An air-ground integrated substation overheat fault detection system based on YOLO algorithm

Abstract

This invention relates to the field of substation equipment condition monitoring and fault diagnosis technology, and particularly to an integrated air-ground substation overheating fault detection system based on the YOLO algorithm. The system includes a pseudo-thermal image identification module, used to determine whether material correction or equipment overheating fault alarm should be performed based on the temperature uniformity of the substation unit; a material correction module, used to perform observation correction or adjacent compensation correction based on thermal correlation; and a misjudgment suppression module, used to determine whether to correct the infrared image based on the UAV rotor speed based on the misjudgment rate. By distinguishing between two causes—material emissivity differences and pseudo-thermal images reflected from adjacent equipment—through thermal correlation, the corrected judgment can truly reflect the overheating state of the tested substation unit itself, rather than being distorted by surface reflection and environmental emissivity, further improving the accuracy of substation overheating fault detection.

Description

Technical Field

[0001] This invention relates to the field of substation equipment condition monitoring and fault diagnosis technology, and in particular to an air-ground integrated substation overheating fault detection system based on the YOLO algorithm. Background Technology

[0002] Conductive connections in substations, such as disconnector contacts, busbar clamps, and bushing joints, can overheat abnormally when bolts loosen, surfaces oxidize, or contact resistance increases. If not addressed promptly, this can lead to joint melting, equipment fires, and even large-scale power outages. Traditional methods rely on maintenance personnel manually inspecting substations using infrared thermal imagers, which is inefficient, risky, and heavily influenced by subjective experience. In recent years, using drones and ground robots for automated infrared thermography has become mainstream. However, the systematic differences in observation angles and material reflectivity between air and ground platforms remain significant challenges in obtaining reliable conclusions about overheating faults.

[0003] Chinese Patent Publication No. CN118198915A discloses a substation inspection method, system, and equipment based on UAV and robot collaboration. The method is used in a substation inspection system, which includes a UAV, a robot, and a backend server. The method includes: the backend server acquiring image data of the substation to be inspected, with the image data collected by the UAV while running in the high-altitude area of ​​the substation and by the robot while running in the ground area of ​​the substation; the image data is processed for recognition to determine the inspection result of the substation, which indicates the equipment and environmental conditions of the substation. It can be seen that the above technical solution has the following problems: it fails to recognize the false thermal images generated by the specular reflection of polished metal surfaces, and is prone to misjudging the reflected false hot spots as real faults, affecting the accuracy of substation overheating fault detection. Summary of the Invention

[0004] To address this issue, the present invention provides an integrated air-ground substation overheating fault detection system based on the YOLO algorithm, which overcomes the problem in the prior art that it fails to identify false thermal images generated by the mirror reflection of polished metal surfaces, easily misjudging the reflected false hot spots as real faults, thus affecting the accuracy of substation overheating fault detection.

[0005] To achieve the above objectives, this invention provides an integrated air-ground substation overheating fault detection system based on the YOLO algorithm, comprising: The air-ground collaborative acquisition module is used to acquire the airborne infrared images, ground infrared images, and lidar point clouds of each substation unit, as well as record the observation position and observation tilt angle when acquiring each image. The target recognition module is used to determine the target box of the substation unit and the target box of the abnormal heating area based on the machine-scanned infrared image and the ground-scanned infrared image, and to determine the substation unit corresponding to each abnormal heating area based on the overlap between the target box of the substation unit and the target box of the abnormal heating area. The pseudo thermal image recognition module is used to determine whether to perform material correction or issue an alarm for equipment overheating fault based on the temperature uniformity of the substation unit. The material correction module is used to perform observation correction or adjacent compensation correction based on the thermal correlation. The misjudgment suppression module is used to determine whether to correct the infrared image based on the UAV rotor speed based on the misjudgment rate; The fault alarm module is used to issue alarm information to the substation unit corresponding to the abnormal heating area.

[0006] Furthermore, the pseudo-thermal image recognition module includes, The uniformity measurement unit is used to determine the temperature uniformity based on the temperature of each pixel within the target box of the transformer unit. The temperature uniformity measures the consistency of the temperature distribution on the surface of the transformer unit.

[0007] Furthermore, the pseudo-thermal image recognition module also includes, The strong interference discrimination unit is used to determine whether to perform observation correction or adjacent compensation correction based on the thermal correlation degree when the temperature uniformity is less than the preset uniformity threshold.

[0008] Furthermore, the material correction module includes, The thermal correlation quantification unit is used to determine the thermal correlation degree based on the temperature sequence of the suspected interference region and the temperature sequence of each interferometric substation.

