Substation equipment fault feature recognition method based on multispectral analysis

Abstract

The application belongs to the technical field of data processing and intelligent analysis, and particularly relates to a substation equipment fault feature identification method based on multispectral analysis, which comprises the following steps: acquiring infrared temperature distribution data and ultraviolet photon distribution data and aligning to obtain an infrared temperature distribution map and an ultraviolet photon distribution map; calculating infrared normalized rank values and ultraviolet normalized rank values based on the cumulative distribution characteristics of coordinate point intensity information, and calculating statistical correlation strength according to local joint distribution; obtaining fault saliency according to the spatial direction consistency of gradient vectors and the tail high value distribution characteristics of the infrared normalized rank values and the ultraviolet normalized rank values, and combining the statistical correlation strength; obtaining a fault region through adaptive threshold segmentation and spatial connectivity analysis, and outputting a position and a fault saliency mean value. The application reduces the risk of weak fault features being submerged by background noise in a complex environment, and improves the accuracy of fault identification.

Description

Technical Field

[0001] This invention relates to the field of data processing and intelligent analysis technology. More specifically, this invention relates to a method for identifying fault features of substation equipment based on multispectral analysis. Background Technology

[0002] As the core hub of the power system, substations often experience partial discharge or overheating of their high-voltage equipment during long-term operation due to insulation aging, poor contact, and other reasons. These phenomena are usually accompanied by thermal radiation and ultraviolet photon emission. Therefore, multispectral detection technology that integrates visible light, infrared thermal imaging, and ultraviolet imaging has become an important means of substation inspection, aiming to obtain equipment operating status information through the complementarity of information from different spectral bands.

[0003] Existing multispectral analysis techniques typically employ methods such as linear weighted averaging, principal component analysis, or wavelet transform to fuse multi-source monitoring data. This involves unifying the spatial reference of the distribution data in different spectral bands, then using numerical linear superposition or data feature weighting to enhance the physical field intensity or local spatial gradient information, and finally determining whether the equipment is faulty based on the fused numerical statistical characteristics.

[0004] However, the existing technologies described above have room for improvement in handling weak fault characteristics in substation environments. Because the physical mechanisms of substation equipment faults exhibit complex nonlinear coupling relationships between the infrared and ultraviolet spectral bands, simple linear weighting cannot accurately describe the higher-order statistical dependencies between temperature distribution and photon count distribution. When the fault is in its early stages and the signal intensity of a single spectral band is lower than the background noise, linear fusion methods tend to treat weak fault characteristics as a simple superposition of independent signals and background noise, causing the fault characteristics to be easily submerged by background noise. Furthermore, existing fusion methods struggle to distinguish between physically homogeneous fault signals and spurious correlation interference caused by environmental reflections, easily leading to false alarms. Summary of the Invention

[0005] To address the technical problem of false alarms caused by the flooding of fault characteristics, this invention provides a method for identifying fault characteristics of substation equipment based on multispectral analysis, including:

[0006] Infrared temperature distribution data and ultraviolet photon distribution data of substation equipment are acquired and spatially aligned to obtain spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps. Based on the cumulative distribution characteristics of intensity information of coordinate points within a preset neighborhood window, the infrared normalized rank value of each coordinate point on the infrared temperature distribution map and the ultraviolet normalized rank value on the ultraviolet photon distribution map are calculated respectively. According to the local joint distribution of the infrared normalized rank value and the ultraviolet normalized rank value, the statistical correlation strength of each coordinate point is calculated. The gradient vectors of the infrared temperature distribution map and the ultraviolet photon distribution map are calculated respectively. Based on the spatial direction consistency of the gradient vector and the tail high value distribution characteristics of the infrared normalized rank value and the ultraviolet normalized rank value, combined with the statistical correlation strength, the fault significance of each coordinate point is obtained. Adaptive threshold segmentation and spatial connectivity analysis are performed on the fault significance of each coordinate point to obtain the fault region, and the location of the fault region is output.

