Distribution line failure discrimination method based on aging identification fusion working condition parameters of image

Abstract

The invention belongs to the field of low-voltage power distribution loop and line safety monitoring, and discloses a power distribution line failure judgment method based on aging identification of an image and fusing working condition parameters, which is a failure mechanism judgment method based on a polyvinyl chloride copper conductor line. According to the method, working condition fusion analysis is carried out on the distribution line surface crack and damage micro-image, analysis processing of corresponding electrical parameters is combined as verification, and a distribution line failure judgment standard based on the distribution line surface crack and damage is given, so that scientific early warning is carried out on potential safety hazards of the distribution line; loss caused by faults of the distribution line is reduced; and a real-time detection and acquisition system is established by using the communication module, and model training is continuously performed on real-time data streams through online learning. The method can be expanded to other actual failure mechanism systems, such as mechanical equipment failure judgment, aerospace equipment failure judgment, medical equipment failure judgment and building facility and structure failure judgment, and has a wider application prospect.

Description

Technical Field

[0001] The present invention belongs to the field of low-voltage power distribution circuit and line safety monitoring, and relates to a power distribution line failure determination method based on image-based aging recognition and fusion of operating condition parameters. Background Art

[0002] In the current social development process, almost all systems are inseparable from the energy supply of the power system, and the distribution line is the lifeblood of the power system's energy transmission. In recent years, with the replacement of power systems and the increase in the operating life of most distribution lines, the aging of distribution lines has become more and more significant, and the failure of lines (i.e. retirement or replacement) has gradually become a research hotspot. The failure of distribution lines refers to the process of replacing or retiring distribution lines due to aging, damage or the need for technological upgrades, which usually involves several aspects such as failure rate and maintenance cost analysis, economic analysis, intelligent detection and early warning mechanism, environmental protection regulations and retirement standards. At present, the research on distribution line failure mainly focuses on three aspects: aging diagnosis technology, life cycle assessment model, and intelligent operation and maintenance and repair technology. However, in actual situations, the materials and working environments of distribution lines themselves are different, and the failure status judgment standards of distribution lines under different materials and different operating parameters are not the same. In today's society, where building distribution lines are generally aging, analyzing various materials for different voltage levels and integrating their operating parameters to quantify the distribution line failure judgment standards, scientifically evaluate the working conditions of distribution lines and provide risk warnings, have a greater preventive effect on building fires and accidents, and can also greatly reduce daily inspection costs and accident losses.

[0003] The image aging recognition technology used in the present invention is a subfield of image processing technology. It is a technology that detects damage to objects or surfaces by analyzing aging in images, and widely covers image fields such as industrial equipment monitoring, building structure maintenance, cultural relics protection, and medical image analysis. With the advancement of integrated circuit technology, the computing power of micro-embedded hardware has gradually increased. Many processors (such as ARM processors, FPGAs, etc.) and dedicated image processing chips can achieve complex image processing tasks under low power consumption conditions, which enables micro-image processing technology to be applied in devices with small size and high energy efficiency requirements, thus making micro-image aging recognition based on micro-embedded hardware possible.

[0004] At present, the industry's failure safety monitoring of distribution lines is mainly based on method demonstration, and there is a lack of quantitative discrimination mechanism for distribution line failure judgment under specific circumstances. Therefore, under this situation, the present invention is based on micro-image processing technology, through hardware platform and real-time processing, for China's building low-voltage power distribution commonly used polyvinyl chloride (PVC) copper wire, by obtaining, pre-processing, analyzing and post-processing the surface cracks and damage related parameters of the micro-distribution line image, combined with the corresponding working temperature, humidity, working condition parameters fusion for analysis and processing as verification, and the image is collected in real time corresponding to the distribution line working electrical parameters for experimental verification, aiming to provide an exemplary distribution line failure status evaluation standard based on distribution line surface cracks and damage, so as to make a scientific early warning of distribution line safety hazards. Summary of the invention

[0005] The purpose of the present invention is to provide a distribution line failure judgment method based on image-based aging recognition and integrated working condition parameters. Based on this technology, the working status of a polyvinyl chloride (PVC) copper conductor distribution line is monitored, and a distribution line failure condition judgment standard based on surface cracks and damage on the distribution line is given.

[0006] The technical method provided by the present invention relies on the feedback of hardware equipment, and is based on the distribution line image acquisition device and the electrical line parameter acquisition device installed in the distribution system distribution cabinet and the corresponding electrical circuit to collect the distribution line micro-image and the corresponding electrical line parameter. The hardware device is composed of the following modules: a single-chip microcomputer based on STM32, a camera module, an infrared measurement module, a humidity measurement module, an RFID module, a 4G communication module, etc. The hardware device is connected to the host computer in a 4G manner, and the data message communication is based on the UDP protocol; the acquisition circuit mainly includes a line sequence measurement circuit, a three-wire conductor impedance acquisition circuit, a voltage and current acquisition circuit, an RCD\GFCI action time acquisition circuit, a dedicated motor phase loss detection circuit, etc.

