An insulation monitoring system between transformer core plates / clamps

Abstract

The present invention provides an insulation monitoring system between transformer core plates / clamps, which belongs to the technical field of transformer cores and comprises: a high-voltage power supply, a non-contact electric target, a measuring circuit, a temperature sensor, a vibration sensor and a current sensor, a data acquisition module, a communication module and a host computer, wherein the host computer is provided with a control module for analyzing measurement data and obtaining insulation monitoring results between transformer core plates / clamps; wherein the high-voltage power supply is used to provide the high-voltage charge required for monitoring; the non-contact electric target is installed at the aggregation of the side edges of the transformer core laminations or at the segmented connection of multiple segments of clamps; the measuring circuit is used to measure the charge induced by the non-contact electric target; the data acquisition module is electrically connected to the temperature sensor, the vibration sensor, the current sensor and the measuring circuit, and is used to collect data detected by these sensors and the measuring circuit; and the communication module is connected to the data acquisition module and is used to send the collected data to the host computer.

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer cores, and in particular relates to an insulation monitoring system between transformer core plates or between clamps. Background Art

[0002] As a crucial current conversion device in power systems, the operating status of transformers is directly related to grid stability and power supply quality. The transformer's laminations are key components, and the insulation performance between laminations and clamps, such as between the cores themselves or between the core and clamps, is a key factor affecting the transformer's service life and safe operation. Therefore, real-time and effective monitoring of the insulation status between transformer laminations and clamps is crucial for preventing insulation failures, extending equipment life, and improving power supply reliability.

[0003] Currently, insulation monitoring between transformer core plates and clamps is typically performed by periodically extracting transformer oil samples and analyzing the dissolved gases in the oil using a gas chromatograph to determine the transformer's operating status. This method can indirectly assess the degree of aging in the transformer's insulation system, but it requires manual sampling, has a long testing cycle, and cannot achieve real-time monitoring.

[0004] 2. Electromagnetic field monitoring: Electromagnetic field sensors are placed on the transformer casing to monitor the transformer's magnetic field distribution in real time. A core insulation fault will cause a localized abnormal magnetic field distribution, allowing the fault location to be determined. This method enables continuous monitoring, but its application is limited due to its strict requirements for sensor placement and the establishment of a magnetic field distribution model.

[0005] 3. Partial Discharge Detection Method: Partial discharge sensors are installed inside or outside the transformer to detect partial discharge signals. When insulation deteriorates, partial discharges occur, allowing the insulation condition to be predicted. This method can detect insulation defects promptly, but it requires high sensor sensitivity and anti-interference capabilities, and cannot precisely locate the fault location.

[0006] The above methods all have the technical problem of low diagnostic accuracy for insulation faults between transformer core plates / clamps and inability to accurately locate the fault location. Summary of the Invention

[0007] In view of this, the present invention provides an insulation monitoring system between transformer core plates / clips, which can solve the technical problems in the prior art of low diagnostic accuracy of insulation faults between transformer core plates / clips and inability to accurately locate the fault location.

[0008] The present invention is achieved in that:

[0009] The first aspect of the present invention provides an insulation monitoring system between transformer core plates / clamps, which includes a high-voltage power supply, a non-contact electric target, a measuring circuit, a temperature sensor, a vibration sensor and a current sensor, a data acquisition module, a communication module and a host computer, wherein the host computer is provided with a control module for analyzing the measurement data and obtaining the insulation monitoring results between the transformer core plates / clamps; wherein the high-voltage power supply is used to provide the high-voltage charge required for monitoring; the non-contact electric target is installed at the side aggregation of the transformer core laminations or the segmented connection of multiple clips; the measuring circuit is used to measure the charge induced by the non-contact electric target; the temperature sensor, vibration sensor and current sensor are respectively used to detect the temperature, vibration and load current data of the transformer during operation; the data acquisition module is electrically connected to the temperature sensor, vibration sensor, current sensor and measuring circuit, and is used to collect data detected by these sensors and the measuring circuit; the communication module is connected to the data acquisition module and is used to send the collected data to the host computer. The non-contact electric target is installed at the side of the transformer core lamination, which means that the sides of the multiple laminations of the transformer core are closely arranged to form a plane, which is recorded as the lamination side plane. Both ends of the non-contact electric target are set on this lamination side plane. The segmented connection of the multi-segment clamp refers to a clamp composed of multiple segments. At the connection position of adjacent segments, an insulator, such as rubber or multi-layer insulating paper, is placed. The connection of the two clamps can be achieved by insulating bolts. The segmented position can be made into an L-shaped connection part and a perforation for the insulating bolt is set. Since these positions have a certain degree of insulation, the numerical detection change of the leakage charge can be formed.

[0010] The control module is configured to perform the following steps:

[0011] S10, establishing the operating characteristics of the transformer core, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance matrix under normal conditions and inter-plate insulation fault conditions;

[0012] S20, receiving the core temperature matrix, core vibration matrix, and core grounding current collected by the data acquisition module as real-time data;

[0013] S30, preprocessing the received real-time data and extracting features of the real-time data, which are recorded as first features;

[0014] S40, matching the first feature with historical features in a preset historical feature database to obtain multiple historical features with the highest matching degree;

[0015] S50, clustering the multiple historical features to obtain cluster features, and using the insulation resistance array of the cluster features as a first predicted insulation resistance array;

[0016] S60, controlling the two ends of the non-contact electric target to move at the two ends of the transformer core, with the movement direction being parallel to the normal direction of the core plate, and controlling the high-voltage power supply to load a high-voltage charge on the non-contact electric target, reading the induced charge value of the non-contact electric target measured by the measurement circuit, and calculating the predicted insulation resistance of the core plate gap based on the induced charge value as a second predicted insulation resistance array;

[0017] S70, fusing the first predicted insulation resistance array and the second predicted insulation resistance array to obtain a comprehensive predicted insulation resistance array;

[0018] S80. Compare the comprehensive predicted insulation resistance array with the insulation resistance threshold of the iron core in a normal state to determine whether there is an insulation fault. If the predicted value is lower than the insulation resistance threshold, it is determined to be an insulation fault and an alarm signal is issued.

