Current transformer airtightness detection method, device, equipment and system

Abstract

The application discloses a current transformer airtightness detection method, device, equipment and system, relates to the technical field of power equipment detection, and comprises the following steps: generating an airtightness detection path based on a three-dimensional reference profile model of a current transformer; collecting sound pressure distribution data, temperature field data and geometric deformation data of each detection area on the surface of the current transformer along the airtightness detection path; calculating the ultrasonic energy attenuation rate, temperature gradient standard deviation and curvature change rate of each detection area based on the sound pressure distribution data, temperature field data and geometric deformation data; performing space-time fusion on the ultrasonic energy attenuation rate, temperature gradient standard deviation and curvature change rate of the same detection area by using a transfer learning model, and calculating the leakage probability of each detection area; and if it is determined that the current transformer has leakage in a target detection area, determining the leakage coordinates and leakage rate of the target detection area on the current transformer. The application can reduce the false positive rate of current transformer airtightness detection and shorten the detection time.

Description

Technical Field

[0001] This application relates to the field of power equipment testing technology, and in particular to a method, apparatus, equipment and system for testing the airtightness of a current transformer. Background Technology

[0002] As a core device for current measurement, relay protection, and energy metering in power systems, the sealing performance of current transformers directly determines their operational safety and service life. During long-term operation of power systems, current transformers are often exposed to complex outdoor environments, enduring multiple stresses such as temperature and humidity fluctuations, electromagnetic interference, mechanical vibration, and chemical corrosion. If the seal fails, external moisture, dust, or corrosive media can easily penetrate the equipment, causing surface discharge of the epoxy resin casting, core corrosion, or aging of the winding insulation. Industry statistics show that faults caused by poor sealing account for 37% of all current transformer failures. For gas-insulated current transformers, when the annual SF6 gas leakage rate exceeds the 0.5% specified in the IEEE C57.13-2016 standard, a dangerous electric arc may ignite due to decreased insulation strength, causing equipment shutdown or even grid accidents. Furthermore, micro-leakage at the flange sealing surface accelerates bolt stress corrosion, further increasing the risk of structural fracture and posing a serious threat to the stable operation of the power system.

[0003] Currently, the industry mainly adopts two technical solutions for the airtightness testing of current transformers: offline testing and online testing. However, both have obvious limitations: First, the early diagnosis capability of micro-leakage is insufficient, and existing solutions cannot effectively detect early sealing hazards such as epoxy resin micro-cracks <0.1mm. Second, the applicable scenarios are limited. Offline solutions require equipment shutdown and some involve destructive operations, while online solutions cannot cover medium and low voltage transformers below 35kV, failing to meet the requirements of uninterrupted power supply maintenance. Third, the detection accuracy and efficiency are low. Multi-physical field data have not been effectively integrated, making them susceptible to electromagnetic interference, resulting in a high false alarm rate. Furthermore, the detection time is long, making it difficult to match the requirements of intelligent operation and maintenance of power systems for correlation analysis between leakage rate and insulation life. Summary of the Invention

[0004] In view of this, this application provides a method, apparatus, equipment and system for airtightness testing of current transformers, which can reduce the false alarm rate of airtightness testing of current transformers, shorten the testing time, and comprehensively solve the limitations of existing testing technologies.

[0005] According to a first aspect of this application, a method for detecting the airtightness of a current transformer is provided, comprising:

[0006] A three-dimensional reference profile model of the current transformer is constructed, and an airtightness detection path covering all parts of the current transformer that are prone to sealing failure is generated based on the three-dimensional reference profile model.

[0007] Along the airtightness detection path, a focused beam and light pulse are emitted toward the current transformer to simultaneously collect sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and to obtain geometric deformation data of each detection area.

[0008] The ultrasonic energy attenuation rate of each detection area is calculated based on the sound pressure distribution data; the temperature gradient standard deviation of each detection area is calculated based on the temperature field data; and the curvature change rate of each detection area is calculated based on the geometric deformation data.

[0009] The leakage probability of each detection area is calculated by using a transfer learning model to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, the temperature gradient standard deviation and the curvature change rate in the same detection area.

[0010] If a target detection area exists with a leakage probability greater than a preset threshold, it is determined that the current transformer has a leakage in the target detection area, and the leakage coordinates and leakage rate of the target detection area on the current transformer are determined.

[0011] According to a second aspect of this application, a current transformer airtightness detection device is provided, comprising:

[0012] The generation module is used to construct a three-dimensional reference contour model of the current transformer and generate an airtightness detection path covering all parts of the current transformer that are prone to sealing failure based on the three-dimensional reference contour model.

[0013] The acquisition module is used to emit a focused beam and light pulse to the current transformer along the airtightness detection path, and simultaneously acquire the sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and obtain the geometric deformation data of each detection area.

[0014] The calculation module is used to calculate the ultrasonic energy attenuation rate of each detection area based on the sound pressure distribution data, calculate the temperature gradient standard deviation of each detection area based on the temperature field data, and calculate the curvature change rate of each detection area based on the geometric deformation data.

[0015] The calculation module is also used to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, the temperature gradient standard deviation, and the curvature change rate in the same detection area using a transfer learning model, and to calculate the leakage probability of each detection area.

[0016] The determination module is used to determine that if there is a target detection area with a leakage probability greater than a preset threshold, the current transformer has a leakage in the target detection area, and to determine the leakage coordinates and leakage rate of the target detection area on the current transformer.

[0017] According to a third aspect of this application, a current transformer airtightness testing device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor. When the processor executes the computer program, it implements the aforementioned current transformer airtightness testing method.

[0018] According to a fourth aspect of this application, a current transformer airtightness detection system is provided, comprising:

[0019] The aforementioned current transformer airtightness testing equipment; and,

[0020] An AR display terminal is used to receive the leakage coordinates in real time and overlay them onto the physical image of the current transformer.

[0021] According to a fifth aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for detecting the airtightness of a current transformer.

[0022] By employing the above technical solutions, this application provides a method, apparatus, equipment, and system for detecting the airtightness of a current transformer. This involves constructing a three-dimensional reference profile model of the current transformer and planning a detection path covering all areas prone to sealing failure, ensuring no blind spots. A focused beam and light pulse are simultaneously emitted along the path, collaboratively acquiring multi-modal data on sound pressure, temperature, and geometric deformation. Wavelet packet decomposition, temperature gradient analysis, and point cloud comparison are used to calculate the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate, respectively. A transfer learning model is then used to perform spatiotemporal fusion of these three types of features to accurately calculate the leakage probability. Finally, the leakage coordinates are located, and the leakage rate is calculated. This solution achieves three core technical effects: First, by capturing micro-leakage characteristics through multi-modal data collaboration, it can effectively detect epoxy resin micro-cracks, breaking through the bottleneck of early diagnosis of micro-leakage; second, it does not require equipment shutdown or destructive operations, and is compatible with medium and low voltage transformers below 35kV, meeting the needs of uninterrupted power supply operation and maintenance scenarios; third, it utilizes a transfer learning model to achieve deep fusion of multi-physics field data, reducing the risk of false alarms caused by electromagnetic interference, while significantly shortening the detection time, and can also output key parameters such as leakage coordinates and rate, providing data support for establishing a model linking leakage rate and insulation life, matching the intelligent operation and maintenance needs of power systems.

