actuated as the comparison set, compare the data of the basic set and the comparison set to determine whether there is any secondary signal abnormality; Make a comprehensive judgment on the image pairing results, sound-vibration correlation and whether the secondary signal is abnormal to determine the hidden dangers and defect types of the equipment.

[0005] Furthermore, after determining the position and its neighborhood where the image offset exists in a device, The acoustic vibration correlation degree between the parent sequence and the child sequence is calculated, and the acoustic vibration correlation degree calculation is not performed for other positions.

[0006] Furthermore, the step of comparing and pairing the captured images of a device includes: obtaining image feature points, describing the feature points, obtaining feature descriptors of all feature points of the image when the image capture point of the same device is in normal operation and action or other starting conditions, and calculating the Hamming distance between the corresponding feature descriptors of the two images. The smaller the distance, the more similar the feature descriptors.

[0007] Furthermore, the method for obtaining the feature point is as follows: taking the corner point as the feature point, using the window sliding model, and judging whether the feature point is a corner point by the size change of the first-order grayscale derivative of the image within a certain range; Define grayscale change function ,in Move the area for the window. is the weight, Indicates the pixel of the image corresponding to the window, Represents the pixel of the corresponding image after the window moves, according to Taylor's formula ,get in , , for Partial derivative of ; find the matrix If the minimum eigenvalue is Greater than threshold , then the pixel The required feature points.

[0008] Furthermore, the method of describing the feature points is: Perform Gaussian filtering on the image, taking the feature point as the center and taking A neighborhood window is randomly selected in the window, and the size of the pixels of the two points is compared to perform binary assignment. in , It's a random point and The pixel value at ; Random selection in window Repeat the above steps for random points to form a 256-bit binary code as a feature descriptor, denoted as , Indicates feature descriptors.

[0009] Furthermore, the steps of calculating the correlation between the parent sequence and the subsequence collected by acoustic vibration include: The denoised signal collected by a certain acoustic vibration sensor is taken as the mother sequence, denoted as ,in, is the parent sequence element, Indicates The sound and vibration characteristic signals of each acoustic and vibration collection point, ; The denoised signals after each primary equipment failure in the acoustic vibration defect library are taken as subsequences, denoted as ,in is the type of primary equipment failure, ; Preprocess the elements of the parent sequence and the child sequence, find the average value of the elements of each sequence, and then divide each element in the sequence by the mean of the sequence to obtain the preprocessed parent sequence elements. and the preprocessed subsequence elements : Forming a standardized matrix ; Calculate the grey correlation coefficient between the subsequence of the kth acoustic vibration acquisition point of the tth sampling channel and the parent sequence in is the resolution coefficient, and the grey relational degree of the t-th sampling channel is ; The one with the largest grey correlation degree indicates that a fault of this type has occurred, and the location of the fault can be determined based on the position of the acoustic vibration sensor.

[0010] Furthermore, before calculating the correlation of the acoustic vibration acquisition sequence, the acquired signal is first denoised by wavelet threshold: Set the number of wavelet decomposition layers, select the wavelet basis function, and construct the wavelet threshold function; Use the selected wavelet threshold function to process the noisy signal using the Mallat algorithm to obtain the wavelet high-frequency coefficients and low-frequency coefficients corresponding to the decomposition layer number; Compare the wavelet coefficient with the wavelet threshold. If the wavelet coefficient is greater than the wavelet threshold, the corresponding data bit is a valid signal, and the wavelet coefficient is retained; otherwise, the signal is a noise signal and is discarded. The retained wavelet signal is reconstructed to obtain the denoised signal.

[0011] Furthermore, the method for determining whether the secondary signal is abnormal is as follows: For auxiliary contacts, signals, and the signal before a device action is taken as the basic set, recorded as , the signal after a device action is used as the comparison set, recorded as , perform XOR calculation on the corresponding elements of the two data sets to obtain the signal generation set, recorded as If all the elements of the signal generation set are 0, it proves that the device has taken action once. If all are 1, it means that the device has not taken action once. If some elements are 1, it means that the device has taken action once but there is a problem with the auxiliary contact where the signal with the element 1 is located.