[0019] Furthermore, the specific steps of S10 include: when the transformer is operating normally, collecting the temperature, vibration and grounding current data of the core through the temperature sensor, vibration sensor and current sensor, and organizing them into the core temperature matrix, core vibration matrix and core grounding current; at the same time, by measuring the insulation resistance between the core plates, establishing the insulation resistance array under the normal state; simulating the insulation fault state between the plates, and collecting the corresponding temperature, vibration and grounding current data, establishing a feature database under the insulation fault state between the plates, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance array under the fault condition.

[0020] Furthermore, the specific steps of S20 include: real-time collection of monitoring data from the temperature sensor, vibration sensor and current sensor to form a core temperature matrix, a core vibration matrix and a core grounding current, respectively, wherein the sampling frequency of the temperature sensor is set to once per minute, and the sampling time is 5 minutes; the sampling frequency of the vibration sensor is set to once per 10 milliseconds, and the sampling time is 1 minute; the sampling frequency of the current sensor is set to once per second, and the sampling time is 10 minutes.

[0021] Furthermore, the method of preprocessing the received real-time data includes denoising, smoothing and standardization.

[0022] Furthermore, the specific steps of S40 include: establishing a historical feature database containing normal operating conditions and inter-board insulation fault conditions; using Euclidean distance or cosine similarity method to compare the first feature with all features in the historical feature database one by one, and selecting the top 3-5 historical features with the highest similarity as matching results.

[0023] Furthermore, the specific steps of S50 include: using the k-means clustering algorithm to perform cluster analysis on the multiple historical features obtained in step S40, and dividing them into several similar clusters; for each cluster, extracting the average value or median of the insulation resistance array as the first predicted insulation resistance array of the cluster feature.

[0024] Furthermore, the specific steps of S60 include: controlling the non-contact electric target to move back and forth at the side convergence of the transformer core laminations or the segmented connection of the multi-segment clamps, with the movement direction parallel to the normal direction of the core plate; using the high-voltage power supply to apply a high-voltage charge of several thousand volts to tens of thousands of volts to the electric target; the measurement circuit detects the charge value sensed by the electric target in real time, and calculates the predicted insulation resistance of the core plate gap according to the formula R=V / I to form a second predicted insulation resistance array.

[0025] Furthermore, the method for fusing the first predicted insulation resistance array and the second predicted insulation resistance array is a weighted average method.

[0026] Compared with the prior art, the insulation monitoring system between transformer core plates / clamps provided by the present invention has the following beneficial effects:

[0027] 1. Comprehensive monitoring indicators comprehensively reflect changes in insulation status. This method not only monitors physical quantities that indirectly reflect insulation status, such as core temperature, vibration, and ground current, but also directly measures the insulation resistance between the core plates, comprehensively understanding the operating characteristics of the transformer core. The integrated analysis of multi-source data enables a more accurate assessment of insulation status trends.

[0028] 2. Flexible monitoring methods enable real-time online monitoring. This method utilizes non-contact electric target detection technology, eliminating the need for numerous sensors within the transformer. Its simple structure allows for easy integration into the transformer itself. Monitoring data can be transmitted to a central control system in real time via a remote communication module, enabling continuous tracking and monitoring of the transformer core insulation condition.

[0029] 3. Accurate fault diagnosis and precise fault location. This method not only determines whether an insulation fault exists, but also accurately predicts insulation resistance trends by analyzing characteristic changes in multiple physical quantities such as temperature, vibration, current, and insulation resistance, combined with pattern matching of historical data, and ultimately locates the specific faulty core plate.

[0030] In summary, the solution of the present invention solves the technical problem in the prior art that the diagnosis accuracy of the insulation fault between the core plates of the transformer is low and the fault location cannot be accurately located. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0032] Figure 1 A schematic diagram of the composition of the system provided by the present invention;

[0033] Figure 2 Schematic diagram of the non-contact electric target structure;

[0034] Figure 3 Flowchart of the steps executed by the control module;

[0035] In the attached figure, 01 is a metal tube and 02 is an insulating support frame. DETAILED DESCRIPTION

[0036] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0037] like Figure 1 As shown, it is a schematic diagram of the composition of an insulation monitoring system between transformer core plates / clamps provided by the present invention. The system includes a high-voltage power supply, a non-contact electric target, a measuring circuit, a temperature sensor, a vibration sensor and a current sensor, a data acquisition module, a communication module and a host computer. A control module is provided in the host computer for analyzing the measurement data and obtaining the insulation monitoring results between the transformer core plates; wherein, the high-voltage power supply is used to provide the high-voltage charge required for monitoring; the non-contact electric target is installed at the side aggregation of the transformer core laminations or the segmented connection of multiple clamps; the measuring circuit is used to measure the charge induced by the non-contact electric target; the temperature sensor, the vibration sensor and the current sensor are respectively used to detect the temperature, vibration and load current data of the transformer during operation; the data acquisition module is electrically connected to the temperature sensor, the vibration sensor, the current sensor and the measuring circuit, and is used to collect data detected by these sensors and the measuring circuit; the communication module is connected to the data acquisition module and is used to send the collected data to the host computer.