[0007] The technical solution of the present invention:

[0008] A distribution line failure judgment method based on image-based aging recognition and fusion of operating condition parameters, the steps are as follows:

[0009] Specifically, the failure mechanism identification method based on the polyvinyl chloride copper wire line is to collect electrical parameters and number the distribution cabinets with the same type of polyvinyl chloride copper wire, and at the same time, use the distribution line image acquisition device to collect the surface damage image of the corresponding numbered distribution line;

[0010] Step S1: installing the electrical circuit parameter acquisition device on the detection area, and using the RFID module to locate and identify the installation position of the electrical circuit parameter acquisition device;

[0011] Step S2: installing the distribution line image acquisition device in the acquisition area, and performing image acquisition on the distribution line in the detection area to obtain a micro-image of the distribution line; at the same time, starting the measurement and acquisition circuit, and acquiring various electrical parameters of the distribution line, including true effective value measurement of AC voltage, peak voltage, frequency, neutral line to ground voltage, three-wire conductor impedance and leakage protector action time, and using the temperature sensor and humidity sensor inside the electrical line parameter acquisition device to collect the working temperature and working humidity of the distribution line in real time; the collected electrical parameters and the working temperature and working humidity of the distribution line constitute the working condition fusion parameters, and the above working condition fusion parameters are wirelessly connected to the host computer through the communication module; the present invention refers to

[0012] Step S3: performing wavelet transform on each row and each column of the distribution line micro-image collected in the working state, and using a separation method to process the two-dimensional image data;

[0013] Assuming that the distribution line micro-image f(x,y) is an m×n two-dimensional matrix, the result after horizontal transformation is expressed as:

[0014] L(x,y)=∑ n f(x,n)·φ(n)

[0015] H(x,y)=∑ n f(x,n)·Ψ(n)

[0016] Among them, L(x,y) is the low-frequency component in the horizontal direction, H(x,y) is the high-frequency component in the horizontal direction; φ(n) is the scale function, which represents the low-frequency part; Ψ(n) is the wavelet function, which represents the high-frequency part;

[0017] Then, the horizontal low-frequency component L(x, y) and the high-frequency component H(x, y) are respectively subjected to vertical wavelet transform to obtain four sub-bands:

[0018] LL subband: Both horizontal and vertical directions are low-frequency components, representing the main contours and rough parts of the image;

[0019] LH subband: The horizontal direction is the low-frequency component, and the vertical direction is the high-frequency component, indicating the detail information in the vertical direction;

[0020] HL subband: The horizontal direction is the high-frequency component, and the vertical direction is the low-frequency component, indicating the detail information in the horizontal direction;

[0021] HH subband: Both horizontal and vertical directions are high-frequency components, representing the diagonal details and edges of the image;

[0022] The corresponding formula is:

[0023] Low-frequency components in the horizontal and vertical directions:

[0024] LL(x,y)=∑m∑nf(m,n)·φ(xm)·φ(yn)

[0025] Horizontal low frequency and vertical high frequency components:

[0026] LH(x,y)=∑m∑nf(m,n)·φ(xm)·Ψ(yn)

[0027] High frequency components in the horizontal direction and low frequency components in the vertical direction:

[0028] HL(x,y)=∑m∑nf(m,n)·Ψ(xm)·φ(yn)

[0029] High frequency components in horizontal and vertical directions:

[0030] HH(x,y)=∑m∑nf(m,n)·Ψ(xm)·Ψ(yn)

[0031] Where f(m,n) represents the pixel value of the input image; x and y are the coordinate indices after transformation;

[0032] Then, the distribution line micro-image collected in the working state is processed from color image to grayscale image. The RGB channels are processed mainly by weighted average method, and the brightness information is merged into a single grayscale value to form the original grayscale image. The calculation formula is as follows:

[0033] G ix =0.299×R+0.587×G+0.114×B

[0034] Among them, G ix Represents the grayscale value of the original grayscale image, R, G, and B represent the pixel values ​​of the red, green, and blue color channels respectively;

[0035] Combined with the working condition fusion parameters in step S2, the grayscale image of the distribution line is compensated. First, the compensation of the grayscale image of the cracks and damage on the surface of the distribution line by the ambient temperature is as follows:

[0036] Ambient temperature T a :

[0037]