[0012] Furthermore, the defect determination method is as follows: when a feature point descriptor of two images changes, the number of feature point descriptors that change within its neighborhood is greater than It means that the feature point has shifted or loosened, and the point is recorded as a valid feature point. Indicates the minimum number of changes in the feature point descriptor. When the number of valid feature points is greater than When the device is offset or loose, Indicates the lower limit of the valid feature point value; The grey correlation degree takes the lower limit value, and a value greater than the lower limit value indicates that effective acoustic vibration information has been collected.

[0013] Further, the comprehensive judgment method includes: Equipment displacement or looseness occurs, but no effective acoustic and vibration information is generated, and no secondary signal abnormality occurs, so it is considered that a general hidden danger has occurred on site; The equipment is offset or loose, and effective acoustic and vibration information is generated, but no secondary signal abnormality occurs, which indicates that a serious hidden danger has occurred on site; When equipment displacement or looseness occurs and effective acoustic vibration information is generated, and all related secondary abnormal signals occur, it is considered that a primary equipment defect has occurred; If there is no equipment displacement or looseness, no effective sound and vibration information occurs, but one or more secondary abnormal signals occur, it is assumed that there is no problem with the primary equipment and the secondary circuit contacts are checked.

[0014] The advantages of the present invention are: Use machine vision technology to detect the characteristics of a device and use image recognition technology to monitor the operating status of the device. This allows for a detailed inspection of the device's appearance and the detection of subtle changes that are difficult for the human eye to detect. Through the joint feature analysis of sound waves and vibration signals, the operating status of key equipment such as transformers can be identified. The sound waves and vibration signals of the equipment during operation can be obtained non-invasively, and the features can be extracted through signal processing technology for auxiliary identification of equipment status. By establishing a connection between the primary equipment and the secondary circuit, problems can be diagnosed more accurately. When an abnormality occurs in the primary equipment, the possible cause of the fault can be inferred by analyzing the signal changes of the secondary circuit connected to it. By correlating machine vision and acoustic vibration signals and adding secondary circuit judgment, hidden dangers and defects of equipment can be identified more accurately. Machine vision technology can accurately identify and locate slight changes in equipment through image recognition and analysis, while acoustic vibration feature detection can penetrate a variety of media and provide detailed information about the internal status of the equipment. The secondary circuit can reduce the amount of manual inspections and combine with the remote monitoring system for comprehensive judgment, which can distinguish different equipment hidden dangers and failures and improve the accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of a comprehensive method for finding invisible defects of substation operating equipment according to the present invention; Figure 2 This is a diagram showing the relationship between the auxiliary contacts and associated signals of the present invention. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] A comprehensive method for finding invisible defects in substation operating equipment, please refer to Figure 1 Flowchart shown.

[0018] Includes steps: S1. Organize the defect database of previous years, sort out the defects of primary equipment abnormalities, and organize the possible causes of a certain primary equipment abnormality, determine and establish the image acquisition points and acoustic vibration acquisition points of the equipment of interest, and form a primary equipment invisible defect database with a one-to-one or many-to-one mapping relationship.

[0019] S2. All parts of the equipment where defects have occurred are taken as observation objects. Each time the equipment moves, a monocular camera is used to collect images of the equipment graphics collection points of interest.

[0020] S21. Taking the corner point as a feature point, using the window sliding model, judging whether the feature point is a corner point by the size change of the first-order grayscale derivative of the image within a certain range.

[0021] Define grayscale change function ,in Move the area for the window. is the weight, Indicates the pixel of the image corresponding to the window, Represents the pixel of the corresponding image after the window moves, according to Taylor's formula ,get in , , for Partial derivative of ; find the matrix If the minimum eigenvalue is Greater than threshold , then the pixel The required feature points.

[0022] S22, perform Gaussian filtering on the image, taking the feature point as the center, and take A neighborhood window is randomly selected in the window, and the size of the pixels of the two points is compared to perform binary assignment. in , It's a random point and The pixel value at ; Random selection in window Repeat the above steps for random points to form a 256-bit binary code. The code is the description of the feature point, that is, the feature descriptor, recorded as , Indicates feature descriptors.