[0023] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0024] Figure 1A schematic flowchart of a current transformer airtightness detection method provided in an embodiment of this application is shown;

[0025] Figure 2 A flowchart illustrating a current transformer airtightness detection method according to another embodiment of this application is shown;

[0026] Figure 3 A schematic diagram of the structure of a current transformer airtightness detection device provided in an embodiment of this application is shown. Detailed Implementation

[0027] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0028] Currently, the industry mainly adopts two technical solutions for the airtightness testing of current transformers: offline testing and online testing. However, both have obvious limitations: First, the early diagnosis capability of micro-leakage is insufficient, and existing solutions cannot effectively detect early sealing hazards such as epoxy resin micro-cracks <0.1mm. Second, the applicable scenarios are limited. Offline solutions require equipment shutdown and some involve destructive operations, while online solutions cannot cover medium and low voltage transformers below 35kV, failing to meet the requirements of uninterrupted power supply maintenance. Third, the detection accuracy and efficiency are low. Multi-physical field data have not been effectively integrated, making them susceptible to electromagnetic interference, resulting in a high false alarm rate. Furthermore, the detection time is long, making it difficult to match the requirements of intelligent operation and maintenance of power systems for correlation analysis between leakage rate and insulation life.

[0029] To address the aforementioned problems, embodiments of the present invention provide a method for detecting the airtightness of a current transformer, such as... Figure 1 As shown, the method includes:

[0030] Step 110: Construct a three-dimensional reference profile model of the current transformer, and generate an airtightness detection path covering all parts of the current transformer that are prone to sealing failure based on the three-dimensional reference profile model.

[0031] Current transformers are key devices in power systems used to proportionally convert high current to low current. Their sealing performance directly affects insulation reliability. Sealing failure can lead to moisture intrusion, gas leakage, and faults such as surface discharge and arcing. The three-dimensional reference profile model is a 1:1 matching 3D digital model of the current transformer, constructed by processing full-area point cloud data collected by a laser scanner. This model serves as a reference for subsequent detection path planning and geometric deformation comparison, accurately reflecting the surface structure of the transformer (such as flange connection surfaces and casting joints). The current transformer is prone to sealing failure. Areas with weak sealing structures and high incidence of leakage faults on the instrument mainly include flange connection surfaces (bolt seals are prone to stress corrosion failure), epoxy resin casting joints (material shrinkage easily produces micro-cracks), and valve interfaces (sealants are prone to aging). These areas account for more than 85% of instrument transformer leakage faults. The airtightness detection path is a detection equipment movement trajectory generated based on a three-dimensional reference contour model, with the principle of full coverage and key focus. It needs to cover all parts of the instrument transformer that are prone to sealing failure, and at the same time, plan repeated scanning paths for weak areas to ensure that there are no blind spots in the detection, and that the path coordinates correspond precisely to the physical spatial position of the instrument transformer.

[0032] In this embodiment of the present disclosure, when constructing a three-dimensional reference contour model of the current transformer, the transformer must first be scanned from all directions using a laser scanner to obtain the three-dimensional coordinates (i.e., point cloud data) of each point on its surface. Then, through point cloud registration, noise reduction, modeling, and other processing, a three-dimensional reference contour model that is completely consistent with the actual structure of the transformer is generated. The three-dimensional reference contour model can clearly present the spatial position and geometric shape of parts prone to sealing failure, such as flange connection surfaces and casting joints. When generating the airtightness detection path based on this, all parts prone to failure must first be marked in the three-dimensional reference contour model. Then, relevant algorithms for grid division and path optimization are used to divide the model surface into multiple uniform grids (ensuring that no micro-leakage areas are missed). With the goal of covering all grids with the shortest path, a spiral or serpentine path extending along the surface of the transformer is planned. Finally, the path in the three-dimensional reference contour model is converted into motion coordinates that can be executed by the detection equipment (such as an ultrasonic phased array probe or an infrared thermal imager), ensuring that the equipment can accurately cover all key areas during the detection process.

[0033] From the perspective of comprehensiveness, the three-dimensional reference profile model can accurately locate all areas prone to sealing failure, avoiding the problem of missed detections in weak areas due to reliance on human experience in traditional testing. Furthermore, the detection path generated based on the three-dimensional reference profile model can achieve full coverage of the transformer surface. In particular, repeated scanning of high-leakage areas such as flanges and casting joints can improve the redundancy of detection data in these areas, providing sufficient data support for subsequent micro-leakage identification. From the perspective of accuracy, the model, as a reference, ensures the precision of the detection path relative to the actual spatial position of the transformer. Matching avoids misalignment between the detection area and the physical part caused by the offset of the detection equipment, thus ensuring that the collected sound pressure, temperature and other data can truly reflect the sealing status of the corresponding part; in terms of scenario adaptability, this path planning method does not require disassembly or modification of the current transformer, and can directly plan the path based on the physical model. It can adapt to current transformers of different voltage levels from 10kV to 750kV (including medium and low voltage current transformers below 35kV), and can meet the detection needs in uninterrupted power supply operation and maintenance scenarios. It can solve the limitations of traditional offline detection that requires equipment shutdown and online detection with narrow coverage.

[0034] Step 120: Along the airtightness detection path, emit a focused beam and light pulse to the current transformer, and simultaneously collect the sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and obtain the geometric deformation data of each detection area.

[0035] The focused beam is a highly concentrated ultrasonic beam emitted by an ultrasonic phased array probe. The beam propagation direction and focal point can be controlled by adjusting the excitation timing of the array elements, allowing for precise application to the current transformer detection area. The optical pulse is a short-duration, high-intensity optical signal emitted by a pulse heating source. When applied to the surface of the current transformer, it achieves localized uniform heating, providing a stable temperature excitation source for temperature field data acquisition and amplifying differences in heat loss due to leakage. The sound pressure distribution data is a dataset reflecting the sound pressure intensity at different locations in the detection area, obtained by converting the signal reflected from the focused beam by the ultrasonic phased array probe after reflection from the current transformer surface into an electrical signal and processing it. The attenuation or abnormal distribution of sound pressure can indirectly reflect medium leakage. The resulting structural or media changes; temperature field data is a dataset collected by an infrared thermal imager, reflecting the spatial distribution and temporal changes of surface temperature in the current transformer detection area. It is usually presented in the form of a two-dimensional temperature image and can capture local temperature anomalies caused by the flow of media at the leak (such as low temperature points or sudden changes in temperature drop rate); geometric deformation data is a dataset obtained by comparing real-time point cloud data acquired by a laser scanner through secondary scanning along the detection path with the point cloud data of the corresponding area of ​​the three-dimensional reference contour model. It reflects the surface morphological changes in the detection area and is mainly reflected in the coordinate difference (i.e., deformation) between the real-time point cloud and the reference point cloud. It can reflect the micro-deformation of the structure caused by the leak (such as surface protrusions / depressions caused by microcracks in epoxy resin).

[0036] In this embodiment of the present disclosure, when performing the test along the planned airtightness testing path, it is necessary to coordinate the linkage of multiple testing components through a synchronous control module: First, the ultrasonic phased array probe can be controlled to move along the path and emit a frequency-adjustable focused beam to each testing area on the surface of the current transformer. After the beam penetrates the surface layer, it is reflected by the internal structure or medium. The probe receives the reflected signal and converts it into sound pressure distribution data. At the same time, the pulse heating light source moves synchronously along the same airtightness testing path and emits light pulses with a set heating pulse width to each testing area to heat the area uniformly for a short time. The infrared thermal imager captures the temperature change of each area in real time after heating and generates temperature field data. In addition, the laser scanner scans the current transformer along the path to obtain the real-time point cloud data of each testing area. It registers and calculates the difference between the point cloud data and the corresponding testing area in the three-dimensional reference contour model to obtain geometric deformation data reflecting the surface morphology change. Throughout the process, the emission of the focused beam and light pulses, as well as the acquisition of multiple types of data, are strictly and synchronously executed along the airtightness detection path. This ensures that each set of data accurately corresponds to the same detection area, laying a foundation for consistency in both spatial and temporal dimensions for subsequent characteristic parameter calculations.