[0038] Where VDD is the power supply voltage of the temperature sensor; VDD 25 is the nominal power supply voltage at 25°C, which is 3.3V; PTAT is the temperature sensor output value read by the ADC inside the main control module; K Tais the sensitivity coefficient of the temperature sensor, which is read from the internal EEPROM of the temperature sensor; PTATSCALE is a calibration coefficient, which is read from the internal EEPROM of the temperature sensor;

[0039] The temperature calculation formula for a single pixel is:

[0040]

[0041] Among them, P ix is the original infrared signal of the ixth pixel, which is sampled by the ADC module inside the main control module; α corr is the correction coefficient, which is read from the EEPROM inside the temperature sensor and is used to compensate for the temperature drift of the temperature sensor; ε is the emissivity of the PVC copper wire line; S ix It is the sensitivity parameter of each pixel, which is read from the internal EEPROM of the temperature sensor;

[0042] Based on the above principles, the temperature compensation formula for the grayscale image of the distribution line is as follows:

[0043]

[0044] Among them, G ix is the gray value of the ixth pixel in the original gray image, representing the uncompensated temperature; is the gray value after compensation, indicating the corrected temperature; G ref It is the reference grayscale value, which indicates the grayscale value corresponding to the reference temperature when the sensor is calibrated, and is obtained by pixel calibration; K S1 is the sensitivity correction factor, which is used to correct the sensitivity as the ambient temperature T a The effect of the change is read from the temperature sensor's internal EEPROM; T ref is the reference temperature of the temperature sensor, which is read from the internal EEPROM of the temperature sensor;

[0045] In addition, the compensation principle of ambient humidity on the grayscale image of cracks and damage on the surface of the distribution line is as follows. The formula of the capacitance value C of the selected capacitive humidity sensor is:

[0046]

[0047] Where C is the capacitance value; ε r is the dielectric constant of the humidity-sensitive material, which changes with humidity; ε 0 is the dielectric constant of vacuum, which is 8.854×10^{-12}F / m; A is the effective area of ​​the electrode, square meters; d is the distance between the electrodes, meters;

[0048] The capacitance value of the capacitive humidity sensor is linear or approximately linear with the relative humidity RH, which can be described by the following empirical formula:

[0049] C=C 0 (1+k·RH);

[0050] Among them, C 0 is the capacitance value in dry air; k is the sensitivity coefficient of the sensor; RH is the relative humidity, ranging from 0 to 1;

[0051] Based on the above principles, consider the initial grayscale value G of the grayscale image ix Affected by humidity RH, the algorithm compensation formula of ambient humidity for the grayscale image of the distribution line is as follows:

[0052]

[0053] in, is the real gray value after compensation; G ix is the gray value of the original image measured under humidity RH; RH 0 is the reference humidity; RH is the humidity corresponding to the current image; k is the humidity compensation coefficient, which is related to the characteristics of the humidity sensor and the sensitivity of the image brightness change;

[0054] Step S4: The analysis of the crack and damage data on the circuit surface relies on fractal theory; to avoid losing image information details, the differential box dimension method is selected to solve the fractal dimension of the grayscale image; where X, Y, and Z represent the position and grayscale value of the pixel in the plane, respectively, and the image grayscale value forms an uneven surface in the three-dimensional space, then the three-dimensional space represented by X, Y, and Z is divided into several cubic boxes with a side length of r; assuming that the maximum and minimum values ​​of the image grayscale at the position X=i, Y=j fall in the Kth box and the Lth box respectively, then:

[0055] N r (x i ,y j )=L-K+1

[0056] Among them, N r (x i ,y j ) is the number of boxes required to cover the image at position (i, j) in the XY plane; the number of boxes required to cover the entire image is:

[0057] N r =∑ i,j N r (x i ,y j )

[0058] The fractal dimension of the image is:

[0059]

[0060] For different numbers of boxes, the two-dimensional fractal dimension D is obtained by the least squares method;

[0061] To calculate the three-dimensional fractal dimension based on the two-dimensional fractal, for each scale r of the cube box side length, the number of non-empty cubes N(r) containing data points in the three-dimensional space grid represented by X, Y, and Z needs to be counted; this counting process needs to be repeated for different r values; as r decreases, N(r) at different scales is recorded; finally, the sizes r1 and r2 with better effects are selected as experimental sizes by fitting the data. At this time, the approximate formula for calculating the three-dimensional fractal dimension is:

[0062]

[0063] Among them, r1 and r2 are different scales, and N(r1) and N(r2) are the number of non-empty cubes counted at these scales respectively;