[0007] This invention acquires infrared temperature distribution data and ultraviolet photon distribution data of substation equipment and aligns them spatially to establish a unified spatial benchmark for multi-source data. Then, based on the cumulative distribution characteristics of intensity information at coordinate points within a preset neighborhood window, it calculates the normalized rank value, mapping temperature and photon fields with different physical dimensions to the same probability space. This eliminates the influence of dimensional differences between data sources on the fusion results. Furthermore, this invention calculates statistical correlation strength by evaluating local joint distribution and, combined with the spatial direction consistency of gradient vectors and the high-value distribution characteristics at the tail, comprehensively evaluates the fault significance of each coordinate point. Utilizing the synergy between the thermal and discharge field diffusion of real faults in the spatial direction, as well as the high-value characteristics of fault signals in statistical distribution, it distinguishes physically homogeneous fault signals from randomly distributed background noise or environmental interference. Finally, it extracts fault regions through adaptive threshold segmentation and spatial connectivity analysis, reducing misjudgments caused by weak signals in a single spectral band or environmental noise interference, thereby improving the accuracy of equipment fault feature identification in complex substation environments.

[0008] Preferably, the step of acquiring infrared temperature distribution data and ultraviolet photon distribution data of substation equipment and aligning them with spatial coordinates to obtain spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps includes:

[0009] Multispectral sensors are placed on substation inspection robots or fixed-point monitoring devices to acquire infrared temperature distribution data and ultraviolet photon distribution data of substation equipment, respectively. The infrared temperature distribution data and ultraviolet photon distribution data are spatially aligned using an affine transformation algorithm to establish a unified two-dimensional coordinate system, resulting in spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps.

[0010] Preferably, the infrared normalized rank value satisfies the following relationship:

[0011] ;

[0012] In the formula, Represents coordinate points Infrared normalized rank values ​​on infrared temperature distribution maps; This indicates the total number of coordinate points within the neighborhood window; Represents a neighborhood window; Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map; Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map; Represents the characteristic function.

[0013] This invention uses a combination of indicative functions and local cumulative counting to calculate the normalized rank value of infrared data. By statistically analyzing the proportion of infrared temperature values ​​less than or equal to the corresponding coordinate points within the neighborhood window of any coordinate point, the absolute temperature value is transformed into a probability rank value reflecting the local relative strength relationship. This approach preserves the details of local data distribution while reducing the impact of numerical fluctuations caused by differences in device emissivity or overall ambient temperature drift, providing unified and stable standardized input data for subsequent statistical dependency analysis of multi-source data.

[0014] Preferably, obtaining the ultraviolet normalized rank value includes:

[0015] The number of coordinate points whose ultraviolet photon values ​​are less than or equal to the corresponding coordinate point within the neighborhood window of any coordinate point on the ultraviolet photon distribution map is counted, and the number of coordinate points is divided by the total number of coordinate points within the neighborhood window to obtain the ultraviolet normalized rank value of the corresponding coordinate point on the ultraviolet photon distribution map.

[0016] Preferably, the statistical association strength satisfies the following relationship:

[0017] ;

[0018] In the formula, Represents coordinate points The statistical correlation strength; and Represents discrete traversal variables; This indicates the total number of coordinate points within the neighborhood window; Indicates the number of elements within the neighborhood window. Infrared normalized rank values ​​of each coordinate point; Indicates the number of elements within the neighborhood window. The normalized rank value of the ultraviolet radiation at each coordinate point; Represents the characteristic function.

[0019] This invention obtains the statistical correlation strength by calculating the Cramér-von Mises distance between the local empirical Copula function value and the theoretical joint distribution value. It utilizes the property of the local empirical Copula function to separate the dependency structure and marginal distribution between variables, evaluates the nonlinear correlation between the infrared normalized rank value and the ultraviolet normalized rank value in the local neighborhood, captures the complex symbiotic relationship between the infrared temperature distribution map and the ultraviolet photon distribution map, and distinguishes the statistical correlation of physically homogeneous fault signals from the random independence of background noise, thereby highlighting the statistical characteristics of the fault area of ​​substation equipment.