[0009] Furthermore, the material correction module also includes, The material attribution unit is used to observe and correct suspected interference areas when the thermal correlation is less than a preset thermal correlation threshold; it is also used to perform adjacent compensation correction when the thermal correlation is greater than or equal to the preset thermal correlation threshold.

[0010] Furthermore, the material correction module also includes, The curve correction unit is used to generate a set of mapping curves for a single suspected interference area to perform observational correction on the suspected interference area.

[0011] Furthermore, the material correction module also includes, Adjacent refraction correction unit, used to perform adjacent compensation correction, includes, Identify and attribute substation units; The average temperature of each pixel in the suspected interference area is corrected based on the attribution transformer unit. The increase in temperature of each pixel within the suspected interference area is positively correlated with the average temperature of the attribution transformer unit.

[0012] Furthermore, the misjudgment suppression module includes, The misjudgment rate quantification unit is used to quantify the misjudgment rate by the number of overheating fault judgment conclusions rejected by the vehicle-mounted infrared acquisition unit based on close-range ground inspection and retesting, and the total number of overheating fault judgment conclusions. When the misjudgment rate is greater than the preset misjudgment threshold, the misjudgment concentration discrimination is performed.

[0013] Furthermore, the misjudgment suppression module also includes, The misjudgment concentration discrimination unit is used to determine the misjudgment concentration based on the acquisition time interval and the misjudgment rate, and to perform infrared image correction when the misjudgment concentration is greater than the preset misjudgment concentration threshold.

[0014] Furthermore, the misjudgment suppression module also includes, Infrared image correction unit, which is used to correct the airborne infrared image based on the rotor speed of the UAV.

[0015] Compared with existing technologies, the beneficial effect of this invention is that the localized high temperatures exhibited by the substation unit in infrared thermal imaging are not necessarily due to overheating faults in the equipment itself. They could be pseudo-thermal images formed by the reflection of thermal radiation from adjacent heat-generating devices onto the measured surface, or apparent temperature non-uniformity caused by differences in the emissivity of the measured surface material itself. The infrared thermal imager detector outputs pixel grayscale values ​​rather than temperatures. The grayscale is first converted to apparent temperature based on the grayscale-temperature mapping curve to characterize the temperature distribution. Temperature uniformity quantifies the reliability of the detection; the greater the uniformity, the more consistent the surface temperature distribution of a single substation unit, and the higher the reliability of the detection. Conversely, the lower the uniformity, the more significant the abnormal detection caused by external interference. Therefore, a uniformity quantification unit is set up to first quantify the temperature uniformity. When a strong interference discrimination unit determines that there is strong interference, it identifies the presence of localized high temperatures requiring further identification and initiates the material correction process. Using temperature uniformity as the first screening criterion, the surface temperature uniformity is separated into two categories: those with self-heating and those with local anomalies, which require further differentiation of causes. At the very beginning of anomaly detection, weak interference situations that do not need further investigation are filtered out. This not only avoids direct false alarms from false thermal images, but also narrows down the scope of processing for subsequent material correction and misjudgment suppression, thereby improving the accuracy of substation overheating fault detection.

[0016] Furthermore, upon entering a strong interference state, a heat-related correlation quantification unit is set up to quantify the temporal correlation between the suspected interference area and adjacent heat-generating equipment using the heat-related correlation degree. Targeted corrections are performed based on the heat-related correlation degree. If the temperature of the two substation units rises and falls together significantly over time, it is determined that the suspected interference area is caused by the same source of thermal radiation from adjacent equipment via reflection. Temperature unevenness caused by differences in the emissivity of the materials themselves has no temporal correlation with adjacent equipment. When the material attribution unit determines that the interference is caused by its own material, the curve correction unit generates a set of gray-scale-temperature mapping curves from multiple observation angles for observation correction; when it determines that the interference is caused by adjacent heat-generating equipment, the adjacent refraction correction unit performs adjacent compensation correction. By differentiating between two causes—the material's own emissivity difference and pseudo-thermal images reflected from adjacent equipment—through the correlation of heating, observational correction and adjacent compensation correction are applied respectively. The true heat source is restored from two dimensions: temperature sequence correlation and spatial radiation transfer. This not only eliminates misjudgments caused by material apparent inhomogeneity but also removes pseudo-thermal images formed by adjacent thermal radiation reflection. The corrected judgment can truly reflect the overheating state of the tested substation unit itself, rather than being distorted by surface reflection and environmental emissivity, thus further improving the accuracy of substation overheating fault detection.