[0038] Among them, the non-contact electric target structure:

[0039] The non-contact electric target consists of a hollow metal tube 01 (such as a copper tube) and two high-voltage insulating support frames 02. The two ends of the metal tube 01 are fixed by the insulating support frames 02, so that the metal tube 01 is suspended in the gap between the transformer core plates.

[0040] The insulating support frame 02 is made of an insulating material (such as ceramic). One end of the insulating support frame 02 is connected to the high-voltage lead, and the other end is connected to the measurement circuit lead. The high-voltage lead is connected to the high-voltage power supply, and the measurement circuit lead is connected to the measurement circuit. The insulating support frame 02 is nested within the empty metal tube 01, and the two can be fitted with a clearance fit. Optionally, balls or integrally molded protrusions can be embedded in the bottom of the cylindrical inner wall of the insulating support frame 02 to reduce friction between the metal tube 01 and the insulating support frame.

[0041] Working principle of non-contact electric target:

[0042] 1. A high-voltage power supply injects high-voltage static charge into the metal tube, causing the metal tube to carry a certain amount of positive charge.

[0043] 2. The control system drives the metal tube to slowly reciprocate along the insulating support frame 02 between the core plates / clamps, with the movement direction parallel to the core plate normal. The same applies to the core and the clamps, or between the clamps. The control system can use a telescopic motor system to drive the metal tube 01, such as a miniature linear screw telescopic motor. The motor is connected to the metal tube 01 through an insulated connection, with the axes of the two aligned. The metal tube 01 can be positioned above the core or, as indicated by the arrow, to the side.

[0044] 3. During the movement, the positively charged metal tube and the two adjacent iron core plates form a simple planar capacitor structure. The two iron core plates are the two electrode plates of the capacitor, and the insulating medium between them is the inter-plate insulation layer.

[0045] 4. When the insulation layer has good performance, the capacitor has good insulation performance, and the positive charge on the metal tube will not leak or transfer significantly.

[0046] 5. When the insulation layer ages or is defective, the insulation performance of the capacitor decreases, and the positive charge on the metal tube will leak proportionally and transfer to the adjacent iron core plate.

[0047] 6. The measurement circuit accurately measures the change in charge on the metal tube and transmits the data to the control system.

[0048] 7. The control system can inversely calculate the insulation resistance value of the insulation layer between the core plates / clamps at that location based on parameters such as the change law of the charge amount and the distance between the metal tube and the plate, using physical models such as the capacitance formula.

[0049] 8. By controlling the metal tube to make scanning motion at the convergence points of different lamination sides or the segmented connections of multi-segment clamps, the insulation resistance distribution diagram of the entire core can be obtained.

[0050] like Figure 2 As shown, the control module is used to perform the following steps:

[0051] S10, establishing the operating characteristics of the transformer core, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance matrix under normal conditions and inter-plate insulation fault conditions;

[0052] S20, receiving the core temperature matrix, core vibration matrix, and core grounding current collected by the data acquisition module as real-time data;

[0053] S30, preprocessing the received real-time data and extracting features of the real-time data, which are recorded as first features;

[0054] S40, matching the first feature with historical features in a preset historical feature database to obtain multiple historical features with the highest matching degree;

[0055] S50, clustering the multiple historical features to obtain cluster features, and using the insulation resistance array of the cluster features as a first predicted insulation resistance array;

[0056] S60, controlling the two ends of the non-contact electric target to move at the two ends of the transformer core, that is, controlling the sliding and extension of the metal tube 01. To control the two ends of the non-contact electric target, the control system can be placed at any end of the metal tube 01, thereby forming a relative movement above or on the side of the two ends of the transformer core; the two ends of the transformer core can be two core laminations forming a detection loop in the embodiment, and the movement direction is parallel to the normal direction of the core plate, and the high-voltage power supply is controlled to load a high-voltage charge on the non-contact electric target, and the induced charge value of the non-contact electric target measured by the measurement circuit is read, and the predicted insulation resistance of the core plate gap is calculated according to the induced charge value as a second predicted insulation resistance array;

[0057] S70, fusing the first predicted insulation resistance array and the second predicted insulation resistance array to obtain a comprehensive predicted insulation resistance array;

[0058] S80. Compare the comprehensive predicted insulation resistance array with the insulation resistance threshold of the iron core in a normal state to determine whether there is an insulation fault. If the predicted value is lower than the insulation resistance threshold, it is determined to be an insulation fault and an alarm signal is issued.

[0059] The specific implementation of the above steps is described in detail below:

[0060] The specific implementation of step S10 is as follows:

[0061] The purpose of this step is to establish the operating characteristics of the transformer core, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance array under normal conditions and inter-plate insulation fault conditions.

[0062] First, during normal operation of the transformer, temperature, vibration, and current sensors are used to collect core temperature, vibration, and ground current data. This data is organized into temperature matrices, vibration matrices, and ground current arrays. Simultaneously, insulation resistance between core plates and clamps is measured to create an insulation resistance array under normal conditions. This data constitutes a characteristic database for the transformer under normal operating conditions.

[0063] Secondly, to simulate inter-board insulation faults, we can artificially create insulation faults, such as intentionally causing insulation damage between core plates. We then collect the corresponding temperature, vibration, and ground current data to build a database of characteristics associated with inter-board insulation faults. This data includes a temperature matrix, a vibration matrix, a ground current matrix, and an insulation resistance matrix under fault conditions.

[0064] The feature database established using this method comprehensively describes the operating characteristics of the transformer core under normal and fault conditions. The temperature matrix reflects the core plate temperature distribution, the vibration matrix reflects the core plate vibration, the ground current array reflects the core ground current, and the insulation resistance array reflects the insulation between the core plates. This feature data provides a foundation for subsequent condition monitoring and fault diagnosis.