[0020] Preferably, the fault saliency satisfies the following relationship:

[0021] ;

[0022] In the formula, Represents coordinate points The severity of the fault; Represents coordinate points The statistical correlation strength; Represents coordinate points Gradient vector on the infrared temperature distribution map; Represents coordinate points Gradient vector on the ultraviolet photon distribution map; Represents coordinate points Infrared normalized rank values ​​on infrared temperature distribution maps; Represents coordinate points The normalized rank value of ultraviolet photon distribution on the ultraviolet photon distribution map; Represents the magnitude of a vector; Represents the maximum value function; This represents the dot product operation.

[0023] This invention applies a fault significance calculation relationship that includes statistical correlation strength, gradient vector, and normalized rank value. By evaluating the consistency of the infrared temperature distribution map and the ultraviolet photon distribution map in the spatial gradient direction, the statistical correlation strength is gain-modulated, which enhances the real fault signal that is consistent with the thermal field diffusion direction and the discharge field diffusion direction. At the same time, the high-value region is weighted by utilizing the high-value distribution characteristics of the tail, which suppresses artifact interference from disordered gradient directions or low intensity of single spectral bands, thereby obtaining a fault significance with a high signal-to-noise ratio.

[0024] Preferably, the step of performing adaptive threshold segmentation and spatial connectivity analysis on the fault saliency of each coordinate point to obtain the fault region includes:

[0025] Based on the Otsu algorithm, the fault significance of each coordinate point is adaptively thresholded to obtain the optimal segmentation threshold. The coordinate points with fault significance greater than the optimal segmentation threshold are extracted as potential fault coordinate points. Spatial connectivity analysis is performed on the potential fault coordinate points to connect broken areas and fill the gaps inside the areas, thus obtaining connected fault areas.

[0026] Preferably, the location of the fault area is output, including:

[0027] Calculate the average fault significance of all coordinate points within the fault area as the basis for assessing the severity of the fault, and output the location of the fault area.

[0028] Preferably, the setting of the preset neighborhood window includes:

[0029] Define a neighborhood window centered on any coordinate point.

[0030] Preferably, obtaining the gradient vector includes:

[0031] The gradient of each coordinate point on the infrared temperature distribution map and the ultraviolet photon distribution map is obtained by performing gradient calculation on the infrared temperature distribution map and the ultraviolet photon distribution map respectively using the Sobel operator.

[0032] This invention utilizes the Sobel operator to perform convolution operations on infrared temperature distribution maps and ultraviolet photon distribution maps to obtain gradient vectors. By using the Sobel operator to perform difference calculations in the horizontal and vertical directions, it captures the rate of change and directional information of the intensity at coordinate points in space. This provides vector data support for evaluating the spatial directional consistency between infrared temperature distribution maps and ultraviolet photon distribution maps, and helps to identify the directional thermal radiation field and ultraviolet photon field characteristics generated when a fault point in a substation equipment diffuses into the surrounding medium, thereby improving the ability to describe the edge of the fault morphology.

[0033] The beneficial effects of this invention are as follows: This invention proposes a fault identification scheme that integrates nonlinear statistical dependency analysis and spatial physical feature verification. Addressing the challenges of weak early-stage fault signals and complex coupling relationships between heterogeneous data in substation equipment, it utilizes probability integral transformation to obtain rank values ​​reflecting the relative magnitude relationships of the data, eliminating the dimensional gap and distribution differences between the infrared temperature field and the ultraviolet photon field. By evaluating the deviation between the joint cumulative distribution and the theoretical independent distribution of data within a local area, it captures the statistical co-occurrence characteristics of infrared and ultraviolet signals hidden under strong background noise, improving the shortcomings of linear weighting methods in describing higher-order dependencies. Simultaneously, this invention combines the spatial orientation consistency and tail-high value characteristics of gradient vectors to establish a signal modulation mechanism based on physical propagation laws. Utilizing the directional coordination generated when the real fault source diffuses outward, it weights and enhances the statistical correlation results and suppresses artifact interference from chaotic directions, thereby achieving high signal-to-noise ratio extraction and accurate localization of weak fault features in substation equipment under complex backgrounds. Attached Figure Description

[0034] Figure 1 The flowchart of the substation equipment fault feature identification method based on multispectral analysis in this invention is illustrated schematically.