[0037] Multiple data types are collected synchronously along the same path, eliminating the need for multiple adjustments to the testing equipment position in stages. This significantly reduces the testing time for a single instrument transformer and eliminates the need for equipment shutdown or destructive operations. Consequently, it is suitable for uninterrupted power supply maintenance scenarios for medium and low voltage instrument transformers below 35kV, solving the problems of traditional offline testing affecting power supply continuity and online testing providing limited data acquisition. Furthermore, the synchronously collected sound pressure, temperature, and geometric deformation data precisely correspond to the same testing area, ensuring data consistency for subsequent multi-physics feature fusion (such as the collaborative calculation of ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate). This reduces misjudgments caused by asynchronous data acquisition, laying the foundation for improving the accuracy of leak detection.

[0038] Step 130: Calculate the ultrasonic energy attenuation rate of each detection area based on the sound pressure distribution data, calculate the temperature gradient standard deviation of each detection area based on the temperature field data, and calculate the curvature change rate of each detection area based on the geometric deformation data.

[0039] Among them, the ultrasonic energy attenuation rate is a quantitative indicator reflecting the degree of energy loss of ultrasonic waves during propagation in the current transformer detection area. It is calculated from sound pressure distribution data. Leakage will cause changes in medium density, which will lead to abnormal ultrasonic energy attenuation. This indicator is the core acoustic characteristic parameter for judging leakage. The temperature gradient standard deviation is a statistical quantity describing the dispersion of temperature gradient distribution in the current transformer detection area. It is calculated based on temperature field data. The flow of medium at the leakage point will cause a sudden change in local temperature gradient. This indicator can quantify the abnormal fluctuation of temperature field and is a key thermal characteristic parameter for identifying leakage. The curvature change rate is a parameter characterizing the degree of deformation of the surface morphology of the current transformer detection area relative to the three-dimensional reference contour model. It is calculated from geometric deformation data. The medium pressure change caused by leakage will lead to micro-deformation of the surface. This indicator can capture this subtle morphological difference and is an important morphological characteristic parameter for judging leakage.

[0040] The above technical solution can produce multi-dimensional synergistic benefits by calculating three types of characteristic parameters: ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate. On the one hand, the three types of parameters rely on wavelet packet decomposition, statistical analysis, point cloud comparison and other technologies to achieve accurate feature extraction from different dimensions of sound, heat and shape. It can capture the minute acoustic energy loss of epoxy resin microcracks <0.1mm through ultrasonic energy attenuation rate, amplify the local thermal differences caused by leakage by temperature gradient standard deviation to avoid interference from natural temperature, and identify micron-level surface deformation through curvature change rate, breaking through the limitation of traditional detection being insensitive to micro-leakage and micro-deformation. On the other hand, the three types of parameters form a mutually verifying feature system based on independent physical principles, which can effectively resist electromagnetic interference in substations, significantly reduce the risk of misjudgment by a single parameter, and significantly improve the accuracy of leakage judgment. At the same time, its calculation method does not require complex equipment or destructive operation, is suitable for uninterrupted power supply detection scenarios of medium and low voltage transformers below 35kV, and can quickly complete the processing through edge computing units, solving the problem of long processing time of traditional detection data. It can provide high-quality data support for subsequent transfer learning model fusion judgment and fully match the power system's needs for detection accuracy, reliability and efficient operation and maintenance.

[0041] Step 140: Use a transfer learning model to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate in the same detection area, and calculate the leakage probability of each detection area.

[0042] Among them, the transfer learning model is a machine learning model that transfers model knowledge trained in the source domain (such as transformer detection data covering multiple sealing failure modes) to the target domain (such as airtightness detection of transformers at a specific voltage level). It does not require training from scratch, can quickly adapt to different scenarios, and improve the model's generalization ability and computational efficiency. Spatiotemporal fusion integrates data from the time and space dimensions for three types of parameters in the same detection area: ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate. This enables collaborative analysis of multi-dimensional features and avoids the limitations of a single dimension or single parameter. The leakage probability is used to characterize the likelihood of sealing failure in the current transformer detection area. Its value ranges from [0,1]. The closer the value is to 1, the higher the leakage risk. It is calculated by fusing multiple feature parameters through the transfer learning model and serves as a quantitative basis for determining leakage.

[0043] In the embodiments of this disclosure, when performing spatiotemporal fusion calculations using a transfer learning model, the ultrasonic energy attenuation rate (E), temperature gradient standard deviation (G), and curvature change rate (G) of the same detection area can first be analyzed. Preprocessing is performed to normalize the three types of parameters to the [0,1] interval, eliminating the interference of dimensional differences on model calculation. Then, a pre-trained transfer learning model is called. This model is trained on 100,000 sets of measured calibration data (source domain data) covering 20 transformer sealing failure modes and 500,000 sets of simulation data at different voltage levels. The weight coefficients (e.g., E corresponds to the weight) have been optimized using gradient descent. The weight is 0.6, and G corresponds to the weight. The weights are 0.3 and K. The parameters are set to 0.1) and a bias term. During model execution, the dynamic changes of the three types of parameters during the detection process are first integrated from a temporal perspective (e.g., E, G, and K value sequences at different sampling times). Then, the distribution characteristics of the parameters within the detection region are correlated from a spatial perspective. Feature mapping is performed through a fully connected hidden layer and the ReLU activation function, and finally, the results are substituted into the formula. (σ is the Sigmoid function), the fusion result is mapped to the [0,1] interval to obtain the leakage probability P of the detection area, thus realizing the spatiotemporal collaborative determination of multiple feature parameters.

[0044] Transfer learning models integrate acoustic, thermal, and shape feature parameters through spatiotemporal fusion, avoiding misjudgments caused by single parameters due to electromagnetic interference, ambient temperature fluctuations, and other factors, thereby improving detection accuracy. Furthermore, transfer learning eliminates the need to retrain models for different voltage levels (especially low- and medium-voltage transformers below 35kV); it can be adapted with only minor adjustments using a small amount of target domain data. This solves the problems of poor generalization ability and limited adaptability of traditional machine learning models, meeting diverse operation and maintenance needs.

[0045] Step 150: If there is a target detection area with a leakage probability greater than a preset threshold, it is determined that the current transformer has a leakage in the target detection area, and the leakage coordinates and leakage rate of the target detection area on the current transformer are determined.

[0046] The preset threshold is a leakage judgment critical value calibrated based on a large number of current transformer sealing failure cases and test data (e.g., it can be set to 0.95). When the leakage probability of the detection area exceeds this value, it can be determined as a target area with leakage risk. The target detection area is the detection area with a leakage probability greater than the preset threshold. It is the core object for subsequent location of leakage coordinates and calculation of leakage rate, corresponding to the specific part of the current transformer where sealing failure may occur. The leakage coordinates are the centroid coordinates of the intersection (overlapping area) of the ultrasonic anomaly area, thermal anomaly area, and deformation anomaly area in the target detection area in the three-dimensional reference contour model of the current transformer. It can be used to accurately map to the specific location of the transformer entity. The leakage rate is a quantitative indicator reflecting the speed of medium leakage in the target detection area. It is calculated based on the medium thermal conductivity coefficient, equivalent leakage area, and temperature drop slope, and is used to assess the severity of leakage.