[0017] Furthermore, even after material correction, the infrared images from drone inspections may still show a lower apparent temperature than the actual equipment temperature due to forced convection cooling of the measured surface caused by the downwash airflow generated by the high-speed rotation of the drone's rotor during hovering. This is a temperature measurement deviation unique to drone inspections and unrelated to the equipment itself. The intensity of downwash cooling increases with rotor speed, and the degree to which the apparent temperature is lowered is positively correlated with the rotor speed. To distinguish between systemic downwash deviations in drone inspections and remaining material-related misjudgments, the misjudgment rate is first quantified by the proportion of drone inspection overheating fault judgments rejected by the vehicle-mounted infrared acquisition unit during close-range ground inspections. This quantifies the misjudgment concentration, which is used to determine whether the misjudgments are concentrated in time and caused by rotor downwash disturbances, triggering the infrared image correction unit; or whether the misjudgments are dispersed in time and caused by material issues, sending the corresponding abnormal heating areas to the material correction module for recalibration. This significantly improves the accuracy of drone inspection overheating fault detection and the reliability of alarm conclusions, further enhancing the accuracy of substation overheating fault detection. Attached Figure Description

[0018] Figure 1 This is a block diagram of an air-ground integrated substation overheating fault detection system based on the YOLO algorithm, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating the steps of an air-ground integrated substation overheating fault detection system based on the YOLO algorithm, as described in an embodiment of the present invention. Figure 3This is a logic diagram of the strong interference discrimination unit in this embodiment of the invention, which determines whether to perform observation correction or adjacent compensation correction based on temperature uniformity. Figure 4 This is a logic diagram for the material attribution unit in an embodiment of the present invention to determine whether to perform observation correction or adjacent compensation correction based on the thermal correlation. Detailed Implementation

[0019] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0020] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0021] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0022] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0023] Please see Figure 1 As shown, it is a block diagram of the air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to an embodiment of the present invention, including, The air-ground collaborative acquisition module is used to acquire the airborne infrared images, ground infrared images, and lidar point clouds of each substation unit, as well as record the observation position and observation tilt angle when acquiring each image. The target recognition module is connected to the air-ground collaborative acquisition module. It is used to determine the target frame of the substation unit and the target frame of the abnormal heating area based on the air-to-ground infrared image and the ground-to-ground infrared image, and to determine the substation unit corresponding to each abnormal heating area based on the overlap between the target frame of the substation unit and the target frame of the abnormal heating area. A pseudo thermal image recognition module, which is connected to the target recognition module, is used to determine whether to perform material correction or issue an alarm for equipment overheating fault based on the temperature uniformity of the substation unit. The material correction module, which is connected to the pseudo thermal image recognition module, is used to perform observation correction or adjacent compensation correction based on the thermal correlation degree. The misjudgment suppression module is connected to the material correction module and the target recognition module respectively, and is used to determine whether to correct the infrared image based on the rotor speed of the UAV based on the misjudgment rate. The fault alarm module is connected to the pseudo thermal image recognition module, the material correction module and the misjudgment suppression module respectively, and is used to issue alarm information to the substation unit corresponding to the abnormal heating area.

[0024] Please see Figure 2 The diagram shows a flowchart of the steps in an air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to an embodiment of the present invention, including: S1, acquire the machine-scan infrared image, ground-scan infrared image and lidar point cloud of each substation unit, and record the observation position and observation tilt angle when acquiring each image; S2, determine the substation unit corresponding to each abnormal heating area based on infrared images; S3, determine the temperature uniformity based on the temperature of each pixel within the target frame of the substation unit; S4, determine whether to perform material correction or issue an alarm for equipment overheating fault based on temperature uniformity; S5, based on the correlation of heating, observation correction or adjacent compensation correction is performed; S6 determines whether to correct the infrared image based on the drone rotor speed based on the false positive rate; S7 issues alarm information to the substation unit corresponding to the abnormal heating area.

[0025] Specifically, the air-to-ground collaborative data acquisition module includes, An airborne infrared acquisition unit, deployed on one side of the UAV, includes an infrared imager and a visible light camera for acquiring airborne infrared images of the substation unit; the substation unit can be a single substation device. An airborne lidar unit, which is deployed on one side of the UAV, includes a lidar sensor for acquiring laser point cloud images of the substation. The vehicle-mounted infrared acquisition unit, which is located on one side of the patrol vehicle, includes an infrared imager and a visible light camera for acquiring ground-based infrared images of the substation unit. The pose recording unit is used to record the observation position and observation tilt angle when acquiring the airborne infrared image and the ground-based infrared image. The observation tilt angle is the angle between the optical axis of the infrared thermal imager and the normal of the measured surface.