[0064] After obtaining the two-dimensional fractal dimension D and three-dimensional fractal dimension E corresponding to the grayscale image of the distribution line after temperature and humidity compensation, combined with the specific experimental conditions, the failure coefficient δ of the distribution line at a certain position (x, y) is defined as the two-dimensional fractal dimension D at that location. (x,y) and the three-dimensional fractal dimension E (x,y) The monitoring matrix is ​​composed of the following conclusions after data analysis in the experimental environment: when the damage feature of the distribution line surface is wear, the failure coefficient δ is between 0.2 and 0.4; when the damage feature is rust, the failure coefficient δ is between 0.4 and 0.5; when the damage feature is scar, the failure coefficient δ is between 0.5 and 0.7;

[0065] Step S5: Compare and verify the failure coefficient with the corresponding real-time electrical parameters of the distribution line in step S2 to improve the scientificity and rationality of the failure coefficient;

[0066] Step S6: The upper computer performs data processing and failure coefficient alarm on the grayscale image of the distribution line collected under the above conclusion on the failure coefficient, and sends the alarm data to the on-site personnel through the communication module with the location information and corresponding ID of the collection device for actual secondary measurement. If there is any problem, it will be replaced in time to prevent accidents.

[0067] Beneficial effects of the present invention:

[0068] (1) The present invention integrates the working condition analysis of the micro-images of the cracks and damages on the surface of the distribution line, and combines the analysis and processing of the corresponding electrical parameters as verification, aiming to provide a failure evaluation standard for the distribution line based on the cracks and damages on the surface of the distribution line, so as to make a scientific early warning of the safety hazards of the distribution line and reduce the losses caused by the failure of the distribution line; and uses the communication module to establish a real-time detection and acquisition system, which can continuously train the model of the real-time data stream through online learning;

[0069] (2) The failure identification method of the present invention can be extended to other actual failure mechanism systems, such as mechanical equipment failure identification, aerospace equipment failure identification, medical equipment failure identification, and building facility and structure failure identification, and has broader application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is the overall flow chart of the present invention;

[0071] Figure 2 It is a module structure diagram of the hardware device for micro-image acquisition of power distribution lines of the present invention;

[0072] Figure 3 It is a module structure diagram of the hardware device for collecting electrical parameters of distribution lines used in the present invention;

[0073] Figure 4 1 is a structural diagram of a power distribution line micro-image acquisition device used in the present invention; wherein (a) is a front view and (b) is a side view;

[0074] Figure 5 It is a structural diagram of the distribution line electrical line parameter acquisition device used in the present invention; wherein (a) is the main view and (b) is the side view. DETAILED DESCRIPTION

[0075] The specific implementation of the present invention is further described below in conjunction with the accompanying drawings and technical solutions.

[0076] The present invention is a method for distinguishing the failure mechanism of a copper wire line made of polyvinyl chloride (PVC). Therefore, during the experiment, the distribution cabinets of the same type of polyvinyl chloride (PVC) wires are used to collect electrical parameters and number them, and at the same time, a micro-image acquisition device is used to collect surface damage images of the corresponding numbered distribution lines. To ensure the accuracy of the experimental results, the wire surface needs to be cleaned before the experiment, and to prevent the wire from being worn, a small amount of grease is applied to the wire surface, and finally alcohol is used to remove the oil stains on the wire surface and the wire is dried naturally.

[0077] Step S1: Figure 5 The electrical line parameter acquisition device is installed on the detection area and uses Figure 3The RFID module locates the installation position of the device and identifies the ID; before collecting electrical parameters, Figure 4 The distribution line image acquisition device is installed in a suitable acquisition area to acquire images of the polyvinyl chloride (PVC) distribution lines in the detection area to obtain a micro-image of the distribution line.

[0078] Step S2: After the distribution line image acquisition device is turned on, the distribution line in the detection area is imaged to obtain a distribution line micro-image; at the same time, the measurement and acquisition circuit is turned on to collect various electrical parameters of the distribution line, including true effective value measurement of AC voltage, peak voltage, frequency, neutral line to ground voltage, three-wire conductor impedance, leakage protector action time and other output data, and the electrical parameter acquisition device installed in the wire detection area in step S1 is used as Figure 2 The temperature sensor and humidity sensor shown in the figure monitor the working temperature and humidity of the power distribution circuit in real time. The data collected in real time in this step is communicated with the host computer through Figure 2 , Figure 3 The communication module in the system is connected wirelessly to achieve the purpose of remote monitoring and collection.

[0079] Step S3: Perform wavelet transform on the distribution line micro-image collected in the working state, so as to obtain the approximate value reflecting the overall situation of the image and the detail value in each direction. For any image (two-dimensional signal), wavelet transform is performed on each row and each column of the image respectively, and separation method is usually used to process the two-dimensional image data.