[0023] S23, perform steps S21-S22 on the two images to obtain all feature descriptors. Since only the part of interest is photographed, each component of the part must not be displaced, and all feature descriptors of the two photos must correspond to each other and be exactly the same. Obtain the Hamming distance of the corresponding feature descriptors of the two images. Assume that the normal operation image The feature descriptor is , when a device is activated or there are other starting conditions, the image The feature descriptor is , calculate the Hamming distance as , Represents two images The smaller the distance, the more similar the feature descriptors are.

[0024] S3. Perform acoustic and vibration collection at the acoustic and vibration collection points of interest. When some abnormal conditions occur, such as the connecting rod of a primary device not being in place, the feature descriptor pairing method can be used for judgment. When the switch operates normally or an abnormal condition occurs, it will be accompanied by abnormal sound and abnormal vibration. If the screw (the structure is symmetrical) is loose and the shooting time is at an indistinguishable moment, the two images will be mismatched. However, the loose screw will definitely be accompanied by abnormal sound and vibration. Abnormal acoustic and vibration information can be used as another condition to determine whether the primary device is abnormal. According to the defect library, sound and vibration sensors are installed at the acoustic and vibration information collection points.

[0025] The sound and vibration characteristic waveforms collected by the acoustic and vibration collection points at the image offset position and its neighborhood are taken as the parent sequence, and the sound and vibration characteristic waveforms of all primary equipment failures are taken as subsequences. The correlation between the parent sequence and the subsequence is calculated.

[0026] S31, first perform wavelet threshold denoising on the collected signal: Set the number of wavelet decomposition layers, select the wavelet basis function, and construct the wavelet threshold function; Use the selected wavelet threshold function to process the noisy signal using the Mallat algorithm to obtain the wavelet high-frequency coefficients and low-frequency coefficients corresponding to the decomposition layer number; Compare the wavelet coefficient with the wavelet threshold. If the wavelet coefficient is greater than the wavelet threshold, the corresponding data bit is a valid signal, and the wavelet coefficient is retained; otherwise, the signal is a noise signal and is discarded. The retained wavelet signal is reconstructed to obtain the denoised signal.

[0027] S32. The parent sequence and subsequence of acoustic vibration collection are calculated using grey correlation: The denoised signal collected by a certain acoustic vibration sensor is taken as the mother sequence, denoted as ,in, is the parent sequence element, Indicates The sound and vibration characteristic signals of each acoustic and vibration collection point, ; The denoised signals after each primary equipment failure in the acoustic vibration defect library are taken as subsequences, denoted as ,in is the type of primary equipment failure, ; Preprocess the elements of the parent sequence and the child sequence, find the average value of the elements of each sequence, and then divide each element in the sequence by the mean of the sequence to obtain the preprocessed parent sequence elements. and the preprocessed subsequence elements : Forming a standardized matrix ; Calculate the grey correlation coefficient between the subsequence of the kth acoustic vibration acquisition point of the tth sampling channel and the parent sequence in is the resolution coefficient, and the grey relational degree of the t-th sampling channel is ; The one with the largest grey correlation degree indicates that a fault of this type has occurred, and the location of the fault can be determined based on the position of the acoustic vibration sensor.

[0028] S4. Number the secondary circuits related to the primary equipment, associate the secondary circuit auxiliary contacts with the relevant signals, and form an associated signal library. Please refer to Figure 2 , the signal before a device is operated is taken as the basic set, and the signal after a device is operated is taken as the comparison set. The data of the basic set and the comparison set are compared to determine whether there is a secondary signal anomaly.

[0029] Assume that For auxiliary contacts, respectively signals, and the signal before a device action is taken as the basic set, recorded as , the signal after a device action is used as the comparison set, recorded as , perform XOR calculation on the corresponding elements of the two data sets to obtain the signal generation set, recorded as If all the elements of the signal generation set are 0, it proves that the device has taken action once. If all are 1, it means that the device has not taken action once. If some elements are 1, it means that the device has taken action once but there is a problem with the auxiliary contact where the signal with the element 1 is located.