[0047] In this embodiment of the present disclosure, when the calculated leakage probability value of a certain detection area exceeds a preset judgment threshold, it is determined that there is a leakage problem caused by sealing failure in that specific detection area of ​​the current transformer. Subsequently, the specific spatial coordinates (i.e., leakage coordinates) of the leakage detection area on the current transformer body are further determined, and the rate of medium leakage in that area (i.e., leakage rate) is calculated. Clear leakage coordinates can directly guide maintenance personnel to locate specific leakage points, avoiding blind investigation and wasting time, and significantly improving maintenance efficiency. Determining the leakage rate helps maintenance personnel quantify the severity of the leakage, distinguish between different maintenance priorities such as emergency handling and planned maintenance, and avoid equipment failure caused by excessive maintenance or delayed handling.

[0048] In summary, the current transformer airtightness detection method provided by this invention, by constructing a three-dimensional reference contour model of the current transformer, can accurately identify all parts prone to sealing failure, such as flange connection surfaces and epoxy resin casting joints. The airtightness detection path generated based on this model can achieve full coverage scanning of these key areas, avoiding missed detections from a spatial perspective and laying the foundation for subsequent accurate capture of micro-leakage characteristics. This effectively detects epoxy resin micro-cracks <0.1mm, overcoming the limitations of insufficient early micro-leakage diagnosis capabilities in existing solutions. Furthermore, by emitting focused beams and light pulses along the planned path to the current transformer, and simultaneously acquiring multi-modal data on sound pressure distribution, temperature field, and geometric deformation, no equipment shutdown or destructive operation is required. This method is adaptable to the structure and operating characteristics of medium and low voltage transformers below 35kV, thus meeting the power supply requirements. This system addresses the need for uninterrupted power supply maintenance, resolving the issues of traditional offline solutions affecting power supply continuity and online solutions having a narrow applicability. Based on multimodal data, it calculates ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate, then uses a transfer learning model to spatiotemporally fuse these three parameters. This fully integrates acoustic, thermal, and shape-based multi-physics information, significantly reducing the impact of electromagnetic interference on individual data, decreasing false alarm rates, simplifying data processing, and shortening detection time. Finally, it pinpoints the target detection area based on leakage probability and determines the leakage coordinates and rate. This not only achieves precise leakage location and quantitative assessment but also provides crucial data support for establishing a correlation model between leakage rate and insulation life. This aligns with the needs of intelligent power system operation and maintenance, solving the problems of low detection accuracy and efficiency in existing solutions and their inability to support insulation life analysis.

[0049] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the implementation of this embodiment, this embodiment also provides another method for detecting the airtightness of a current transformer, such as... Figure 2 As shown, the method includes:

[0050] Step 210: Construct a three-dimensional reference profile model of the current transformer, and generate an airtightness detection path covering all parts of the current transformer that are prone to sealing failure based on the three-dimensional reference profile model.

[0051] For the specific implementation process of the embodiments disclosed herein, please refer to the relevant description in step 110 of the embodiment, which will not be repeated here.

[0052] Step 220: Along the airtightness detection path, emit a focused beam and light pulse to the current transformer, and simultaneously collect the sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and obtain the geometric deformation data of each detection area.

[0053] In this embodiment of the disclosure, when performing airtightness testing, multiple devices need to be coordinated to operate synchronously along a pre-planned airtightness testing path: First, an ultrasonic phased array probe can be controlled to move along the path and emit a 1MHz-10MHz tunable focused beam to each testing area of ​​the current transformer. After the beam contacts the surface of the transformer, it is reflected. The probe receives the reflected signal and converts it into an electrical signal. After filtering, amplification, and other processing, sound pressure distribution data reflecting the sound pressure intensity distribution of each area is generated. At the same time, a pulse heating light source moves synchronously along the same path and emits light pulses of a set pulse width to each testing area to heat the area uniformly for a short time to provide temperature excitation. An infrared thermal imager captures the temperature changes of each area in real time after heating and generates dynamic temperature field data. In addition, a laser scanner scans the current transformer along the path to obtain real-time point cloud data of each testing area. This data is then registered with the reference point cloud data of the corresponding area in the three-dimensional reference contour model, and the difference is calculated to finally obtain geometric deformation data reflecting the surface morphology changes. Throughout the process, the emission of focused beams and light pulses, as well as the data acquisition of the four types of equipment, are strictly synchronized according to the path sequence to ensure that each set of sound pressure, temperature, and deformation data accurately corresponds to the same detection area, providing a spatiotemporally consistent data foundation for subsequent characteristic parameter calculations.

[0054] Accordingly, step 220 of the embodiment may specifically include the following steps: controlling the ultrasonic phased array probe to move along the airtightness detection path, emitting focused beams to each detection area on the surface of the current transformer, and receiving the reflected signals of the focused beams in each detection area, converting the reflected signals into electrical signals and processing them into sound pressure distribution data for each detection area; controlling the pulse heating light source to move along the airtightness detection path, emitting light pulses to each detection area to provide a temperature excitation source for the current transformer; using an infrared thermal imager to capture the temperature changes of each detection area in real time, generating temperature field data for each detection area; using a laser scanner to scan the current transformer along the airtightness detection path, obtaining real-time point cloud data for each detection area, and comparing it with the reference point cloud data of each detection area in the three-dimensional reference contour model to obtain surface geometric deformation data for each detection area.

[0055] Step 230: Calculate the ultrasonic energy attenuation rate of each detection area based on the sound pressure distribution data, calculate the temperature gradient standard deviation of each detection area based on the temperature field data, and calculate the curvature change rate of each detection area based on the geometric deformation data.

[0056] In the embodiments of this disclosure, when calculating the ultrasonic energy attenuation rate of each detection area based on sound pressure distribution data, firstly, the preset number of wavelet packet decomposition layers (usually 3-5 layers) can be determined. Wavelet packet decomposition is then performed on the sound pressure distribution data of each detection area according to this layer, decomposing the original time-domain sound pressure signal into sub-signals of different frequencies, simultaneously generating a total coefficient set containing all energy information, as well as detail coefficients reflecting subtle high-frequency fluctuations at each decomposition layer; secondly, through the formula... (in, Let be the energy of all detail coefficients at the j-th decomposition level, and n be the number of detail coefficients contained at the j-th decomposition level. For each detail coefficient at each decomposition level (the k-th detail coefficient at the j-th decomposition level), the energy is calculated, and the energy of each detail coefficient at each level is summed to obtain the total energy of the detail coefficients. Simultaneously, the total energy of the original sound pressure data is calculated based on the total set of coefficients. Finally, the ratio of the sum of the detail coefficient energies to the total energy is used as the ultrasonic energy attenuation rate for the detection area. If there is a sealing failure in the detection area (such as epoxy resin microcracks), the energy loss during ultrasonic wave propagation is aggravated, the energy proportion of the detail coefficients increases, and the corresponding ultrasonic energy attenuation rate increases significantly. This quantitative indicator reflects the leakage characteristics.

[0057] Accordingly, step 230 of the embodiment may specifically include: performing wavelet packet decomposition on the sound pressure distribution data of each detection area according to a preset number of decomposition layers to obtain the total coefficient set and the detail coefficients under each decomposition layer; calculating the sum of the energy of multiple detail coefficients under each decomposition layer and the total energy of the total coefficient set; and determining the ratio of the sum of the energy of multiple detail coefficients under each decomposition layer to the total energy as the ultrasonic energy attenuation rate of the corresponding detection area.