[0080] Assuming that the image f(x,y) is an m×n two-dimensional matrix, the result after horizontal transformation can be expressed as

[0081] L(x,y)=∑ n f(x,n)·φ(n)

[0082] H(x,y)=∑ n f(x,n)·Ψ(n)

[0083] Here, L(x, y) is the low-frequency component in the horizontal direction, H(x, y) is the high-frequency component in the horizontal direction; φ(n) is the scale function, which represents the low-frequency part; Ψ(n) is the wavelet function, which represents the high-frequency part;

[0084] Then perform vertical wavelet transform on the low frequency L(x, y) and high frequency H(x, y) in the horizontal direction to obtain four sub-bands:

[0085] LL subband: It is a low-frequency component in both horizontal and vertical directions, representing the main contour and rough parts of the image.

[0086] LH subband: The horizontal direction is the low-frequency component, and the vertical direction is the high-frequency component, indicating the detail information in the vertical direction.

[0087] HL subband: The horizontal direction is the high-frequency component, and the vertical direction is the low-frequency component, indicating the detail information in the horizontal direction.

[0088] HH subband: Both the horizontal and vertical directions are high-frequency components, representing the diagonal details and edges of the image.

[0089] The corresponding formula is:

[0090] Low-frequency components in the horizontal and vertical directions:

[0091] LL(x,y)=∑m∑nf(m,n)·φ(xm)·φ(yn)

[0092] Horizontal low frequency and vertical high frequency components:

[0093] LH(x,y)=∑m∑nf(m,n)·φ(xm)·Ψ(yn)

[0094] High frequency components in the horizontal direction and low frequency components in the vertical direction:

[0095] HL(x,y)=∑m∑nf(m,n)·Ψ(xm)·φ(yn)

[0096] High frequency components in horizontal and vertical directions:

[0097] HH(x,y)=∑m∑nf(m,n)·Ψ(xm)·Ψ(yn)

[0098] Among them, f(m,n) represents the pixel value of the input image, and x and y are the coordinate indices after transformation.

[0099] Then, the distribution line micro-image collected in the working state is processed into grayscale image by color image conversion. The RGB channels are processed mainly by weighted average method, and their brightness information is merged into a single grayscale value to form a grayscale image. The calculation formula is as follows

[0100] GRAY=0.299×R+0.587×G+0.114×B

[0101] Where GRAY represents the grayscale value of the grayscale image, and R, G, and B represent the pixel values ​​of the red, green, and blue color channels respectively. This weighting coefficient reflects the sensitivity of the human eye to different colors, with green being the most sensitive and blue being the least sensitive.

[0102] Combined with the working condition fusion parameters in step S2, the grayscale image of the polyvinyl chloride (PVC) distribution line is algorithmically compensated. First, the algorithm compensation principle of the ambient temperature on the grayscale image of cracks and damage on the distribution line surface is as follows:

[0103] Ambient temperature T a :

[0104]

[0105] Where: VDD is the current power supply voltage. 25 is the nominal supply voltage at 25°C (usually 3.3V). PTAT is the raw temperature sensor output value read by the ADC. K Ta is the sensitivity coefficient of the temperature sensor, read from the EEPROM. PTATSCALE is a calibration coefficient used to adjust the ratio in the formula, usually read from the EEPROM.

[0106] The temperature calculation formula for a single pixel is:

[0107]

[0108] In the formula, where: P ix is the original infrared signal of the ixth pixel (obtained through ADC sampling). corr is the correction factor, read from the EEPROM, used to compensate for the temperature drift of the sensor. ε is the emissivity of the object (usually the emissivity is assumed to be 1, if the emissivity of the target object is known, it can be adjusted accordingly). ix It is the sensitivity parameter of each pixel, which is read from EEPROM.

[0109] Based on the above principles, the temperature compensation formula for the grayscale image of the distribution line is as follows:

[0110]

[0111] Where: G ix is the gray value of the ixth pixel in the original gray image, representing the uncompensated temperature; is the gray value after compensation, indicating the corrected temperature; G ref is the reference grayscale value, which indicates the grayscale value corresponding to the reference temperature when the sensor is calibrated, and can be obtained through pixel calibration; ε is the emissivity of the device surface (usually obtained from the physical properties of the device material). If it is uncertain, ε = 1 is usually assumed (assuming that the device surface is close to blackbody radiation); K S1 is the sensitivity correction factor, which is used to correct the sensitivity as the ambient temperature T a To avoid the effect of changes, this value is read from the EEPROM.a is the ambient temperature measured by the sensor, and how to calculate it has been discussed previously. ref is the reference temperature of the sensor, and this value is also read from the EEPROM.