[0030] S5. Comprehensive judgment. In the actual operation of the equipment, the equipment needs to be sufficiently offset or loose before the acoustic vibration signal will occur. It is necessary to comprehensively consider the image information and acoustic vibration information to determine the location of the fault or hidden danger. When the equipment is offset or loose, the image feature descriptor will change continuously in pieces; it needs to be determined that the offset or looseness will occur first to generate effective acoustic vibration information, and the gray correlation degree needs to take the lower limit value. If it is greater than the lower limit value, it means that effective acoustic vibration information has been collected.

[0031] Assume that the feature point descriptors of the two images are changes, and the number of feature point descriptors that change within its neighborhood is greater than ( Indicates the minimum number of changes in the feature point descriptor) indicates that the feature point has shifted or loosened, and the point is recorded as a valid feature point; when the number of valid feature points is greater than , it indicates that the equipment is offset or loose. When the equipment is offset or loose, but there is no valid sound and vibration information and no secondary abnormal signal, it can be considered that a general hidden danger has occurred on site; when the equipment is offset or loose, and valid sound and vibration information occurs, but no secondary abnormal signal occurs, it indicates that the hidden danger on site is very serious and needs to be verified and handled; when the equipment is offset or loose, and valid sound and vibration information occurs, and all the related secondary abnormal signals occur, it indicates that a primary equipment defect has occurred; if the feature point descriptors of the two images are all similar (the equipment is not offset or loose), there is no valid sound and vibration information, and only individual secondary abnormal signals occur, it indicates that there is no problem with the primary equipment and the secondary circuit contacts need to be checked.

[0032] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A comprehensive method for finding invisible defects of substation operating equipment, characterized in that: Includes steps: Read the data in the historical defect database, determine and establish the image acquisition points and acoustic vibration acquisition points of the equipment of interest, and form a one-to-one or many-to-one mapping relationship of the primary equipment invisible defect library; Take all the parts of the primary equipment where defects have occurred as observation objects, collect images at the image collection points of the equipment of interest, compare and match the images of the primary equipment in normal operation, the images of the primary equipment in action or other starting conditions, and determine whether there is image deviation of the primary equipment based on the similarity; Acoustic vibration collection is performed at the acoustic vibration collection points of interest, and the collected sound and vibration characteristic waveforms are used as the parent sequence, and the sound and vibration characteristic waveforms of all primary equipment failures are used as subsequences. According to the correlation between the parent sequence and the subsequence, it is determined whether there are abnormal acoustic and vibration signals in the primary equipment; Number the secondary circuits related to the primary equipment, associate the secondary circuit auxiliary contacts with related signals to form an associated signal library, take the signal before the primary equipment is actuated as the basic set, and the signal after the primary equipment is actuated as the comparison set, compare the data of the basic set and the comparison set to determine whether there is any secondary signal abnormality; Make a comprehensive judgment on the image pairing results, sound-vibration correlation and whether the secondary signal is abnormal to determine the hidden dangers and defect types of the equipment.

2. According to claim 1, the method for finding hidden defects of substation operating equipment is characterized in that: Determine the location and neighborhood of the image offset of a device The acoustic vibration correlation degree between the parent sequence and the child sequence is calculated, and the acoustic vibration correlation degree calculation is not performed for other positions.

3. The method for finding invisible defects of substation operating equipment according to claim 1 or 2, characterized in that: The steps of comparing and pairing the captured images of a device include: obtaining image feature points, describing the feature points, obtaining feature descriptors of all feature points of the image when the image capture point of the same device is in normal operation and action or other starting conditions, and calculating the Hamming distance between the corresponding feature descriptors of the two images. The smaller the distance, the more similar the feature descriptors are.

4. According to claim 3, the method for finding invisible defects of substation operating equipment is characterized in that: The method for obtaining the feature point is: taking the corner point as the feature point, using the window sliding model, and judging whether the feature point is a corner point by the size change of the first-order grayscale derivative of the image within a certain range; Define grayscale change function ,in Move the area for the window. is the weight, Indicates the pixel of the image corresponding to the window, Represents the pixel of the corresponding image after the window moves, according to Taylor's formula ,get in , , for Partial derivative of ; find the matrix If the minimum eigenvalue is Greater than threshold , then the pixel The required feature points.