[0058] The preset decomposition level refers to the signal decomposition levels pre-defined during wavelet packet decomposition (e.g., L≥3). The level setting must balance feature extraction accuracy and computational efficiency to ensure effective separation of effective signals and interference noise in the sound pressure data. Wavelet packet decomposition is a signal processing algorithm that decomposes sound pressure distribution data (time-domain signal) into sub-signals of different frequency bands. Through multi-scale analysis, it achieves separation of effective signals and noise, capturing subtle signal changes in high-frequency bands more precisely than traditional wavelet decomposition. The total coefficient set represents all decomposition coefficients obtained after wavelet packet decomposition. The set of numbers encompasses all energy information of the original sound pressure data; the detail coefficients are the decomposition coefficients of the corresponding high-frequency signals in each level of wavelet packet decomposition, mainly reflecting subtle fluctuations in the sound pressure data (such as abnormal signals of sound energy loss caused by leakage), and are key coefficients for extracting leakage correlation features; the sum of the energy of the detail coefficients is the value obtained by summing the energy of all detail coefficients under each decomposition level, representing the total energy of the effective high-frequency signals in the original sound pressure data; the total energy is the total energy of the original sound pressure data calculated based on the total coefficient set, and is the benchmark for measuring the energy proportion of the detail coefficients.

[0059] When calculating the standard deviation of the temperature gradient for each detection area based on temperature field data, firstly, effective pixels can be selected from the temperature field data collected by the infrared thermal imager. This involves removing abnormal pixels that deviate from the average temperature of the area by more than ±5℃ and invalid pixels at the boundaries, and then counting the total number of remaining effective pixels (N). Secondly, for each effective pixel, the temporal temperature gradient reflecting the temperature change of that point over time can be calculated. Simultaneously, the temperature difference between that point and its neighboring pixels at the same time point can be calculated to obtain the spatial temperature gradient reflecting the temperature difference between that point and the surrounding area. These two values ​​are then fused to obtain the temperature gradient value of the effective pixel. Next, based on the temperature gradient values ​​of all valid pixels... The average temperature gradient of the detection area was calculated using an arithmetic mean. Finally, substitute the values ​​into the standard deviation formula. Combining the effective pixel count N and the gradient value of each pixel and gradient mean The standard deviation of the temperature gradient in the detection area is calculated, thus completing the transformation from raw temperature data to thermal characteristic quantification parameters.

[0060] Accordingly, step 230 of the embodiment may specifically include: determining the number of effective pixels in each detection area based on temperature field data; calculating the temporal temperature gradient and spatial temperature gradient of each effective pixel, and calculating the temperature gradient value of each effective pixel based on the temporal temperature gradient and spatial temperature gradient; calculating the average temperature gradient of all effective pixels in each detection area based on the temperature gradient value; and calculating the standard deviation of the temperature gradient of each detection area based on the number of effective pixels, the temperature gradient value of each effective pixel, and the average temperature gradient.

[0061] Among them, the effective pixels are those selected from the pixels in the temperature field data that can truly reflect the temperature state of the detection area. Abnormally high / low temperature pixels (such as extreme temperature points caused by environmental interference) that deviate from the average temperature of the area by more than ±5℃ and invalid pixels at the boundary of the detection area (such as background pixels that exceed the physical range of the transformer) need to be removed; the effective pixel count is the total number of effective pixels (represented by N); the temporal temperature gradient is the gradient value calculated based on the temperature change of the same effective pixel at different times, reflecting the rate of temperature change of the pixel over time. The flow of medium at the leak will cause abnormal temporal temperature gradients; the spatial temperature gradient is the gradient value calculated based on the temperature difference of adjacent effective pixels at the same time, reflecting the spatial distribution difference of temperature in the detection area. The flow of medium caused by the leak will cause abrupt changes in the local spatial temperature gradient; the temperature gradient value is the comprehensive gradient quantization value obtained by fusing the temporal and spatial temperature gradients of the same effective pixel, which can fully reflect the degree of temperature anomaly of the pixel; the temperature gradient mean is the arithmetic mean of the temperature gradient values ​​of each effective pixel; the temperature gradient standard deviation is a statistic describing the degree of dispersion of the temperature gradient values ​​of each effective pixel from the gradient mean.

[0062] When calculating the rate of curvature change of each detection area based on geometric deformation data, firstly, two types of core coordinates can be extracted from the geometric deformation data: one is the real-time point cloud coordinates of each sampling point in the detection area obtained by the laser scanner in real time, and the other is the reference point cloud coordinates in the three-dimensional reference contour model that completely correspond to the position of the detection area, ensuring that the two types of coordinates correspond one-to-one in spatial position; secondly, for each corresponding sampling point, the distance difference between the real-time point cloud coordinates and the reference point cloud coordinates can be calculated by the spatial distance formula, and the average (or the maximum difference) of the distance differences of all sampling points is taken to obtain the real-time point cloud deformation (ΔL) of the detection area; then, in the three-dimensional reference contour model, the length of the reference model along the key force direction (such as the radial direction of the flange sealing surface, the axial direction of the epoxy resin casting) of the detection area is determined. This ensures that the length accurately reflects the standard scale of the detection area; finally, it is substituted into the curvature change rate calculation formula specified in the disclosure document. The ratio of real-time point cloud deformation to baseline model length is converted into a percentage to obtain the curvature change rate of the detection area, thus completing the conversion from geometric deformation data to shape feature quantification parameters.

[0063] Accordingly, step 230 of the embodiment may specifically include: extracting the real-time point cloud coordinates of each detection area from the geometric deformation data of each detection area, and extracting the reference point cloud coordinates of each detection area from the three-dimensional reference contour model of the current transformer; calculating the distance difference between the real-time point cloud coordinates and the reference point cloud coordinates in each detection area to obtain the real-time point cloud deformation of each detection area; determining the reference model length of each detection area in the three-dimensional reference contour model; and calculating the curvature change rate of each detection area based on the real-time point cloud deformation and the reference model length.

[0064] Among them, the real-time point cloud coordinates are the three-dimensional spatial coordinates (X real-time, Y real-time, Z real-time) of each sampling point on the surface of the detection area acquired in real time when the laser scanner scans the detection area along the airtightness detection path, which can accurately reflect the current actual geometric shape of the detection area; the reference point cloud coordinates are the three-dimensional spatial coordinates (X reference, Y reference, Z reference) of the sampling points in the three-dimensional reference contour model that correspond one-to-one with the real-time point cloud coordinates of the detection area, representing the standard geometric shape coordinates when the detection area is leak-free; the real-time point cloud deformation is the distance difference (ΔL) between the real-time point cloud coordinates of the same sampling point and the reference point cloud coordinates, calculated by the formula: It directly reflects the degree of geometric deformation of the surface of the detection area caused by leakage; the length of the reference model is the standard length of the detection area along a specific direction (such as the axial or radial direction of the transformer) in the three-dimensional reference contour model. It is usually taken as the straight-line distance between the cloud coordinates of the two reference points in the detection area as the reference scale for quantifying the rate of curvature change.

[0065] Step 240: Input the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate of the same detection area into the pre-trained transfer learning model to calculate the leakage probability of the corresponding detection area.

[0066] The transfer learning model is trained using 100,000 sets of measured calibration data covering 20 failure modes, including epoxy resin cracking and sealing ring aging, as source domain data, combined with 500,000 sets of simulated operating condition data containing electromagnetic interference characteristics from 10kV to 750kV. This machine learning model eliminates the need for training from scratch for new testing scenarios, as it incorporates optimized weight coefficients and bias terms, allowing for rapid adaptation to transformer testing at different voltage levels. In the leakage probability calculation, based on the model weight coefficients and bias terms obtained from the pre-trained transfer learning model, spatiotemporal fusion calculations are performed on the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate to obtain the leakage probability for the corresponding testing area.