[0026] Specifically, the target recognition module includes, It has a built-in lightweight YOLO model, which is used to determine the target bounding box of the substation unit and the target bounding box of the abnormal heating area based on the machine-scanned infrared image and the ground-scanned infrared image. In a single embodiment, the lightweight YOLO model incorporates a SIMAM module as a spatial attention module into the backbone network, enabling the model to focus more on temperature anomaly regions in the infrared images and suppress background interference. A bidirectional feature pyramid network is used in the feature fusion network to improve the detection capability of abnormal heating regions of varying scales. A lightweight design employs depthwise separable convolution combined with model pruning and quantization to adapt to the computational limitations of edge computing units. The SIoU loss function is used to optimize the accuracy of bounding box regression. The training dataset consists of infrared thermal images of substation units under different operating conditions, and is expanded through random flipping, random rotation, color adjustment, and noise addition.

[0027] Specifically, the pseudo-thermal image recognition module includes, The uniformity quantification unit is used to determine the temperature uniformity based on the temperature of each pixel within the target box of the substation unit. Temperature uniformity quantifies the detection reliability of the substation unit. The greater the temperature uniformity, the higher the detection reliability for a single substation unit. The lower the temperature uniformity, the more significant the abnormal detection caused by external interference to the substation unit.

[0028] The uniformity measurement unit converts the grayscale values ​​of each pixel within the target box of the substation into apparent temperature based on the grayscale-temperature mapping curve, and then calculates the temperature uniformity of the substation.

[0029] In a single embodiment, the uniformity quantization unit used to determine the gray-temperature mapping curve includes aligning an infrared thermal imager with a standard radiation reference source of known temperature, acquiring infrared images at several known temperature points, recording the correspondence between the pixel gray values ​​output by the detector and the corresponding known temperatures, and establishing a functional relationship between the pixel gray values ​​and apparent temperature for the material type point by point to determine the gray-temperature mapping curve.

[0030] In a single embodiment, the uniformity measurement unit used to determine temperature uniformity includes: determining the temperature corresponding to each pixel within the target box of the substation unit based on the gray-temperature mapping curve; calculating the standard deviation of each temperature; calculating the temperature difference between the maximum value of each temperature and the ambient temperature; calculating the temperature ratio between the temperature standard deviation and the temperature difference; and solving for the difference between the temperature ratio to obtain the temperature uniformity.

[0031] Please see Figure 3As shown, this is a logic diagram of the strong interference discrimination unit in this embodiment of the invention, which determines whether to perform observation correction or adjacent compensation correction based on temperature uniformity. The pseudo thermal image recognition module also includes... The strong interference discrimination unit is used to determine that a single substation unit is in a strong interference state when the temperature uniformity is less than a preset uniformity threshold. It then performs observational correction or adjacent compensation correction based on the heating correlation. Localized high-temperature areas appear on the surface of the substation unit. However, this localized high temperature may originate from heat radiation from adjacent heating devices reflected by the surface rather than from a self-heating fault, or it may be due to uneven apparent temperature caused by differences in the emissivity of the substation unit's own material. Therefore, the heating correlation is further used to distinguish between these two causes and correct them accordingly.

[0032] The strong interference discrimination unit is also used to determine that a single substation unit is in a weak interference state when the temperature uniformity is greater than or equal to a preset uniformity threshold. The surface temperature of the substation unit is uniform and there are no local high temperature areas. The control fault alarm module issues alarm information for the abnormal heating area.

[0033] In a single embodiment, the strong interference discrimination unit is used to determine a preset uniformity threshold. The preset uniformity threshold is used to determine whether the surface temperature of the substation unit is in a uniform state. The unit retrieves healthy sample infrared images of each substation unit under the same operating conditions and confirmed by operation and maintenance to have no overheating faults from the historical inspection database to obtain a sample set of temperature uniformity of each healthy sample infrared image. The minimum temperature uniformity value in the sample set is used as the preset uniformity threshold.

[0034] In a single embodiment, the localized high temperatures observed in infrared thermal imaging of a substation unit may not necessarily originate from an overheating fault in the equipment itself. They could be pseudo-thermal images formed by the reflection of thermal radiation from adjacent heat-generating devices onto the measured surface, or apparent temperature non-uniformity caused by differences in the emissivity of the measured surface material. The infrared thermal imager detector outputs pixel grayscale values ​​rather than temperatures. The grayscale is first converted to apparent temperature based on a grayscale-temperature mapping curve to characterize the temperature distribution. Temperature uniformity quantifies the reliability of the detection; higher uniformity indicates a more consistent temperature distribution on the surface of a single substation unit, resulting in higher detection reliability. Lower uniformity indicates more significant abnormal detection due to external interference. Therefore, a uniformity quantification unit is set up to first quantify the temperature uniformity. When a strong interference discrimination unit determines that there is strong interference, it identifies the presence of localized high temperatures requiring further identification and initiates the material correction process. Using temperature uniformity as the first screening criterion, the surface temperature uniformity is separated into two categories: those with self-heating and those with local anomalies, which require further differentiation of causes. At the very beginning of anomaly detection, weak interference situations that do not need further investigation are filtered out. This not only avoids direct false alarms from false thermal images, but also narrows down the scope of processing for subsequent material correction and misjudgment suppression, thereby improving the accuracy of substation overheating fault detection.