[0112] In addition, the algorithm compensation principle of ambient humidity for the grayscale image of cracks and damage on the surface of the distribution line is as follows. The formula of the capacitance value C of the selected capacitive humidity sensor is:

[0113]

[0114] Where: C is the capacitance value (unit: Farad); ε r is the dielectric constant of the humidity-sensitive material, which changes with humidity; ε 0 is the dielectric constant of vacuum, which is approximately 8.854×10^{-12}F / m; A is the effective area of ​​the electrode (unit: square meters); d is the distance between the electrodes (unit: meter).

[0115] In general, the capacitance value of a capacitive humidity sensor is linear or approximately linearly related to the relative humidity RH, which can be described by the following empirical formula:

[0116] C=C 0 (1+k·RH)

[0117] Where: C 0 is the capacitance value in dry air; k is the sensitivity coefficient of the sensor; RH is the relative humidity, which ranges from 0 to 1 (or 0% to 100%).

[0118] Based on the above principles, consider the gray value G of the image ix Affected by humidity RH, the algorithm compensation formula of ambient humidity for the grayscale image of the distribution line is as follows:

[0119]

[0120] in: is the actual gray value after compensation (corresponding to humidity RH 0 ). ix is the grayscale value of the original image measured under humidity RH. 0 is the reference humidity (e.g., humidity under standard humidity, which may be 50%RH or 0%RH). RH is the humidity corresponding to the current image. k is the humidity compensation coefficient, which is related to the sensor characteristics and the sensitivity of image brightness changes.

[0121] Step S4: The analysis of the crack and damage data on the line surface relies on fractal theory, which is mainly used to study the non-smooth and irregular geometric shapes that appear in nature or complex nonlinear systems. It starts directly from the nonlinear system itself and understands the internal laws from the unsimplified and abstract research object itself. The fractal dimension is a very important indicator in fractal theory to characterize the surface state of an object, and is used to describe the roughness of the surface morphology of an object. The two-dimensional fractal dimension is between 1 and 2, and is used to obtain information on the surface state of the image focus area; the three-dimensional fractal dimension is between 2 and 3, and is used to characterize the unevenness of the image.

[0122] In order to avoid losing image information details, the differential box dimension method is used to solve the fractal dimension of the grayscale image; where X, Y, and Z represent the position and grayscale value of the pixel in the plane respectively, and the image grayscale value forms an uneven surface in the three-dimensional space. The three-dimensional space represented by X, Y, and Z is divided into several cubic boxes with a side length of r. Assume that the maximum and minimum values ​​of the image grayscale at the position X=i, Y=j fall in the Kth box and the Lth box respectively, then

[0123] N r (x i ,y j )=L-K+1

[0124] Among them, N r (x i ,y j ) is the number of boxes required to cover the image in position (i, j) in the XY plane. The number of boxes required to cover the entire image is

[0125] N r =∑ i,j N r (x i ,y j )

[0126] The fractal dimension of the image is

[0127]

[0128] For different numbers of boxes, the two-dimensional fractal dimension D can be obtained by the least squares method.

[0129] To calculate the three-dimensional fractal dimension based on the two-dimensional fractal, for each scale r, count the non-empty cubes containing data points in the grid. For each cube, count the data difference (voxel value) between the minimum and maximum values ​​it contains. In three-dimensional space, this is equivalent to considering the voxel distribution in the grid box. For each scale r, count the number of all non-empty cubes N(r). This counting process needs to be repeated for different r values. As r decreases, record N(r) at different scales. Usually multiple scales r are selected. Finally, the sizes r1 and r2 with better effects are selected as experimental sizes by fitting the data. At this time, the approximate formula for calculating the three-dimensional fractal dimension is:

[0130]

[0131] Here, r1 and r2 are different scales, and N(r1) and N(r2) are the number of non-empty cubes counted at these scales, respectively.

[0132] After obtaining the two-dimensional fractal dimension D and three-dimensional fractal dimension E corresponding to the grayscale image of the distribution line after temperature and humidity compensation, combined with the specific experimental conditions, the failure coefficient δ of the distribution line at a certain position (x, y) is defined as the two-dimensional fractal dimension D at that location. (x,y) and the three-dimensional fractal dimension E (x,y) The monitoring matrix composed of the components is composed of the components of the monitoring matrix ...

[0133] Step S5: Compare and verify the failure coefficient with the corresponding real-time electrical parameters of the distribution line in step S2 to improve the scientificity and rationality of the failure coefficient.

[0134] Step S6: The host computer extracts and highlights the distribution line image collected under the current model, and sends the alarm data to the on-site personnel through the communication module with the location information and corresponding ID of the collection device for actual secondary measurement. If there is a problem, it will be replaced in time to prevent accidents.