[0067] Step 250: If there is a target detection area with a leakage probability greater than the preset threshold, it is determined that the current transformer has a leakage in the target detection area. Based on the ultrasonic energy attenuation rate, temperature gradient standard deviation and curvature change rate of the target detection area, the ultrasonic abnormal area, thermal abnormal area and deformation abnormal area are marked in the target detection area respectively.

[0068] Among them, the ultrasonic anomaly region is a sub-region in the target detection area where the ultrasonic energy attenuation rate (E) meets the criterion of E > 15%. This region is characterized by abnormal ultrasonic energy loss due to medium leakage, and is an area with abnormal acoustic characteristics of leakage. The thermal anomaly region is a sub-region in the target detection area where the standard deviation of the temperature gradient (G) meets the criterion of G > 0.5℃ / cm. This region is characterized by abrupt changes in the local temperature gradient caused by medium flow at the leakage point, and is an area with abnormal thermal characteristics of leakage. The deformation anomaly region is a sub-region in the target detection area where the rate of change of curvature (E) meets the criterion of E > 15%. )satisfy The sub-region with a judgment standard of >0.1% / mm is a region where surface micro-deformation is caused by pressure changes due to leakage, and is an abnormal region with leakage characteristics.

[0069] In this embodiment of the disclosure, when performing leakage determination and abnormal area marking, the leakage probability of each detection area can first be compared with a preset threshold (e.g., 0.95) to filter out target detection areas with leakage probabilities exceeding the threshold, and to preliminarily determine that there is a leak in the area; then, for the target detection area, the three types of characteristic parameters that have been calculated are called, and the ultrasonic abnormal area with abnormal sound energy loss is marked with the ultrasonic energy attenuation rate E > 15% as the standard; the thermal abnormal area with abrupt temperature gradient change is marked with the temperature gradient standard deviation G > 0.5℃ / cm as the standard; and the curvature change rate is used as the standard. Using a standard of >0.1% / mm, abnormal deformation areas with excessive surface micro-deformation within the specified region are marked. The entire process uses quantitative thresholds as the basis for marking, ensuring that the marking of the three types of abnormal areas accurately corresponds to the acoustic, thermal, and morphological anomalies caused by the leak. This lays the foundation for subsequent location of the leak coordinates by the overlap of multiple abnormal areas.

[0070] Step 260: Calculate the intersection area and union area of ​​the ultrasound abnormality area, thermal abnormality area and deformation abnormality area.

[0071] The intersection area refers to the area of ​​the spatial overlap between the ultrasonic anomaly area, the thermal anomaly area, and the deformation anomaly area. This overlapping area is the area that the three types of leakage characteristics (sound, heat, and shape) point to together, and is likely the area where the actual leakage point is located. The union area refers to the total area of ​​all spatial ranges covered by the ultrasonic anomaly area, the thermal anomaly area, and the deformation anomaly area. It includes the independent parts of the three types of anomaly areas and the overlapping parts, representing all possible suspicious areas related to the leakage.

[0072] In the embodiments of this disclosure, when calculating the intersection area and union area of ​​the three types of abnormal regions, it is necessary to first rely on the spatial coordinate data of the marked ultrasonic abnormal region, thermal abnormal region, and deformation abnormal region within the target detection area, and through spatial geometric operations, determine the part of the three types of regions that overlap simultaneously in the three-dimensional model coordinate system, measure the area of ​​the overlapping part and record it as the intersection area; at the same time, integrate all the coverage areas of the three types of regions, remove the overlapping parts that are repeatedly calculated, and measure the total area of ​​all remaining spatial ranges to obtain the union area.

[0073] Step 270: Determine the ratio of the intersection area to the union area as the overlap degree of the multi-sensor abnormal area. When the overlap degree of the multi-sensor abnormal area is greater than the preset overlap degree threshold, use the centroid coordinates of the overlapping area corresponding to the intersection area in the current transformer as the leakage coordinates of the target detection area on the current transformer.

[0074] Among them, the overlap of the multi-sensor abnormal area is a percentage value calculated by "intersection area / union area × 100%", which quantifies the spatial overlap of the three types of abnormal areas. The higher the value, the more concentrated the areas pointed to by the three types of leakage characteristics are, and the greater the possibility of the leakage point.

[0075] In this embodiment of the disclosure, when determining the leakage coordinates, the overlap is first calculated using the formula "overlap of multiple sensor anomaly areas (intersection area / union area) × 100%" based on the intersection and union areas of the three types of anomaly areas (ultrasonic, thermal, and deformation) already calculated within the target detection area. This overlap is then compared to a preset overlap threshold (e.g., 80%). If the overlap exceeds the preset threshold, it indicates that the three types of leakage features are highly concentrated in the same area, and this overlapping area can be determined as the location of the actual leakage point. Finally, in the three-dimensional reference contour model of the current transformer, the geometric center (centroid) of this overlapping area can be calculated, and its corresponding three-dimensional spatial coordinates (Xc, Yc, Zc) are the leakage coordinates of the target detection area on the transformer entity. The entire process uses the overlap of multiple feature spaces as the core criterion, ensuring the accuracy of leakage coordinate location.

[0076] Step 280: Determine the thermal conductivity coefficient and equivalent leakage area of ​​the medium corresponding to the target detection area, calculate the temperature drop slope of the target detection area based on the temperature field data, and calculate the leakage rate of the target detection area on the current transformer according to the thermal conductivity coefficient, equivalent leakage area and temperature drop slope.

[0077] Among them, the thermal conductivity coefficient of the medium is an inherent thermophysical property parameter of the medium filling the current transformer; the equivalent leakage area is the area of ​​the actual leakage channel in the target detection area equivalent to a circular or rectangular opening, which can be calculated based on the size of the deformation anomaly area associated with leakage in the three-dimensional reference contour model, and characterizes the actual flow capacity of the leakage channel; the temperature drop slope is the rate of temperature change with time in the target detection area, which is calculated by linear fitting of the temperature-time curve of the same effective pixel in the temperature field data.

[0078] In the embodiments of this disclosure, when calculating the leakage rate of the target detection area on the current transformer based on the thermal conductivity of the medium, the equivalent leakage area, and the temperature drop slope, the thermal conductivity of the medium, the equivalent leakage area, and the temperature drop slope can be substituted into the leakage rate calculation formula. The thermal conductivity of the medium Equivalent leakage area With temperature drop slope Multiplying the absolute values ​​yields the leakage rate of the target detection area. This completes the entire process from parameter acquisition to rate quantization.

[0079] In summary, the technical solution of this application ensures no blind spots in detection by constructing a three-dimensional baseline contour model and generating a detection path covering all areas prone to sealing failure. Along the detection path, the ultrasonic phased array probe, pulse heating light source, infrared thermal imager, and laser scanner are simultaneously controlled to collaboratively acquire sound pressure distribution, temperature field, and real-time point cloud data, which are then converted into geometric deformation data without requiring shutdown or destructive operations, thus adapting to uninterrupted power supply scenarios for medium and low voltage transformers below 35kV. Based on multimodal data, wavelet packet decomposition, effective pixel selection and statistical analysis, and point cloud comparison are used to accurately calculate the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate, capturing micro-leakage features from acoustic, thermal, and morphological dimensions. A pre-trained transfer learning model, combined with weight coefficients and bias terms, is used to perform spatiotemporal fusion calculations of the three types of parameters to calculate the leakage probability, and a preset threshold is used to lock the target detection area. Furthermore, three types of abnormal areas are marked, and their intersection and union areas are calculated. After verification with an overlap threshold, the leakage coordinates are determined, and the leakage rate is calculated by combining the medium's thermal conductivity coefficient, equivalent leakage area, and temperature drop slope. It can achieve accurate identification of micro-leakage as a whole, significantly reduce the false judgment rate caused by electromagnetic interference, shorten the detection time, and output positioning and quantitative evaluation results. It can provide data support for intelligent operation and maintenance of power systems and comprehensively solve the problems of weak early diagnosis, limited applicable scenarios, and low accuracy and efficiency of existing detection technologies.