[0035] Figure 2 This diagram illustrates the distribution of statistical association strength.

[0036] Figure 3 This diagram illustrates the distribution of fault saliency.

[0037] Figure 4 A schematic diagram of the fault area is shown. Detailed Implementation

[0038] This invention discloses a method for identifying fault features of substation equipment based on multispectral analysis, referring to... Figure 1 This includes steps S100-S400:

[0039] S100. Acquire infrared temperature distribution data and ultraviolet photon distribution data of substation equipment and perform spatial coordinate alignment to obtain spatially aligned infrared temperature distribution map and ultraviolet photon distribution map.

[0040] It should be noted that, because infrared temperature distribution data and ultraviolet photon distribution data have different fields of view and resolutions, directly superimposing the two sets of distribution data will lead to physical position misalignment, making feature fusion based on the same physical point impossible. Therefore, this invention uses spatial coordinate alignment to ensure that the infrared temperature distribution data and ultraviolet photon distribution data correspond to the same physical position on the device under the same coordinate system.

[0041] Specifically, multispectral sensors are placed on substation inspection robots or fixed-point monitoring devices to acquire infrared temperature distribution data and ultraviolet photon distribution data of substation equipment. A feature-point-based data registration algorithm is used to align the spatial coordinates of the infrared temperature distribution data and ultraviolet photon distribution data, establishing a unified two-dimensional coordinate system to obtain spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps. For example, the data registration algorithm is an affine transformation algorithm, which is existing technology and will not be elaborated upon here.

[0042] Thus, the infrared temperature distribution map and the ultraviolet photon distribution map were obtained.

[0043] S200. Based on the cumulative distribution characteristics of the intensity information of coordinate points within a preset neighborhood window, calculate the infrared normalized rank value of each coordinate point on the infrared temperature distribution map and the ultraviolet normalized rank value on the ultraviolet photon distribution map; calculate the statistical correlation strength of each coordinate point according to the local joint distribution of the infrared normalized rank value and the ultraviolet normalized rank value.

[0044] It should be noted that, since infrared temperature distribution maps represent temperature information and ultraviolet photon distribution maps represent photon count information, their physical dimensions and distribution characteristics differ, making it difficult to eliminate the influence of marginal distribution differences through direct numerical calculations. Therefore, this invention introduces a probability integral transformation to convert physical values ​​into rank values ​​that reflect relative magnitude relationships, thereby eliminating dimensional differences and mapping data from different sources to a unified probability space.

[0045] Specifically, based on the distribution of each coordinate point within its local neighborhood, the infrared normalized rank value of each coordinate point on the infrared temperature distribution map is calculated, including:

[0046] coordinate points Set a neighborhood window centered on the target area. For example, the neighborhood window size is 5×5. It should be noted that for coordinate points at the edge, the neighborhood window is constructed using mirror filling or zero-padding.

[0047] coordinates The normalized rank values ​​of the infrared temperature distribution map satisfy the following relationship:

[0048] ;

[0049] In the formula, Represents coordinate points Infrared normalized rank values ​​on infrared temperature distribution maps; This indicates the total number of coordinate points within the neighborhood window; Represents a neighborhood window; Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map; Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map; This represents an indicator function, which takes the value 1 when the condition within the parentheses is met, and 0 otherwise.

[0050] In this relation, This represents a binary intensity comparison performed on a single coordinate point within a neighborhood window. The infrared temperature value is less than or equal to the coordinate point The contribution value is 1 if it is active, and 0 otherwise. This represents the local cumulative count of the above comparison results, that is, the count within the neighborhood window where the infrared temperature value does not exceed the coordinate point. The total number of coordinate points. The higher the local cumulative count, the more coordinate points... The infrared temperature value is higher than the infrared temperature values ​​of most coordinate points in the local neighborhood, coordinate point A value that is locally relatively high indicates a high-temperature point in the infrared temperature distribution map; conversely, the smaller the value, the lower the coordinate point. The infrared temperature value is lower than that of most coordinate points in the local neighborhood. It is in a locally relatively low value state, that is, a low temperature point.