[0035] Specifically, the material correction module includes, The thermal correlation quantification unit is used to determine the thermal correlation degree based on the temperature sequence of the suspected interference area and the temperature sequences of each interfering substation. The thermal correlation degree is used to determine the degree of correlation between the temperature time series of the suspected interference area and the temperature time series of adjacent heat-generating equipment. The higher the thermal correlation degree, the more significant the co-occurrence of temperature increases and decreases between the two substations over time. In a single embodiment, the heat correlation quantification unit is used to determine the heat correlation degree, obtain the average temperature of the suspected interference area at several historical detection time points to obtain the receiving point temperature sequence, obtain the average temperature of the interferometric substation at several historical detection time points for a single interferometric substation to obtain the interference temperature sequence, calculate the absolute value of the covariance between the receiving point temperature sequence and the interference temperature sequence, calculate the product of the standard deviation of the receiving point temperature sequence and the standard deviation of the interference temperature sequence, and solve the ratio of the absolute value of the covariance to the product of the standard deviations to obtain the heat correlation degree.

[0036] The suspected interference area refers to a locally high-temperature connected region within the target frame of the substation unit, surrounded by closed isotherms. The geometric center of this region is the center point of the suspected interference area. The temperature within the locally high-temperature connected region is greater than the preset abnormal temperature threshold.

[0037] Specifically, the method for determining the preset abnormal temperature threshold is to retrieve the infrared images of healthy samples that have the same equipment type as each substation unit and are in the same operating condition range from the historical inspection data, and which have been confirmed by operation and maintenance to have no overheating faults. This will obtain the surface temperature set of each healthy sample infrared image within the target frame of the substation unit, count the peak temperature in the surface temperature set of each healthy sample, obtain the peak temperature sample set corresponding to each healthy sample, and take the maximum value in the peak temperature sample set as the preset abnormal temperature threshold.

[0038] The interference substation unit is each substation unit within a coverage area centered on the center point of the suspected interference area and with a preset adjacent length as the radius.

[0039] In a single embodiment, the heat-related quantification unit is used to determine the adjacent length by statistically analyzing samples from historical inspection data where false thermal images occur due to reflection, and taking the maximum actual distance between the reflection source device and the target device in these samples as the adjacent length.

[0040] Please see Figure 4 As shown, this is a logic diagram of the material attribution unit in an embodiment of the present invention determining whether to perform observation correction or adjacent compensation correction based on the thermal correlation. The material correction module further includes: The material attribution unit is used to determine that the temperature unevenness corresponding to the abnormal heating area is caused by the surface material of the transformer unit under test when the heating correlation degree is less than the preset heating correlation degree threshold, and to observe and correct the suspected interference area; it is also used to determine that the temperature unevenness corresponding to the abnormal heating area is caused by the adjacent heating equipment when the heating correlation degree is greater than or equal to the preset heating correlation degree threshold, and to perform adjacent compensation correction.

[0041] In a single embodiment, the material attribution unit is used to determine a preset heat-related degree threshold. It retrieves infrared images of healthy samples from historical inspection data, which have been confirmed by operation and maintenance to have no overheating faults and whose surface temperature non-uniformity is caused by the difference in directional emissivity of different surface areas of the tested substation unit. For each healthy sample, it outputs the target box of its abnormal heat-generating area and the target box of adjacent heat-generating equipment to determine the heat-related degree set. The maximum value of the heat-related degree of the healthy samples in the heat-related degree set is used as the preset heat-related degree threshold.

[0042] Specifically, the material correction module also includes, The curve correction unit is used to generate a set of mapping curves for a single suspected interference area to perform observational correction on the suspected interference area.

[0043] In a single embodiment, the curve correction unit is used to generate a set of mapping curves for a single suspected interference region, acquire infrared images of the suspected interference region at several observation angles, determine the grayscale value of each pixel and the corresponding known actual temperature, and obtain grayscale-temperature mapping curves for the suspected interference region at each observation angle. The set of mapping curves for a single suspected interference region includes grayscale-temperature mapping curves at several observation angles.