[0065] The specific implementation of step S20 is as follows:

[0066] The purpose of this step is to receive the core temperature matrix, core vibration matrix, and core grounding current collected by the data acquisition module as real-time data.

[0067] Specifically, the data acquisition module collects real-time monitoring data from temperature sensors, vibration sensors, and current sensors. This data forms the core temperature matrix, core vibration matrix, and core ground current matrix, respectively. The control module receives this real-time data in preparation for subsequent preprocessing and feature extraction.

[0068] It's important to note that to ensure data reliability and accuracy, the data collection process must adhere to a specific sampling frequency and duration. For example, a temperature sensor can be sampled every minute for a five-minute period; a vibration sensor can be sampled every 10 milliseconds for a one-minute period; and a current sensor can be sampled every 1 second for a ten-minute period. This ensures sufficient data is captured for subsequent analysis.

[0069] The specific implementation of step S30 is as follows:

[0070] The purpose of this step is to preprocess the received real-time data, including denoising, smoothing and standardization, and extract the features of the real-time data, which are recorded as the first features.

[0071] First, the collected raw data needs to be denoised. Since data collected by various sensors in real-world environments is inevitably subject to noise, a filtering algorithm is needed to denoise the data. Common filtering methods include median filtering and low-pass filtering.

[0072] Secondly, the denoised data needs to be smoothed. Since the temperature, vibration, and current data will fluctuate during transformer operation, a sliding average algorithm can be used to smooth the data to better reflect trend changes.

[0073] Third, the preprocessed data needs to be standardized. Since different types of data may have different dimensions and value ranges, in order to eliminate the influence of dimensions, methods such as minimum-maximum normalization or Z-score normalization can be used to standardize the data.

[0074] Finally, features are extracted from the preprocessed temperature matrix, vibration matrix, and current data. Common feature extraction methods include statistical features (mean, standard deviation, skewness, kurtosis, etc.), frequency domain features (Fourier transform, wavelet transform, etc.), and time-frequency features (short-time Fourier transform, wavelet packet analysis, etc.). These features together form the first eigenvector, which provides a basis for subsequent feature matching and fault diagnosis.

[0075] The specific implementation of step S40 is as follows:

[0076] The purpose of this step is to match the first feature with historical features in a preset historical feature database to obtain multiple historical features with the highest matching degree.

[0077] First, a historical feature database needs to be established, which contains the feature data of the transformer in normal operation and various fault conditions. These historical feature data can come from previous test data, actual operation data or fault simulation data.

[0078] Then, an appropriate similarity measurement method is used to compare the first feature extracted in step S30 with all features in the historical feature database one by one, and calculate the similarity between them. Commonly used similarity measurement methods include Euclidean distance, cosine similarity, etc.

[0079] Finally, the top n historical features with the highest similarity are selected as the matching results. Here, n is typically set to around 3-5 and can be adjusted based on actual needs. These highly similar historical features provide an important reference for subsequent fault diagnosis.

[0080] The specific implementation of step S50 is as follows:

[0081] The purpose of this step is to cluster the multiple historical features to obtain cluster features, and use the insulation resistance array of the cluster features as the first predicted insulation resistance array.

[0082] First, cluster analysis is performed on the multiple historical features obtained in step S40. Common clustering algorithms include k-means. Through cluster analysis, these historical features can be divided into several similar clusters, each of which represents a typical operation or failure mode.

[0083] For each cluster, the average or median of the insulation resistance array is extracted as the first predicted insulation resistance array of the cluster feature. These predicted insulation resistance arrays reflect the insulation conditions between transformer core plates / clamps under different operating conditions.

[0084] It should be noted that the selection of cluster centers and the determination of the number of clusters need to be optimized according to the actual situation. Usually, indicators such as silhouette coefficient and cohesion coefficient can be used to evaluate the clustering effect and select the optimal clustering solution.

[0085] The specific implementation of step S60 is as follows:

[0086] The purpose of this step is to control the two ends of the non-contact electric target to move at the two ends of the transformer core, with the movement direction parallel to the normal direction of the core plate, and control the high-voltage power supply to load high-voltage charge on the non-contact electric target, read the induced charge value of the non-contact electric target measured by the measurement circuit, and calculate the predicted insulation resistance of the core plate gap based on the induced charge value as the second predicted insulation resistance array.

[0087] First, the control system needs to precisely control the non-contact electric target to move back and forth within the gap between the transformer core plate and / or the clamp, with the movement direction parallel to the normal direction of the core plate. This movement process can be achieved by a motor drive or linear motor.

[0088] Secondly, during the movement of the target, the control system needs to apply a high-voltage charge to the target. This high-voltage power supply can be a DC high-voltage power supply, with a voltage generally ranging from several thousand volts to tens of thousands of volts.

[0089] Then, the measurement circuit will detect the charge value induced by the target electrode in real time. This induced charge value is inversely proportional to the insulation resistance of the core plate gap and can be calculated using the formula:

[0090]

[0091] Where R is the insulation resistance, V is the high voltage applied to the target electrode, and I is the induced charge value.

[0092] Finally, a second predicted insulation resistance array is generated based on the calculated insulation resistance values. This array reflects the real-time insulation status of the core plate gap.

[0093] It is worth mentioning that safety precautions must be taken when applying high voltage to avoid harm to personnel and equipment. At the same time, to improve measurement accuracy, multiple electrodes can be used for joint measurement to enhance data reliability.

[0094] The specific implementation of step S70 is as follows:

[0095] The purpose of this step is to merge the first predicted insulation resistance array (from step S50 ) and the second predicted insulation resistance array (from step S60 ) to obtain a comprehensive predicted insulation resistance array.