[0051] Preferably, the coordinates of the statistical ultraviolet photon distribution map are... The ultraviolet photon value within the neighborhood window is less than or equal to the coordinate point. The number of coordinate points is calculated, and this number is divided by the total number of coordinate points within the neighborhood window to obtain the coordinate points. Ultraviolet normalized rank values ​​on ultraviolet photon distribution maps .

[0052] Thus, the infrared normalized rank value and the ultraviolet normalized rank value corresponding to each coordinate point were obtained.

[0053] It should be noted that, since infrared and ultraviolet signals often exhibit a symbiotic relationship (i.e., statistical dependence) when a fault occurs, while background noise is usually random and independent, this invention characterizes the fundamental statistical dependence strength between infrared and ultraviolet signals within a local region by calculating the Cramér-von Mises distance between the local empirical Copula and the theoretical joint distribution value. This is because the Copula function can separate the joint distribution function from the marginal distribution function, thus focusing on the dependency structure between variables.

[0054] Preferably, the statistical correlation strength of each coordinate point is calculated based on the rank value distribution within the local neighborhood. The statistical association strength satisfies the following relationship:

[0055] ;

[0056] In the formula, Represents coordinate points The statistical correlation strength; and This represents a discrete traversal variable, with values ​​ranging from 0 to 1. ; This indicates the total number of coordinate points within the neighborhood window; Indicates the number of elements within the neighborhood window. Infrared normalized rank values ​​of each coordinate point; Indicates the number of elements within the neighborhood window. The normalized rank value of the ultraviolet radiation at each coordinate point; Represents the characteristic function.

[0057] In this relation, This indicates that at discrete grid points... The local empirical Copula function value calculated in real time reflects the joint cumulative distribution of the actual observation data; This represents the theoretical joint distribution value when the infrared normalized rank value and the ultraviolet normalized rank value are completely independent. Represents coordinate points The statistical correlation strength is defined as the Cramér-von Mises distance between the local empirical Copula function value and the theoretical joint distribution value. This Cramér-von Mises distance is calculated by letting... and Traversing the integer range from 1 to N, the squared deviations at each grid point are accumulated. The larger the Cramér-von Mises distance, the stronger the dependence between the infrared and ultraviolet normalized rank values ​​in the local neighborhood, and the further away from the independent state, corresponding to the fault region. Conversely, the smaller the Cramér-von Mises distance, the more independent the infrared and ultraviolet normalized rank values ​​tend to be, corresponding to the background noise region or the non-fault region.

[0058] For example, Figure 2 This is a schematic diagram of the statistical correlation strength distribution. The value of each coordinate point represents the statistical correlation strength between the infrared temperature distribution map and the ultraviolet photon distribution map in the local neighborhood. The larger the value, the more obvious the local joint distribution deviates from the independent state, corresponding to the fault area.

[0059] Thus, the statistical correlation strength of each coordinate point was obtained.

[0060] S300. Calculate the gradient vectors of the infrared temperature distribution map and the ultraviolet photon distribution map respectively; based on the spatial direction consistency of the gradient vectors and the tail high value distribution characteristics of the infrared normalized rank value and the ultraviolet normalized rank value, combined with the statistical correlation strength, obtain the fault significance of each coordinate point.

[0061] It should be noted that, considering two physical characteristics of real fault points: first, the thermal field and discharge field spread outward from the fault point, causing their spatial gradient directions to tend to be consistent; second, the fault signal is usually located in the high-value region of the data distribution. Therefore, this invention introduces gradient information and rank values ​​to nonlinearly modulate the statistical correlation strength of each coordinate point, thereby suppressing spurious correlations that are not physically homogeneous and enhancing the signal response in severely faulted regions, generating the fault salience of each coordinate point.

[0062] Specifically, based on gradient information and rank values, the fault significance at each coordinate point is calculated, including:

[0063] Gradient calculations were performed on the infrared temperature distribution map and the ultraviolet photon distribution map respectively to obtain the coordinate points. The gradient vectors on the infrared temperature distribution map and the ultraviolet photon distribution map. For example, the gradient vectors are calculated using the Sobel operator, which is prior art and will not be described in detail here.