[0044] Specifically, the material correction module also includes, Adjacent refraction correction unit, used to perform adjacent compensation correction, includes, Identify and attribute substation units; The average temperature of each pixel in the suspected interference area is corrected based on the attribution transformer unit. The increase in temperature of each pixel within the suspected interference area is positively correlated with the average temperature of the attribution transformer unit.

[0045] The greater the increase in temperature of each pixel within the suspected interference area, the more dominant the contribution of reflected radiation is in the apparent temperature. The high-temperature area is more likely to be formed by the reflected thermal radiation from adjacent attribution devices rather than by the heating of the surface under test itself. The greater the average temperature of the attribution transformer unit, the greater the energy radiated outward with its absolute temperature, and the stronger the thermal radiation reflected from the surface under test into the infrared thermal imager.

[0046] In a single embodiment, the identification attribution transformer unit is used to correct the temperature of each pixel in the suspected interference area based on the average temperature of the attribution transformer unit. The corrected temperature of each pixel in the suspected interference area is determined based on the average temperature of the attribution transformer unit and the adjacent interference coefficient. The corrected temperature is equal to the original apparent temperature of the pixel minus the product of the average temperature of the attribution transformer unit and the adjacent interference coefficient.

[0047] The adjacent interference coefficient is determined by obtaining the surface reflectivity of the substation based on its material type, determining the spatial distance from the center of the suspected interference area to the center of the attribution substation and the incident angle between the radiation incident direction and the normal of the surface under test based on the three-dimensional surface model constructed by the lidar point cloud, obtaining the geometric attenuation of radiation propagation from the spatial distance, obtaining the coupling strength of the reflection direction from the incident angle, and finally multiplying the reflectivity, geometric attenuation factor and the coupling strength of the reflection direction to obtain the adjacent interference coefficient. The larger the coefficient, the greater the intensity of the thermal radiation of the attribution substation coupled into the suspected interference area after being reflected by the surface under test.

[0048] The attribution substation unit is the substation unit with the highest correlation to the heating of the substation unit under test within the suspected interference area.

[0049] In a single embodiment, after entering a strong interference state, a heat-related correlation quantification unit is set up to quantify the temporal correlation between the suspected interference area and adjacent heat-generating devices using heat-related correlation degree. Targeted correction is performed based on the heat-related correlation degree. If the temperature of the two substation units rises and falls together significantly over time, it is determined that the suspected interference area is caused by the same source of thermal radiation from adjacent devices via reflection. Temperature unevenness caused by differences in the emissivity of the materials themselves has no temporal correlation with adjacent devices. When the material attribution unit determines that the interference is caused by its own material, the curve correction unit generates a set of gray-scale-temperature mapping curves from multiple observation angles for observation correction; when it determines that the interference is caused by adjacent heat-generating devices, the adjacent refraction correction unit performs adjacent compensation correction. By differentiating between two causes—the material's own emissivity difference and pseudo-thermal images reflected from adjacent equipment—through the correlation of heating, observational correction and adjacent compensation correction are applied respectively. The true heat source is restored from two dimensions: temperature sequence correlation and spatial radiation transfer. This not only eliminates misjudgments caused by material apparent inhomogeneity but also removes pseudo-thermal images formed by adjacent thermal radiation reflection. The corrected judgment can truly reflect the overheating state of the tested substation unit itself, rather than being distorted by surface reflection and environmental emissivity, thus further improving the accuracy of substation overheating fault detection.

[0050] Specifically, the misjudgment suppression module includes, The false alarm rate quantification unit is used to determine the false alarm rate based on the abnormal temperature of ground patrol and the abnormal temperature of air patrol. In a single embodiment, the false positive rate quantization unit is used to determine the false positive rate, including: By identifying abnormal heating areas through infrared imaging, the abnormal heating areas can be obtained and the average temperature of a single abnormal heating area can be determined. Acquire ground-based infrared images to identify each mapped abnormal heating area corresponding to the machine-surveyed abnormal heating area in each ground-based infrared image; determine the mapped average temperature of a single mapped abnormal heating area. For the average temperature of the machine patrol and the corresponding mapped abnormal heating area, determine the absolute value of the difference between the average temperature of the machine patrol and the average temperature of the mapped area, and obtain the temperature deviation of a single set of mapped images. A single set of mapping images with a temperature deviation greater than the preset temperature deviation is identified as an abnormal mapping set. The false positive rate is obtained by calculating the ratio of the number of abnormal mapping groups to the total number of mapping images in each group.