Claims

1. A distribution line failure identification method based on image-based aging recognition and fusion of operating condition parameters, characterized in that: Here are the steps: Specifically, the failure mechanism identification method based on the polyvinyl chloride copper wire line is to collect electrical parameters and number the distribution cabinets with the same type of polyvinyl chloride copper wire, and at the same time, use the distribution line image acquisition device to collect the surface damage image of the corresponding numbered distribution line; Step S1: installing the electrical circuit parameter acquisition device on the detection area, and using the RFID module to locate and identify the installation position of the electrical circuit parameter acquisition device; Step S2: installing the distribution line image acquisition device in the acquisition area, and performing image acquisition on the distribution line in the detection area to obtain a micro-image of the distribution line; at the same time, starting the measurement and acquisition circuit, and acquiring various electrical parameters of the distribution line, including true effective value measurement of AC voltage, peak voltage, frequency, neutral line to ground voltage, three-wire conductor impedance and leakage protector action time, and using the temperature sensor and humidity sensor inside the electrical line parameter acquisition device to collect the working temperature and working humidity of the distribution line in real time; the collected electrical parameters and the working temperature and working humidity of the distribution line constitute the working condition fusion parameters, and the above working condition fusion parameters are wirelessly connected to the host computer through the communication module; the present invention refers to Step S3: performing wavelet transform on each row and each column of the distribution line micro-image collected in the working state, and using a separation method to process the two-dimensional image data; Assuming that the distribution line micro-image f(x,y) is an m×n two-dimensional matrix, the result after horizontal transformation is expressed as: L(x,y)=Σ n f(x,n)·φ(n) H(x,y)=Σ n f(x,n)·Ψ(n) Among them, L(x,y) is the low-frequency component in the horizontal direction, H(x,y) is the high-frequency component in the horizontal direction; φ(n) is the scale function, which represents the low-frequency part; Ψ(n) is the wavelet function, which represents the high-frequency part; Then, the horizontal low-frequency component L(x, y) and the high-frequency component H(x, y) are respectively subjected to vertical wavelet transform to obtain four sub-bands: LL subband: Both horizontal and vertical directions are low-frequency components, representing the main contours and rough parts of the image; LH subband: The horizontal direction is the low-frequency component, and the vertical direction is the high-frequency component, indicating the detail information in the vertical direction; HL subband: The horizontal direction is the high-frequency component, and the vertical direction is the low-frequency component, indicating the detail information in the horizontal direction; HH subband: Both horizontal and vertical directions are high-frequency components, representing the diagonal details and edges of the image; The corresponding formula is: Low-frequency components in the horizontal and vertical directions: LL(x,y)=Σm∑nf(m,n)·φ(xm)·φ(yn) Horizontal low frequency and vertical high frequency components: LH(x,y)=∑m∑nf(m,n)·φ(xm)·Ψ(yn) High frequency components in the horizontal direction and low frequency components in the vertical direction: HL(x,y)=∑m∑nf(m,n)·Ψ(xm)·φ(yn) High frequency components in horizontal and vertical directions: HH(x,y)=∑m∑nf(m,n)·Ψ(xm)·Ψ(yn) Where f(m,n) represents the pixel value of the input image; x and y are the coordinate indices after transformation; Then, the distribution line micro-image collected in the working state is processed from color image to grayscale image. The RGB channels are processed mainly by weighted average method, and the brightness information is merged into a single grayscale value to form the original grayscale image. The calculation formula is as follows: G ix =0.299×R+0.587×G+0.114×B Among them, G ix Represents the grayscale value of the original grayscale image, R, G, and B represent the pixel values ​​of the red, green, and blue color channels respectively; Combined with the working condition fusion parameters in step S2, the grayscale image of the distribution line is compensated. First, the compensation of the grayscale image of the cracks and damage on the surface of the distribution line by the ambient temperature is as follows: Ambient temperature T a : Where VDD is the power supply voltage of the temperature sensor; VDD 25 is the nominal power supply voltage at 25°C, which is 3.3V; PTAT is the temperature sensor output value read by the ADC inside the main control module; K Ta is the sensitivity coefficient of the temperature sensor, which is read from the internal EEPROM of the temperature sensor; PTATSCALE is a calibration coefficient, which is read from the internal EEPROM of the temperature sensor; The temperature calculation formula for a single pixel is: Among them, P ix is the original infrared signal of the ixth pixel, which is sampled by the ADC module inside the main control module; α corr is the correction coefficient, which is read from the EEPROM inside the temperature sensor and is used to compensate for the temperature drift of the temperature sensor; ε is the emissivity of the PVC copper wire line; S ix It is the sensitivity parameter of each pixel, which is read