[0096] To achieve this goal, the following methods can be used for data fusion:

[0097] 1. Weighted average method: Give the first and second predictions a certain weight, and then calculate the weighted average as the final prediction. The weights can be adjusted according to their respective credibility.

[0098] 2. Dempster-Shafer evidence theory: The two prediction results are regarded as independent evidence, and the DS theory is used to calculate the comprehensive prediction result and its credibility.

[0099] 3. Fuzzy comprehensive evaluation method: Construct a fuzzy membership function, evaluate the membership of the first prediction and the second prediction respectively, and then use weighted average and other methods to obtain the comprehensive prediction result.

[0100] Regardless of the fusion method used, the goal is to fully utilize the advantages of both prediction results, suppress their respective limitations, and obtain a more reliable comprehensive prediction. This comprehensive prediction insulation resistance array provides an important basis for subsequent fault diagnosis.

[0101] The specific implementation of step S80 is as follows:

[0102] The purpose of this step is to compare the comprehensive predicted insulation resistance array with the insulation resistance threshold of the core under normal conditions to determine whether there is an insulation fault.

[0103] First, we need to determine the threshold for the insulation resistance between the core plates during normal transformer operation. This threshold can be determined through statistical analysis of historical data and is generally set at 3-5 times the standard deviation of the insulation resistance under normal conditions. For example, if the average insulation resistance between the core plates under normal conditions is 1000 megohms and the standard deviation is 50 megohms, the threshold can be set at 900 megohms.

[0104] Then, the comprehensive predicted insulation resistance array obtained in step S70 is compared with the above threshold value. If any of the predicted values ​​is lower than the threshold value, it is determined that an insulation fault exists.

[0105] Finally, once an insulation fault is diagnosed, the control module needs to immediately issue an alarm signal to notify the operation and maintenance personnel to promptly handle it. At the same time, it can also record information such as the time, location, and severity of the fault, providing a basis for subsequent analysis and maintenance.

[0106] In order to better describe the specific embodiments of the present invention, the above specific embodiments are described in more detail below with reference to formulas and variables:

[0107] The specific implementation of step S10 is as follows:

[0108] The purpose of this step is to establish the operating characteristics of the transformer core, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance array under normal conditions and inter-plate insulation fault conditions.

[0109] First, when the transformer is operating normally, the temperature, vibration and ground current data of the core are collected through temperature sensors, vibration sensors and current sensors. The data collected by the temperature sensors can be organized into an n×m-dimensional temperature matrix T normal , where n represents the number of rows of the core plate and m represents the number of columns of the core plate; the data collected by the vibration sensor can be organized into an n×m-dimensional vibration matrix V normal The data collected by the current sensor can form an n-dimensional ground current array I normal At the same time, by measuring the insulation resistance between the core plates, an n×m-dimensional insulation resistance array R under normal conditions is established. normal .

[0110] Secondly, to simulate the insulation fault state between boards, we can artificially create insulation faults, such as deliberately creating insulation damage between core plates, and collect the corresponding temperature, vibration and ground current data to establish a feature database under the insulation fault state between boards. These data include the temperature matrix T under the fault condition. fault , vibration matrix V fault , ground current array I fault and insulation resistance array R fault .

[0111] The characteristic database established by the above method can fully describe the operating characteristics of the transformer core under normal and fault conditions. This characteristic data provides a basic basis for subsequent condition monitoring and fault diagnosis.

[0112] The specific implementation of step S20 is as follows:

[0113] The purpose of this step is to receive the core temperature matrix, core vibration matrix, and core grounding current collected by the data acquisition module as real-time data.

[0114] Specifically, the data acquisition module collects real-time monitoring data from temperature sensors, vibration sensors, and current sensors. Assuming the sampling interval is Δt, at time t, the data collected by the temperature sensor forms a real-time temperature matrix T(t), the data collected by the vibration sensor forms a real-time vibration matrix V(t), and the data collected by the current sensor forms a real-time ground current matrix I(t). The control module receives this data in real time to prepare for subsequent preprocessing and feature extraction.

[0115] In order to ensure the reliability and accuracy of the data, the data acquisition process needs to follow the following sampling parameters: the sampling frequency f of the temperature sensor T =1 / 60Hz, sampling time T T =5min; sampling frequency f of vibration sensor V =100Hz, sampling time T V =1min; sampling frequency of current sensor f I =1Hz, sampling time T I = 10 min. This ensures that sufficient data is obtained for subsequent analysis.

[0116] The specific implementation of step S30 is as follows:

[0117] The purpose of this step is to preprocess the received real-time data, including denoising, smoothing and standardization, and extract the features of the real-time data, which are recorded as the first features.

[0118] First, the collected raw data needs to be denoised. The median filter algorithm can be used to filter the temperature matrix T(t), vibration matrix V(t) and ground current array I(t) respectively to obtain the denoised data. and The formula for median filtering is:

[0119]

[0120] Among them, x i is the original data, is the denoised data, and k is the filter window size.

[0121] Secondly, the denoised data needs to be smoothed. The M-point sliding average algorithm can be used to obtain the smoothed temperature matrix Vibration Matrix and ground current array The formula for the sliding average is:

[0122]

[0123] in, is the denoised data, is the smoothed data, and M is the sliding window size.

[0124] Again, the preprocessed data needs to be standardized. The Z-score standardization method can be used to obtain the standardized temperature matrix T * (t), vibration matrix V * (t) and the ground current array I * (t). The formula for Z-score standardization is:

[0125]

[0126] Among them, x i is the original data, μ is the mean value of the data, σ is the standard deviation of the data, The data are standardized.