[0051] The preset temperature deviation can be determined by retrieving several matching sample pairs from the historical inspection database of the machine inspection infrared images and ground inspection infrared images of the same substation unit, which are confirmed by operation and maintenance to point to the same real overheating fault. The average temperature of the abnormal heating area of ​​the machine inspection in each matching sample pair and the average temperature of the corresponding mapped abnormal heating area are obtained. The absolute value of the difference between the average temperature of the machine inspection and the average temperature of the mapped area of ​​each matching sample pair is calculated to obtain the temperature deviation sample of each matching sample pair, thereby forming a temperature deviation sample set.

[0052] The maximum value in the temperature deviation sample set is used as the preset temperature deviation.

[0053] The misjudgment rate quantification unit is used to perform misjudgment concentration judgment when the misjudgment rate is greater than a preset misjudgment threshold. A misjudgment rate greater than the preset misjudgment threshold indicates a large number of misjudgments, and it is determined that there may be batch misjudgments in this inspection, further analyzing the degree of concentration of misjudgments in the time dimension.

[0054] If the false positive rate is less than or equal to the preset false positive threshold, it indicates that the reliability of the current overheating fault judgment conclusion meets the requirements, and the current parameters are maintained for continued monitoring. In this case, the number of false positives is small, and the overall judgment conclusion of this inspection is considered reliable, with the false positives being of an occasional nature.

[0055] In a single embodiment, the misjudgment rate quantification unit is used to determine a preset misjudgment threshold. It retrieves healthy inspection samples from the historical inspection database that have been confirmed by operation and maintenance and whose machine inspection overheating fault judgment is consistent with the ground inspection retest conclusion. The samples are grouped according to the same operating condition interval. The misjudgment rate of each group in one inspection cycle is calculated. After obtaining the sample set of the misjudgment rate of the healthy samples in the group, the maximum value of the misjudgment rate of the healthy samples in the set is directly used as the preset misjudgment threshold of the group.

[0056] Specifically, the misjudgment suppression module also includes, The misjudgment concentration discrimination unit is used to determine the misjudgment concentration based on the misjudgment rate corresponding to each collection time interval. When the misjudgment concentration is greater than the preset misjudgment concentration threshold, it is determined that the misjudgment is concentrated in time and is caused by the drone rotor downwash disturbance, and infrared image correction is performed.

[0057] If the concentration of misjudgments is less than or equal to the preset concentration threshold for misjudgments, it is determined that the misjudgments are dispersed in time and are caused by the material problem of the transformer unit itself. The suspected interference area is then observed and corrected. In a single embodiment, the misjudgment concentration discrimination unit is used to determine the misjudgment concentration, obtain the misjudgment rate in each collection time interval, take the maximum value of the misjudgment rate in each time interval, and divide the maximum value by the average value of the misjudgment rate in each time interval to obtain the misjudgment concentration.

[0058] In a single embodiment, the misjudgment concentration discrimination unit is further used to determine a preset misjudgment concentration threshold. It retrieves healthy inspection samples from the historical inspection database that have been confirmed by operation and maintenance and whose machine inspection overheating fault judgment is consistent with the ground inspection retest conclusion. For healthy samples that have been rejected by the ground inspection retest at least once, they are grouped according to the same operating condition interval, and the misjudgment concentration of each sample is calculated. After obtaining the misjudgment concentration sample set of the healthy samples in the group, the maximum value of the misjudgment concentration of the healthy samples in the set is directly used as the preset misjudgment concentration threshold of the group.

[0059] Specifically, the misjudgment suppression module also includes, The infrared image correction unit is used to correct the drone's infrared image based on the drone's rotor speed. The temperature reduction of each pixel in the infrared image is positively correlated with the drone's rotor speed.

[0060] The rotor speed of a drone is used to determine the degree of cooling and pressure reduction of the apparent temperature of the measured surface caused by the downwash airflow of the drone rotor. The higher the rotor speed, the stronger the downwash convection cooling, the lower the apparent temperature is compressed, and the higher the drone rotor speed.

[0061] In a single embodiment, the infrared image correction unit is used to correct the airborne infrared image based on the UAV rotor speed. A fixed standard radiation reference source with a known true temperature is selected, and the UAV is controlled to hover at several different rotor speeds. The apparent temperature of the reference source after being subjected to the downwash airflow is measured by an airborne infrared imager. The true temperature of the reference source is subtracted from the apparent temperature at each speed to obtain the downwash cooling temperature drop corresponding to each speed. A correspondence between the UAV rotor speed and the UAV rotor speed is established with rotor speed as the independent variable and downwash cooling temperature drop as the dependent variable. During inspection, the actual rotor speed of the UAV while hovering is read, and the corresponding downwash cooling temperature drop compensation value is determined by the UAV rotor speed correspondence. The apparent temperature of each pixel in the airborne infrared image is added to the compensation value to obtain the corrected apparent temperature.