from the internal EEPROM of the temperature sensor; Based on the above principles, the temperature compensation formula for the grayscale image of the distribution line is as follows: Among them, G ix is the gray value of the ixth pixel in the original gray image, representing the uncompensated temperature; is the gray value after compensation, indicating the corrected temperature; G ref It is the reference grayscale value, which indicates the grayscale value corresponding to the reference temperature when the sensor is calibrated, and is obtained by pixel calibration; K S1 is the sensitivity correction factor, which is used to correct the sensitivity as the ambient temperature T a The effect of the change is read from the temperature sensor's internal EEPROM; T ref is the reference temperature of the temperature sensor, which is read from the internal EEPROM of the temperature sensor; In addition, the compensation principle of ambient humidity on the grayscale image of cracks and damage on the surface of the distribution line is as follows. The formula of the capacitance value C of the selected capacitive humidity sensor is: Where C is the capacitance value; ε r is the dielectric constant of the hygroscopic material, which changes with humidity; ε0 is the dielectric constant of vacuum, which is 8.854×10^{-12}F / m; A is the effective area of ​​the electrode, in square meters; d is the distance between the electrodes, in meters; The capacitance value of the capacitive humidity sensor is linear or approximately linear with the relative humidity RH, which can be described by the following empirical formula: C = C0 (1 + k·RH); Where C0 is the capacitance value in dry air; k is the sensitivity coefficient of the sensor; RH is the relative humidity, ranging from 0 to 1; Based on the above principles, consider the initial grayscale value G of the grayscale image ix Affected by humidity RH, the algorithm compensation formula of ambient humidity for the grayscale image of the distribution line is as follows: in, is the real gray value after compensation; G ix is the grayscale value of the original image measured under humidity RH; RH0 is the reference humidity; RH is the humidity corresponding to the current image; k is the humidity compensation coefficient, which is related to the characteristics of the humidity sensor and the sensitivity of the image brightness change; Step S4: The analysis of the crack and damage data on the circuit surface relies on fractal theory; to avoid losing image information details, the differential box dimension method is selected to solve the fractal dimension of the grayscale image; where X, Y, and Z represent the position and grayscale value of the pixel in the plane, respectively, and the image grayscale value forms an uneven surface in the three-dimensional space, then the three-dimensional space represented by X, Y, and Z is divided into several cubic boxes with a side length of r; assuming that the maximum and minimum values ​​of the image grayscale at the position X=i, Y=j fall in the Kth box and the Lth box respectively, then: N r (x i ,y j )=L-K+1 Among them, N r (x i ,y j ) is the number of boxes required to cover the image at position (i, j) in the XY plane; the number of boxes required to cover the entire image is: N r ∑ i,j N r (x i ,y j ) The fractal dimension of the image is: For different numbers of boxes, the two-dimensional fractal dimension D is obtained by the least squares method; To calculate the three-dimensional fractal dimension based on the two-dimensional fractal, for each scale r of the cube box side length, the number of non-empty cubes N(r) containing data points in the three-dimensional space grid represented by X, Y, and Z needs to be counted; this counting process needs to be repeated for different r values; as r decreases, N(r) at different scales is recorded; finally, the sizes r1 and r2 with better effects are selected as experimental sizes by fitting the data. At this time, the approximate formula for calculating the three-dimensional fractal dimension is: Among them, r1 and r2 are different scales, and N(r1) and N(r2) are the number of non-empty cubes counted at these scales respectively; After obtaining the two-dimensional fractal dimension D and three-dimensional fractal dimension E corresponding to the grayscale image of the distribution line after temperature and humidity compensation, combined with the specific experimental conditions, the failure coefficient δ of the distribution line at a certain position (x, y) is defined as the two-dimensional fractal dimension D at that location. (x,y) and the three-dimensional fractal dimension E (x,y) The monitoring matrix is ​​composed of the following conclusions after data analysis in the experimental environment: when the damage feature of the distribution line surface is wear, the failure coefficient δ is between 0.2 and 0.4; when the damage feature is rust, the failure coefficient δ is between 0.4 and 0.5; when the damage feature is scar, the failure coefficient δ is between 0.5 and 0.7; Step S5: Compare and verify the failure coefficient with the corresponding real-time electrical parameters of the distribution line in step S2 to improve the scientificity and rationality of the failure coefficient; Step S6: The upper computer performs data processing and failure coefficient alarm on the grayscale image of the distribution line collected under the above conclusion on the failure coefficient, and sends the alarm data to the on-site personnel through the communication module with the location information and corresponding ID of the collection device for actual secondary measurement. If there is any problem, it will be replaced in time to prevent accidents.

Images