[0080] Furthermore, as Figure 1 and Figure 2 The specific implementation of the method shown in this embodiment provides a current transformer airtightness detection device, such as... Figure 3As shown, the device includes: a generation module 31, a data acquisition module 32, a calculation module 33, and a determination module 34.

[0081] The generation module 31 can be used to construct a three-dimensional reference profile model of the current transformer and generate an airtightness detection path covering all parts of the current transformer that are prone to sealing failure based on the three-dimensional reference profile model.

[0082] The acquisition module 32 can be used to emit focused beams and light pulses to the current transformer along the airtightness detection path, and simultaneously acquire sound pressure distribution data, temperature field data and geometric deformation data of each detection area on the surface of the current transformer.

[0083] Calculation module 33 can be used to calculate the ultrasonic energy attenuation rate of each detection area based on sound pressure distribution data, calculate the temperature gradient standard deviation of each detection area based on temperature field data, and calculate the curvature change rate of each detection area based on geometric deformation data.

[0084] The calculation module 33 can also be used to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, temperature gradient standard deviation and curvature change rate of the same detection area using a transfer learning model, and calculate the leakage probability of each detection area.

[0085] The determination module 34 can be used to determine if there is a leakage in the target detection area if the leakage probability is greater than a preset threshold, and to determine the leakage coordinates and leakage rate of the target detection area on the current transformer.

[0086] In some embodiments of this application, the acquisition module 32 can be specifically used to control the ultrasonic phased array probe to move along the airtightness detection path, emit focused beams to each detection area on the surface of the current transformer, and receive the reflected signals of the focused beams in each detection area, convert the reflected signals into electrical signals and process them into sound pressure distribution data for each detection area; control the pulse heating light source to move along the airtightness detection path, emit light pulses to each detection area to provide a temperature excitation source for the current transformer; use an infrared thermal imager to capture the temperature changes of each detection area in real time, and generate temperature field data for each detection area; use a laser scanner to scan the current transformer along the airtightness detection path, obtain real-time point cloud data of each detection area, and compare it with the reference point cloud data of each detection area in the three-dimensional reference contour model to obtain the surface geometric deformation data of each detection area.

[0087] In some embodiments of this application, when calculating the ultrasonic energy attenuation rate of each detection area based on sound pressure distribution data, the calculation module 33 can be specifically used to perform wavelet packet decomposition on the sound pressure distribution data of each detection area according to a preset number of decomposition layers to obtain the total coefficient set and the detail coefficients under each decomposition layer; calculate the sum of the energy of multiple detail coefficients under each decomposition layer and the total energy of the total coefficient set; and determine the ratio of the sum of the energy of multiple detail coefficients under each decomposition layer to the total energy as the ultrasonic energy attenuation rate of the corresponding detection area.

[0088] In some embodiments of this application, when calculating the standard deviation of the temperature gradient of each detection area based on temperature field data, the calculation module 33 can be specifically used to determine the number of effective pixels in each detection area based on temperature field data; calculate the temporal temperature gradient and spatial temperature gradient of each effective pixel, and calculate the temperature gradient value of each effective pixel based on the temporal temperature gradient and spatial temperature gradient; calculate the mean temperature gradient of all effective pixels in each detection area based on the temperature gradient value; and calculate the standard deviation of the temperature gradient of each detection area based on the number of effective pixels, the temperature gradient value of each effective pixel, and the mean temperature gradient.

[0089] In some embodiments of this application, when calculating the rate of curvature change of each detection area based on geometric deformation data, the calculation module 33 can specifically be used to extract the real-time point cloud coordinates of each detection area from the geometric deformation data of each detection area, and the reference point cloud coordinates of each detection area in the three-dimensional reference contour model of the current transformer; calculate the distance difference between the real-time point cloud coordinates and the reference point cloud coordinates in each detection area to obtain the real-time point cloud deformation of each detection area; determine the reference model length of each detection area in the three-dimensional reference contour model; and calculate the rate of curvature change of each detection area based on the real-time point cloud deformation and the reference model length.

[0090] In some embodiments of this application, when using a transfer learning model to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate of the same detection area to calculate the leakage probability of each detection area, the calculation module 33 can specifically be used to input the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate of the same detection area into the pre-trained transfer learning model to calculate the leakage probability of the corresponding detection area. In the process of calculating the leakage probability, based on the model weight coefficients and bias terms obtained from the pre-training of the transfer learning model, the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate are spatiotemporally fused to calculate the leakage probability of the corresponding detection area.

[0091] In some embodiments of this application, when determining the leakage coordinates and leakage rate of the target detection area on the current transformer, the determination module 34 can be specifically used to mark ultrasonic abnormal areas, thermal abnormal areas, and deformation abnormal areas in the target detection area according to the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate of the target detection area; calculate the intersection area and union area of ​​the ultrasonic abnormal areas, thermal abnormal areas, and deformation abnormal areas; determine the ratio of the intersection area to the union area as the overlap degree of the multi-sensor abnormal area; when the overlap degree of the multi-sensor abnormal area is greater than a preset overlap degree threshold, use the centroid coordinates of the overlapping area corresponding to the intersection area in the current transformer as the leakage coordinates of the target detection area on the current transformer; determine the dielectric thermal conductivity coefficient and equivalent leakage area corresponding to the target detection area; calculate the temperature drop slope of the target detection area based on temperature field data; and calculate the leakage rate of the target detection area on the current transformer according to the dielectric thermal conductivity coefficient, equivalent leakage area, and temperature drop slope.

[0092] It should be noted that other corresponding descriptions of the functional units involved in the current transformer airtightness testing device provided in this embodiment can be found in [reference needed]. Figure 1 and Figure 2 The corresponding descriptions in [the document] will not be repeated here.

[0093] Based on the above, Figure 1 and Figure 2 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 and Figure 2 The method for testing the airtightness of current transformers is shown.

[0094] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0095] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 3 To achieve the above objectives, this application also provides a current transformer airtightness testing device, specifically a personal computer, tablet computer, server, or other network device. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described... Figure 1 and Figure 2 The method for testing the airtightness of current transformers is shown.

[0096] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0097] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0098] Based on the above, Figure 1 and Figure 2 The method shown, and Figure 3 To achieve the above objectives, the virtual device embodiment shown in this application also provides a current transformer airtightness detection system, including: the aforementioned current transformer airtightness detection device; and an AR display terminal, which is used to receive leakage coordinates in real time and overlay them onto the physical image of the current transformer.

[0099] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0100] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0101] This invention, through the construction of a three-dimensional baseline contour model and the generation of a detection path covering all areas prone to sealing failure, ensures blind-spot-free detection. Along the detection path, an ultrasonic phased array probe, pulse heating light source, infrared thermal imager, and laser scanner are simultaneously controlled to collaboratively acquire sound pressure distribution, temperature field, and real-time point cloud data, converting them into geometric deformation data without requiring shutdown or destructive operations. This allows it to adapt to uninterrupted power supply scenarios for medium and low voltage transformers below 35kV. Based on multimodal data, wavelet packet decomposition, effective pixel selection and statistical analysis, and point cloud comparison are used to accurately calculate the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate, capturing micro-leakage features from acoustic, thermal, and morphological dimensions. A pre-trained transfer learning model, combined with weight coefficients and bias terms, is used to perform spatiotemporal fusion calculations of the three types of parameters to determine the leakage probability, and a preset threshold is used to lock the target detection area. Furthermore, three types of abnormal areas are marked, and their intersection and union areas are calculated. After verification with an overlap threshold, the leakage coordinates are determined, and the leakage rate is calculated by combining the medium's thermal conductivity coefficient, equivalent leakage area, and temperature drop slope. It can achieve accurate identification of micro-leakage as a whole, significantly reduce the false judgment rate caused by electromagnetic interference, shorten the detection time, and output positioning and quantitative evaluation results. It can provide data support for intelligent operation and maintenance of power systems and comprehensively solve the problems of weak early diagnosis, limited applicable scenarios, and low accuracy and efficiency of existing detection technologies.