[0062] In a single embodiment, even after material correction, the infrared image from the drone inspection may still show a lower apparent temperature than the actual equipment temperature due to forced convection cooling of the measured surface caused by the downwash airflow generated by the high-speed rotation of the drone's rotor during hovering. This is a temperature measurement deviation unique to drone inspection and unrelated to the equipment itself. The intensity of downwash cooling increases with rotor speed, and the degree to which the apparent temperature is lowered is positively correlated with the rotor speed. To distinguish between systemic downwash deviations in drone inspection and remaining material-related misjudgments, the misjudgment rate is first quantified by the misjudgment rate quantification unit, which measures the proportion of drone inspection overheating fault judgments rejected by the vehicle-mounted infrared acquisition unit during close-range ground inspection retests to the total number of drone inspection judgments. This further yields the misjudgment concentration, used to determine whether the misjudgments are concentrated in time and caused by rotor downwash disturbances, triggering the infrared image correction unit; or whether the misjudgments are dispersed in time and caused by material issues, sending the corresponding abnormal heating areas to the material correction module for recalibration. This significantly improves the accuracy of drone inspection overheating fault detection and the reliability of alarm conclusions, further enhancing the accuracy of substation overheating fault detection.

[0063] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A YOLO-based integrated air-ground substation overheating fault detection system, characterized in that, include: The air-ground collaborative acquisition module is used to acquire the airborne infrared images, ground infrared images, and lidar point clouds of each substation unit, as well as record the observation position and observation tilt angle when acquiring each image. The target recognition module is used to determine the target box of the substation unit and the target box of the abnormal heating area based on the machine-scanned infrared image and the ground-scanned infrared image, and to determine the substation unit corresponding to each abnormal heating area based on the overlap between the target box of the substation unit and the target box of the abnormal heating area. The pseudo thermal image recognition module is used to determine whether to perform material correction or issue an alarm for equipment overheating fault based on the temperature uniformity of the substation unit. The material correction module is used to perform observation correction or adjacent compensation correction based on the thermal correlation. The misjudgment suppression module is used to determine whether to correct the infrared image based on the UAV rotor speed based on the misjudgment rate; The fault alarm module is used to issue alarm information to the substation unit corresponding to the abnormal heating area.

2. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 1, characterized in that, The pseudo-thermal image recognition module includes, The uniformity measurement unit is used to determine the temperature uniformity based on the temperature of each pixel within the target box of the transformer unit.

3. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 2, characterized in that, The pseudo-thermal image recognition module also includes, The strong interference discrimination unit is used to determine whether to perform observation correction or adjacent compensation correction based on the thermal correlation degree when the temperature uniformity is less than the preset uniformity threshold.

4. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 3, characterized in that, The material correction module includes, The thermal correlation quantification unit is used to determine the thermal correlation degree based on the temperature sequence of the suspected interference region and the temperature sequence of each interferometric substation.

5. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 4, characterized in that, The material correction module also includes, The material attribution unit is used to observe and correct suspected interference areas when the thermal correlation is less than a preset thermal correlation threshold; it is also used to perform adjacent compensation correction when the thermal correlation is greater than or equal to the preset thermal correlation threshold.

6. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 5, characterized in that, The material correction module also includes, The curve correction unit is used to generate a set of mapping curves for a single suspected interference area to perform observational correction on the suspected interference area.

7. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 6, characterized in that, The material correction module also includes, Adjacent refraction correction unit, used to perform adjacent compensation correction, includes, Identify and attribute substation units; The average temperature of each pixel in the suspected interference area is corrected based on the attribution transformer unit. The increase in temperature of each pixel within the suspected interference area is positively correlated with the average temperature of the attribution transformer unit.

8. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 7, characterized in that, The misjudgment suppression module includes, The misjudgment rate quantification unit is used to determine the misjudgment rate based on the abnormal temperature of ground patrol and the abnormal temperature of air patrol, and to perform misjudgment concentration discrimination when the misjudgment rate is greater than the preset misjudgment threshold.

9. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 8, characterized in that, The misjudgment suppression module also includes, The misjudgment concentration discrimination unit is used to determine the misjudgment concentration based on the misjudgment rate in each acquisition time interval, and to perform machine-scan infrared image correction when the misjudgment concentration is greater than the preset misjudgment concentration threshold.

10. The air-ground integrated substation overheating fault detection system based on the YOLO algorithm according to claim 9, characterized in that, The misjudgment suppression module also includes, Infrared image correction unit, which is used to correct the airborne infrared image based on the rotor speed of the UAV.

Images