[0127] Finally, from the preprocessed temperature matrix T * (t), vibration matrix V * (t) and the ground current array I * (t) to extract statistical features, frequency domain features and time-frequency features to form the first feature vector F1(t). These features include:

[0128] Statistical characteristics: mean μ, standard deviation σ, skewness γ1, kurtosis γ2, etc.

[0129] Frequency domain features: amplitude spectrum, phase spectrum, etc. after Fourier transform

[0130] Time-frequency characteristics: wavelet coefficients after continuous wavelet transform, etc.

[0131] These features provide a basis for subsequent feature matching and fault diagnosis.

[0132] In addition, the preprocessing process may also adopt a customized method, which is not limited to the above description.

[0133] The specific implementation of step S40 is as follows:

[0134] The purpose of this step is to match the first feature F1(t) with the historical features in a preset historical feature database to obtain multiple historical features with the highest matching degree.

[0135] First, a historical feature database needs to be established It contains the characteristic data of the transformer in normal operation and various fault conditions. These historical characteristic data can come from previous test data, actual operation data or fault simulation data.

[0136] Then, the Euclidean distance is used as the similarity measurement method to compare the first feature F1(t) extracted in step S30 with the historical feature database Compare all features in one by one and calculate the Euclidean distance between them:

[0137]

[0138] Among them, F his，i is the i-th feature vector in the historical feature database, and n is the dimension of the feature vector.

[0139] Finally, the top k historical features with the smallest Euclidean distance are selected as the matching results, which are recorded as Here, k is usually set to 3 to 5. These highly similar historical features provide important references for subsequent fault diagnosis.

[0140] The specific implementation of step S50 is as follows:

[0141] The purpose of this step is to Clustering is performed to obtain cluster features, and an insulation resistance array of the cluster features is used as a first predicted insulation resistance array.

[0142] First, the step S40 obtained Perform cluster analysis. The k-means algorithm can be used, and its objective function is:

[0143]

[0144] Among them, k is the number of clusters, C j is the jth cluster, μ j is the cluster center of the jth cluster. Through iterative optimization, Divide into k similar clusters.

[0145] For each cluster C j , extract the insulation resistance array {R his，i ∣x i ∈C j}, as the first predicted insulation resistance array R of the cluster feature pred1，j These predicted insulation resistance arrays reflect the insulation conditions between transformer core plates under different operating conditions.

[0146] It should be noted that the cluster center {μ jThe selection of} and the determination of the number of clusters k need to be optimized according to the actual situation. Indicators such as silhouette coefficient and cohesion coefficient can be used to evaluate the clustering effect and select the optimal clustering scheme.

[0147] The specific implementation of step S60 is as follows:

[0148] The purpose of this step is to control the two ends of the non-contact electric target to move at the two ends of the transformer core, with the movement direction parallel to the normal direction of the core plate, and control the high-voltage power supply to load high-voltage charge on the non-contact electric target, read the induced charge value of the non-contact electric target measured by the measurement circuit, and calculate the predicted insulation resistance of the core plate gap based on the induced charge value as the second predicted insulation resistance array.

[0149] First, the control system needs to accurately control the non-contact electric target to move back and forth at the side convergence of the transformer core laminations or the segment connection of the multi-segment clamps, and the movement direction is parallel to the normal direction of the core plate. Assume that the position of the electric target at time t is The (x, y) plane represents the plane of the core laminations, and the z-axis is the normal. The control system must maintain the z component of x(t) constant while the (x, y) component moves back and forth at the convergence of the core laminations or at the segmented connections of multi-segment clamps.

[0150] Secondly, during the movement of the target, the control system needs to apply a high voltage charge V to the target. This high voltage power supply can be a DC high voltage power supply, and the voltage is generally between several thousand volts and tens of thousands of volts.

[0151] Then, the measurement circuit will detect the charge value Q induced by the target electrode in real time. This induced charge value is inversely proportional to the insulation resistance R of the core plate gap and can be calculated using the formula:

[0152]

[0153] Where I is the induced current and Δt is the sampling time interval.

[0154] Finally, based on the calculated insulation resistance value R, a second predicted insulation resistance array R is formed. pred2 This array reflects the real-time insulation status of the core plate gap.

[0155] It is worth mentioning that safety precautions must be taken when applying high voltage to avoid harm to personnel and equipment. At the same time, to improve measurement accuracy, multiple electrodes can be used for joint measurement to enhance data reliability.

[0156] The specific implementation of step S70 is as follows:

[0157] The purpose of this step is to convert the first predicted insulation resistance array R pred1(from step S50) and the second predicted insulation resistance array R pred2 (from step S60) are merged to obtain a comprehensive predicted insulation resistance array R pred .

[0158] In order to achieve this goal, the weighted average method can be used for data fusion:

[0159] R pred，i =αR pred1，i +(1-α)R pred2，i

[0160] Among them, R pred，i is the comprehensive prediction result, R pred1，i and R pred2，i are the first prediction and the second prediction respectively, and α is a weight coefficient, ranging from 0 to 1. The weight α can be adjusted according to the credibility of each prediction, for example, it can be determined based on the variance or correlation coefficient of historical data.

[0161] By weighted averaging, we can fully utilize the advantages of the two prediction results, suppress their respective limitations, and obtain a more reliable comprehensive prediction insulation resistance array R pred This array provides an important basis for subsequent fault diagnosis.

[0162] The specific implementation of step S80 is as follows:

[0163] The purpose of this step is to combine the comprehensive predicted insulation resistance array R pred The insulation resistance threshold R of the core in normal state th Compare and determine whether there is an insulation fault.