[0102] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0103] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for detecting the airtightness of a current transformer, characterized in that, include: A three-dimensional reference profile model of the current transformer is constructed, and an airtightness detection path covering all parts of the current transformer that are prone to sealing failure is generated based on the three-dimensional reference profile model. Along the airtightness detection path, a focused beam and light pulse are emitted toward the current transformer to simultaneously collect sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and to obtain geometric deformation data of each detection area. The ultrasonic energy attenuation rate of each detection area is calculated based on the sound pressure distribution data; the temperature gradient standard deviation of each detection area is calculated based on the temperature field data; and the curvature change rate of each detection area is calculated based on the geometric deformation data. The leakage probability of each detection area is calculated by using a transfer learning model to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, the temperature gradient standard deviation and the curvature change rate in the same detection area. If a target detection area exists with a leakage probability greater than a preset threshold, it is determined that the current transformer has a leakage in the target detection area, and the leakage coordinates and leakage rate of the target detection area on the current transformer are determined.

2. The method according to claim 1, characterized in that, Along the airtightness detection path, a focused beam and light pulse are emitted towards the current transformer to simultaneously collect sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and to obtain geometric deformation data of each detection area, including: The ultrasonic phased array probe is controlled to move along the airtightness detection path, emits a focused beam to each detection area on the surface of the current transformer, and receives the reflected signal of the focused beam in each detection area. The reflected signal is converted into an electrical signal and processed into sound pressure distribution data of each detection area. The pulse heating light source is controlled to move along the airtightness detection path and emit light pulses to each detection area to provide a temperature excitation source for the current transformer. The temperature changes of each detection area are captured in real time using an infrared thermal imager, and temperature field data of each detection area is generated. The current transformer is scanned along the airtightness detection path using a laser scanner to obtain real-time point cloud data of each detection area. This data is then compared with the reference point cloud data of each detection area in the three-dimensional reference contour model to obtain the surface geometric deformation data of each detection area.

3. The method according to claim 1, characterized in that, The calculation of the ultrasonic energy attenuation rate of each detection area based on the sound pressure distribution data includes: The sound pressure distribution data of each detection area are decomposed by wavelet packet according to the preset number of decomposition layers to obtain the total set of coefficients and the detail coefficients of each decomposition layer. Calculate the sum of the energies of multiple detail coefficients under each decomposition layer, and the total energy of the total coefficient set; The ratio of the sum of the energies of multiple detail coefficients under each decomposition layer to the total energy is determined as the ultrasonic energy attenuation rate of the corresponding detection area.

4. The method according to claim 1, characterized in that, The calculation of the standard deviation of the temperature gradient in each detection area based on the temperature field data includes: The effective pixel count of each effective pixel point in each detection area is determined based on the temperature field data. Calculate the temporal temperature gradient and spatial temperature gradient of each effective pixel, and calculate the temperature gradient value of each effective pixel based on the temporal temperature gradient and the spatial temperature gradient; Based on the temperature gradient value, calculate the average temperature gradient of all valid pixels in each detection area; Based on the number of effective pixels, the temperature gradient value of each effective pixel, and the mean of the temperature gradient, the standard deviation of the temperature gradient of each detection area is calculated.

5. The method according to claim 1, characterized in that, The calculation of the curvature change rate of each detection region based on the geometric deformation data includes: From the geometric deformation data of each detection area, extract the real-time point cloud coordinates of each detection area, and the reference point cloud coordinates of each detection area in the three-dimensional reference contour model of the current transformer; Calculate the distance difference between the real-time point cloud coordinates and the reference point cloud coordinates within each detection area to obtain the real-time point cloud deformation of each detection area; Determine the reference model length of each detection region in the three-dimensional reference contour model; Based on the real-time point cloud shape and the length of the baseline model, the rate of curvature change of each detection region is calculated.

6. The method according to claim 1, characterized in that, The method of using a transfer learning model to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, the temperature gradient standard deviation, and the curvature change rate in the same detection area to calculate the leakage probability of each detection area includes: The ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate of the same detection area are input into a pre-trained transfer learning model to calculate the leakage probability of the corresponding detection area. In the process of calculating the leakage probability, the ultrasonic energy attenuation rate, temperature gradient standard deviation, and curvature change rate are spatiotemporally fused based on the model weight coefficients and bias terms obtained from the pre-training of the transfer learning model to obtain the leakage probability of the corresponding detection area.

7. The method according to claim 1, characterized in that, Determining the leakage coordinates and leakage rate of the target detection area on the current transformer includes: Based on the ultrasonic energy attenuation rate, the temperature gradient standard deviation, and the curvature change rate of the target detection area, ultrasonic abnormality areas, thermal abnormality areas, and deformation abnormality areas are marked in the target detection area, respectively. Calculate the intersection area and the union area of ​​the ultrasound abnormality region, the thermal abnormality region, and the deformation abnormality region; The ratio of the intersection area to the union area is determined as the overlap degree of the multi-sensor abnormal area. When the overlap degree of the multi-sensor abnormal area is greater than the preset overlap degree threshold, the centroid coordinates of the overlapping area corresponding to the intersection area in the current transformer are used as the leakage coordinates of the target detection area in the current transformer. Determine the thermal conductivity coefficient of the medium and the equivalent leakage area corresponding to the target detection area; Calculate the temperature drop slope of the target detection area based on the temperature field data; The leakage rate of the target detection area on the current transformer is calculated based on the thermal conductivity coefficient of the medium, the equivalent leakage area, and the temperature drop slope.

8. A current transformer airtightness testing device, characterized in that, include: The generation module is used to construct a three-dimensional reference contour model of the current transformer and generate an airtightness detection path covering all parts of the current transformer that are prone to sealing failure based on the three-dimensional reference contour model. The acquisition module is used to emit a focused beam and light pulse to the current transformer along the airtightness detection path, and simultaneously acquire the sound pressure distribution data and temperature field data of each detection area on the surface of the current transformer, and obtain the geometric deformation data of each detection area. The calculation module is used to calculate the ultrasonic energy attenuation rate of each detection area based on the sound pressure distribution data, calculate the temperature gradient standard deviation of each detection area based on the temperature field data, and calculate the curvature change rate of each detection area based on the geometric deformation data. The calculation module is also used to perform spatiotemporal fusion of the ultrasonic energy attenuation rate, the temperature gradient standard deviation, and the curvature change rate in the same detection area using a transfer learning model, and to calculate the leakage probability of each detection area. The determination module is used to determine that if there is a target detection area with a leakage probability greater than a preset threshold, the current transformer has a leakage in the target detection area, and to determine the leakage coordinates and leakage rate of the target detection area on the current transformer.

9. A current transformer airtightness testing device, comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

10. A current transformer airtightness detection system, characterized in that, include: The current transformer airtightness testing device as described in claim 9; as well as, An AR display terminal is used to receive the leakage coordinates in real time and overlay them onto the physical image of the current transformer.

Images