[0064] coordinates The saliency of the fault satisfies the following relationship:

[0065] ;

[0066] In the formula, Represents coordinate points The severity of the fault; Represents coordinate points The statistical correlation strength; Represents coordinate points Gradient vector on the infrared temperature distribution map; Represents coordinate points Gradient vector on the ultraviolet photon distribution map; Represents coordinate points Infrared normalized rank values ​​on infrared temperature distribution maps; Represents coordinate points The normalized rank value of ultraviolet photon distribution on the ultraviolet photon distribution map; Represents the magnitude of a vector; Represents the maximum value function; This represents the dot product operation.

[0067] In this relation, Cosine similarity representing the gradient direction, when coordinates point When the gradient directions are consistent on the infrared temperature distribution map and the ultraviolet photon distribution map, the cosine similarity is close to 1, indicating that the thermal field diffusion direction is consistent with the discharge field diffusion direction, corresponding to fault signals with the same physical origin; the smaller the cosine similarity or the negative one, the more it corresponds to environmental interference or background noise with chaotic directions. This represents the tail-high weighting term based on the normalized rank values ​​of infrared and ultraviolet. A larger tail-high weighting term indicates that both the infrared temperature distribution map and the ultraviolet photon distribution map are in a locally relatively high state, corresponding to a severe fault region. Conversely, a smaller tail-high weighting term indicates that at least one of the normalized rank values ​​of infrared and ultraviolet is relatively small, i.e., it is in a background region or a single-mode noise region. The tail-high weighting term decreases, thereby suppressing the response in non-fault regions.

[0068] For example, Figure 3 This is a schematic diagram of the fault significance distribution. The value of each coordinate point is the fault significance obtained by combining the statistical correlation strength, the consistency of the gradient direction of the infrared temperature distribution map and the ultraviolet photon distribution map, and the high value weight of the tail. It is used to reflect the significance of physically homologous faults.

[0069] Thus, the fault significance of each coordinate point was obtained.

[0070] S400. Perform adaptive threshold segmentation and spatial connectivity analysis on the fault saliency of each coordinate point to obtain the fault region and output the location of the fault region.

[0071] It should be noted that in order to provide clear diagnostic suggestions to maintenance personnel, it is necessary to extract the specific fault area from the fault saliency of each coordinate point. In addition, the fault saliency of each coordinate point may have discrete noise points in its spatial distribution. Therefore, this invention achieves fault location through adaptive threshold segmentation and spatial connectivity analysis.

[0072] Specifically, an adaptive threshold segmentation method is used to statistically analyze the fault saliency of each coordinate point to obtain the optimal segmentation threshold. Coordinate points with fault saliency greater than the optimal segmentation threshold are extracted as potential fault coordinate points. Spatial connectivity analysis is then performed on these potential fault coordinate points to obtain spatially continuous fault regions. The mean fault saliency of all coordinate points within the fault region is calculated as the basis for assessing the fault severity and is output synchronously with the location of the fault region. For example, the adaptive threshold segmentation uses the Otsu algorithm, and the spatial connectivity analysis uses a spatial neighborhood topology constraint mechanism. Both the Otsu algorithm and the spatial neighborhood topology constraint mechanism are existing technologies and will not be elaborated upon here. This completes the identification and location of substation equipment fault characteristics.

[0073] For example, Figure 4This is a schematic diagram of the fault area, which identifies the fault area obtained through adaptive threshold segmentation and spatial connectivity analysis. It is used to locate substation equipment faults and assess the severity of the faults.