[0164] First, it is necessary to determine the threshold value R of the insulation resistance between the core plates under normal operation of the transformer. th This threshold can be obtained by statistically analyzing historical data R normal It can be obtained, generally taken as 3-5 times the standard deviation of the insulation resistance under normal conditions:

[0165]

[0166] in, and R normal The mean and standard deviation of .

[0167] Then, the comprehensive predicted insulation resistance array R obtained in step S70 is pred With the above threshold R th Perform an element-wise comparison:

[0168] F={R pred，i <R th}

[0169] Where F is a Boolean vector of length n×m. If R pred，i <R th , then F i =true, otherwise F i =false.

[0170] Finally, if any true element in F exists, an insulation fault is detected, and the control module must immediately issue an alarm signal to notify the operation and maintenance personnel to promptly address the problem. At the same time, information such as the time, location, and severity of the fault can be recorded to provide a basis for subsequent analysis and maintenance.

[0171] The following is an explanation table of the formulas or variables used in the above description:

[0172]

[0173]

[0174] Specifically, the basic idea of ​​the present invention is:

[0175] 1. In step S10, a database of temperature, vibration, current, and insulation resistance characteristics of the transformer in normal and fault states is established to provide a basis for subsequent monitoring;

[0176] 2. In steps S20-S30, these physical quantity data are collected and preprocessed in real time to extract representative features;

[0177] 3. In steps S40-S50, a first predicted insulation resistance array is obtained using pattern matching and clustering algorithms;

[0178] 4. In step S60, a second predicted insulation resistance array is obtained through non-contact electrical target detection;

[0179] 5. In step S70, the two prediction results are combined to obtain a comprehensive predicted insulation resistance array;

[0180] 6. In step S80, the prediction result is compared with the threshold value to determine whether there is an insulation fault and issue an alarm signal.

[0181] The specific principle of the present invention is:

[0182] 1. Leverage multi-source physical quantity data to comprehensively reflect insulation status. Changes in the transformer core's insulation status can cause variations in multiple physical quantities, including temperature, vibration, and ground current. The method of the present invention simultaneously monitors these indirect physical quantities that reflect insulation status and, combined with directly measured inter-plate insulation resistance, constructs a comprehensive database of insulation status characteristics. This comprehensive utilization of multi-source data allows for a more accurate description of insulation status variations.

[0183] 2. Accurately diagnose faults using pattern matching technology. The method of the present invention establishes a database of characteristics of the transformer under normal and fault conditions. Using a pattern matching algorithm, it compares real-time monitoring data with historical data to identify the most similar operating mode. Combined with the predicted insulation resistance value, it can accurately determine whether the current insulation state is abnormal and locate the specific faulty core plate. This pattern recognition-based fault diagnosis method can significantly improve diagnostic accuracy compared to empirical judgment based on a single physical quantity.

[0184] 3. Fusion of multi-source prediction results improves monitoring reliability. The method of the present invention utilizes two different insulation resistance prediction methods: one based on indirect physical quantity characteristics, and one based on direct measurement of a non-contact electrical target. These two prediction results are weighted and fused to complement and verify each other, ultimately resulting in a more reliable comprehensive prediction. This data fusion technology can effectively mitigate the limitations of a single prediction method and improve the accuracy and stability of overall monitoring.

[0185] A specific embodiment of the present invention is provided below:

[0186] The present invention provides an online monitoring method for the insulation status between transformer core plates based on multi-source data fusion, which is described below in conjunction with specific data.

[0187] 1. System composition

[0188] (1) High voltage power supply: It can provide 10kV DC high voltage charge, which is used to load high voltage to the non-contact target electrode.

[0189] (2) Non-contact electric target: It is installed at the side of the transformer core lamination or the segmented connection of the multi-section clamp. It can move back and forth at the side of the core lamination or the segmented connection of the multi-section clamp. The movement direction is parallel to the normal direction of the core plate.

[0190] (3) Measurement circuit: used to detect the charge value sensed by the non-contact electric target in real time.

[0191] (4) Temperature sensor, vibration sensor and current sensor: used to detect the temperature, vibration and grounding current of the transformer core respectively.

[0192] (5) Data acquisition module: connected to temperature sensor, vibration sensor, current sensor and measurement circuit to collect various physical quantity data.

[0193] (6) Communication module: connected to the data acquisition module to transmit the monitoring data to the host computer in real time.

[0194] (7) Control module: responsible for the control of the monitoring process and fault diagnosis and analysis.

[0195] The monitoring system can collect multi-source physical quantity data such as transformer core temperature, vibration, grounding current and insulation resistance of core plate gap in real time, and transmit the data to the host computer for analysis and diagnosis.

[0196] 2. Data Collection and Preprocessing

[0197] (1) Temperature, vibration and current data acquisition

[0198] Temperature sensors are placed on each of the transformer's 24 core plates, with a sampling frequency of 1 time per minute and a sampling duration of 5 minutes. The collected temperature data can be combined into a 24×24-dimensional temperature matrix T.

[0199] Vibration sensors are placed at the four corners of the transformer core, with a sampling frequency of 100 Hz and a sampling duration of 1 minute. The collected vibration data can be combined into a 4×4 vibration matrix V.

[0200] The current sensor is placed on the neutral grounding wire of the transformer, with a sampling frequency of 1 Hz and a sampling duration of 10 minutes. The collected current data can be combined into a 24-dimensional ground current array I.

[0201] (2) Temperature, vibration and current data preprocessing

[0202] The collected raw temperature, vibration and current data have some noise interference and need to be denoised. Here we use the median filter algorithm with a filter window size of 5. The temperature matrix after filtering is recorded as The vibration matrix is ​​recorded as The ground current array is recorded as

[0203] Next, in order to eliminate the dimensional differences between temperature, vibration and current data, the preprocessed data is standardized. Here, the Z-score standardization method is used to obtain the standardized temperature matrix T * , vibration matrix V * and ground current array I * .