5. According to claim 4, the method for finding hidden defects of substation operating equipment is characterized in that: The method of describing the feature points is: Perform Gaussian filtering on the image, taking the feature point as the center and taking A neighborhood window is randomly selected in the window, and the size of the pixels of the two points is compared to perform binary assignment. in , It's a random point and The pixel value at ; Random selection in window Repeat the above steps for random points to define a 256-bit binary code as a feature descriptor, denoted as , Indicates feature descriptors.

6. The method for finding invisible defects of substation operating equipment according to claim 1 or 2, characterized in that: The steps for calculating the correlation between the parent sequence and the subsequence collected by acoustic vibration include: The denoised signal collected by a certain acoustic vibration sensor is taken as the mother sequence, denoted as ,in, is the parent sequence element, Indicates The sound and vibration characteristic signals of each acoustic vibration collection point, ; The denoised signals after each primary equipment failure in the acoustic vibration defect library are taken as subsequences, denoted as ,in is the type of primary equipment failure, ; Preprocess the elements of the parent sequence and the child sequence, find the average value of the elements of each sequence, and then divide each element in the sequence by the mean of the sequence to obtain the preprocessed parent sequence elements. and the preprocessed subsequence elements : Forming a standardized matrix ; Calculate the grey correlation coefficient between the subsequence of the kth acoustic vibration acquisition point of the tth sampling channel and the parent sequence in is the resolution coefficient, and the grey relational degree of the t-th sampling channel is ; The one with the largest grey correlation degree indicates that a fault of this type has occurred, and the location of the fault can be determined based on the position of the acoustic vibration sensor.

7. The method for finding invisible defects of substation operating equipment according to claim 1 or 2, characterized in that: Before calculating the correlation of the acoustic vibration acquisition sequence, the acquired signal is first denoised by wavelet threshold: Set the number of wavelet decomposition layers, select the wavelet basis function, and construct the wavelet threshold function; Use the selected wavelet threshold function to process the noisy signal using the Mallat algorithm to obtain the wavelet high-frequency coefficients and low-frequency coefficients corresponding to the decomposition layer number; Compare the wavelet coefficient with the wavelet threshold. If the wavelet coefficient is greater than the wavelet threshold, the corresponding data bit is a valid signal, and the wavelet coefficient is retained. Otherwise, the signal is a noise signal and is discarded; The retained wavelet signal is reconstructed to obtain the denoised signal.

8. The method for finding hidden defects of substation operating equipment according to claim 1 or 2, characterized in that: The method for judging secondary signal abnormality is: For auxiliary contacts, respectively signals, and the signal before a device action is taken as the basic set, recorded as , the signal after a device action is used as the comparison set, recorded as , perform XOR calculation on the corresponding elements of the two data sets to obtain the signal generation set, recorded as If all the elements of the signal generation set are 0, it proves that the device has taken action once. If all are 1, it means that the device has not taken action once. If some elements are 1, it means that the device has taken action once but there is a problem with the auxiliary contact where the signal with the element 1 is located.

9. The method for finding invisible defects of substation operating equipment according to claim 1 or 2, characterized in that: The defect determination method is: when the descriptor of a feature point of two images changes, the number of feature point descriptors that change within its neighborhood is greater than It means that the feature point has shifted or loosened, and the point is recorded as a valid feature point. Indicates the minimum number of changes in the feature point descriptor. When the number of valid feature points is greater than When the device is offset or loose, Indicates the lower limit of the valid feature point value; The grey correlation degree takes the lower limit value, and a value greater than the lower limit value indicates that effective acoustic vibration information has been collected.

10. The method for finding hidden defects of substation operating equipment according to claim 9 is characterized in that: Comprehensive judgment methods include: Equipment displacement or looseness occurs, but no effective acoustic and vibration information is generated, and no secondary signal abnormality occurs, so it is considered that a general hidden danger has occurred on site; The equipment is offset or loose, and effective acoustic and vibration information is generated, but no secondary signal abnormality occurs, which indicates that a serious hidden danger has occurred on site; When equipment displacement or looseness occurs and effective acoustic vibration information is generated, and all related secondary abnormal signals occur, it is considered that a primary equipment defect has occurred; If there is no equipment displacement or looseness, no effective sound and vibration information occurs, but one or more secondary abnormal signals occur, it is assumed that there is no problem with the primary equipment and the secondary circuit contacts are checked.

Images