[0074] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for identifying fault features of substation equipment based on multispectral analysis, characterized in that, include: Infrared temperature distribution data and ultraviolet photon distribution data of substation equipment are acquired and spatially aligned to obtain spatially aligned infrared temperature distribution map and ultraviolet photon distribution map; Based on the cumulative distribution characteristics of the intensity information of coordinate points within a preset neighborhood window, the infrared normalized rank value of each coordinate point on the infrared temperature distribution map and the ultraviolet normalized rank value on the ultraviolet photon distribution map are calculated respectively; the infrared normalized rank values ​​satisfy the following relationship: , Represents coordinate points Infrared normalized rank values ​​on infrared temperature distribution maps This indicates the total number of coordinate points within the neighborhood window. Represents the neighborhood window, Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map Represents coordinate points Infrared temperature values ​​on an infrared temperature distribution map Indicates characteristic functions; Based on the local joint distribution of the infrared normalized rank values ​​and the ultraviolet normalized rank values, the statistical correlation strength of each coordinate point is calculated, satisfying the following relationship: , Represents coordinate points The statistical correlation strength, and Represents discrete traversal variables. This indicates the total number of coordinate points within the neighborhood window. , These represent the first and second elements within the neighboring window, respectively. Infrared normalized rank values ​​and ultraviolet normalized rank values ​​of each coordinate point; Calculate the gradient vectors of the infrared temperature distribution map and the ultraviolet photon distribution map respectively; based on the spatial orientation consistency of the gradient vectors and the high tail distribution characteristics of the infrared and ultraviolet normalized rank values, combined with the statistical correlation strength, obtain the fault significance of each coordinate point, satisfying the following relationship: , Represents coordinate points The severity of the fault Represents coordinate points Gradient vector on the infrared temperature distribution map, , Representing coordinate points Gradient vector and normalized rank value of ultraviolet photon distribution map Represents the magnitude of a vector. Represents the maximum value function. This represents the dot product operation; Adaptive threshold segmentation and spatial connectivity analysis are performed on the fault saliency of each coordinate point to obtain the fault region, and the location of the fault region is output.

2. The method for identifying fault features of substation equipment based on multispectral analysis according to claim 1, characterized in that, The process of acquiring infrared temperature distribution data and ultraviolet photon distribution data of substation equipment and aligning them with spatial coordinates to obtain spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps includes: Multispectral sensors are placed on substation inspection robots or fixed-point monitoring devices to acquire infrared temperature distribution data and ultraviolet photon distribution data of substation equipment, respectively. The infrared temperature distribution data and ultraviolet photon distribution data are spatially aligned using an affine transformation algorithm to establish a unified two-dimensional coordinate system, resulting in spatially aligned infrared temperature distribution maps and ultraviolet photon distribution maps.

3. The method for identifying fault features of substation equipment based on multispectral analysis according to claim 1, characterized in that, The acquisition of the ultraviolet normalized rank value includes: The number of coordinate points whose ultraviolet photon values ​​are less than or equal to the corresponding coordinate point within the neighborhood window of any coordinate point on the ultraviolet photon distribution map is counted, and the number of coordinate points is divided by the total number of coordinate points within the neighborhood window to obtain the ultraviolet normalized rank value of the corresponding coordinate point on the ultraviolet photon distribution map.

4. The substation equipment fault feature identification method based on multispectral analysis according to claim 1, characterized in that, The adaptive threshold segmentation and spatial connectivity analysis of the fault significance at each coordinate point yields the fault region, including: Based on the Otsu algorithm, the fault significance of each coordinate point is adaptively thresholded to obtain the optimal segmentation threshold. The coordinate points with fault significance greater than the optimal segmentation threshold are extracted as potential fault coordinate points. Spatial connectivity analysis is performed on the potential fault coordinate points to connect broken areas and fill the gaps inside the areas, thus obtaining connected fault areas.

5. The method for identifying fault features of substation equipment based on multispectral analysis according to claim 1, characterized in that, The location of the fault area is output, including: Calculate the average fault significance of all coordinate points within the fault area as the basis for assessing the severity of the fault, and output the location of the fault area.

6. The method for identifying fault features of substation equipment based on multispectral analysis according to claim 1, characterized in that, The settings for the preset neighborhood window include: Define a neighborhood window centered on any coordinate point.

7. The method for identifying fault features of substation equipment based on multispectral analysis according to claim 1, characterized in that, The acquisition of the gradient vector includes: The gradient of each coordinate point on the infrared temperature distribution map and the ultraviolet photon distribution map is obtained by performing gradient calculation on the infrared temperature distribution map and the ultraviolet photon distribution map respectively using the Sobel operator.

Images