[0204] (3) Feature extraction

[0205] From the normalized temperature matrix T * , vibration matrix V * and ground current array I * In the CNN, statistical features, frequency domain features and time-frequency features are extracted respectively to form the first feature vector F1.

[0206] Statistical features include: mean value μ T , μ V , μ I , standard deviation σ T ,σ V ,σ I , skewness γ 1，T , γ 1，V , γ 1，I , kurtosis γ 2，T , γ 2，V , γ 2，I wait.

[0207] Frequency domain features include: amplitude spectrum and phase spectrum obtained by Fourier transform of temperature matrix, vibration matrix and ground current array.

[0208] The time-frequency characteristics include wavelet coefficients obtained by continuous wavelet transform of temperature matrix, vibration matrix and ground current matrix.

[0209] 3. Establishment of historical feature database

[0210] In order to achieve accurate diagnosis of the transformer core insulation status, it is necessary to establish a historical feature database containing normal and fault conditions in advance.

[0211] (1) Normal state characteristic data

[0212] During normal operation of the transformer, temperature, vibration, and ground current data were collected for a period of one year, totaling 526,560 samples (1 time / min, 5 minutes sampling, for a total of one year). After preprocessing and feature extraction, these data formed the temperature matrix T under normal conditions. normal , vibration matrix V normal , ground current array I normal and insulation resistance array R normal Among them, T normal and V normal The scale is 24×24, I normal The scale is 24, R normal The scale is 24×24.

[0213] (2) Fault status characteristic data

[0214] In order to simulate transformer insulation failure, six different degrees of insulation damage were artificially created in the laboratory environment, and the corresponding temperature, vibration and ground current data were collected. After preprocessing and feature extraction, the temperature matrix T under the fault state was formed. fault , vibration matrix V fault , ground current array I fault and insulation resistance array R fault .

[0215] Organize the above feature data under normal and fault conditions into a historical feature database Provide a basis for subsequent pattern matching and fault diagnosis.

[0216] 4. Real-time monitoring and fault diagnosis

[0217] (1) Real-time data collection

[0218] During transformer operation, the data acquisition module collects temperature, vibration, and ground current data in real time. The temperature sensor samples at a frequency of 1 Hz and a sampling period of 5 minutes, generating a real-time temperature matrix T(t). The vibration sensor samples at a frequency of 100 Hz and a sampling period of 1 minute, generating a real-time vibration matrix V(t). The current sensor samples at a frequency of 1 Hz and a sampling period of 10 minutes, generating a real-time ground current matrix I(t).

[0219] (2) Real-time data preprocessing

[0220] De-noise, smooth and standardize the collected real-time temperature, vibration and current data to obtain pre-processed data and Then, statistical features, frequency domain features and time-frequency features are extracted from these data to form the first feature vector F1(t).

[0221] (3) Historical feature matching

[0222] The first eigenvector F1(t) is compared with the historical feature database Calculate the Euclidean distance of all the features in the dataset and get the top 5 historical features with the highest similarity.

[0223] (4) Cluster analysis

[0224] right Perform k-means cluster analysis and divide it into 3 clusters, C1, C2, and C3. Calculate the average value of the insulation resistance array of each cluster, R pred1，1 , R pred1，2 , R pred1，3 , as the first predicted insulation resistance.

[0225] (5) Contact measurement

[0226] The non-contact electric target is controlled to move back and forth at the side convergence of the core laminations or the segmented connection of the multi-segment clamps, with a movement range of (x, y)∈[-0.5, 0.5]m, (z)=0.1m. At the same time, a high voltage of 10kV is applied to the electric target. The measurement circuit detects the charge value Q induced by the electric target in real time, and calculates the real-time insulation resistance R(t) according to the formula R=V / I=V / (Q / Δt), which is used as the second predicted insulation resistance R pred2 (t).

[0227] (6) Data fusion and fault diagnosis

[0228] The first predicted insulation resistance R pred1 and the second predicted insulation resistance R pred2 (t) Perform weighted average fusion to obtain the comprehensive predicted insulation resistance R pred (t). The weight of the first prediction result is α=0.6, and the weight of the second prediction result is 1-α=0.4, which is determined based on the variance characteristics of historical data.

[0229] Then, the comprehensive predicted insulation resistance R pred (t) and the threshold value R of the insulation resistance under normal conditions th =900MΩΩ for comparison. If R pred If any value of (t) is lower than the threshold, it is judged that there is an insulation fault and an alarm signal is immediately issued.

[0230] The above real-time monitoring and fault diagnosis process allows for a comprehensive assessment of the insulation condition between transformer core plates. A detailed analysis of the monitoring results is provided below.

[0231] 5. Analysis of monitoring results

[0232] Taking a routine inspection as an example, the monitoring results of the insulation status between the transformer core plates are analyzed.

[0233] (1) Real-time data analysis

[0234] The inspection time is 9:00-14:00 on March 1, 2024. Figure 2 The temperature matrix T during this period is given * (t), vibration matrix V * (t) and the ground current array I * The normalized results for (t) show that the temperature and vibration fluctuate significantly over time, and the ground current also increases to a certain extent. These changes may be caused by the degradation of the insulation performance of one of the core plates.

[0235] (2) Pattern matching analysis

[0236] The first feature vector F1(t) is compared with the historical feature database Perform Euclidean distance matching to obtain the five historical features with the highest similarity Through k-means cluster analysis, we found It is mainly concentrated in cluster C2, and the average insulation resistance of this cluster is R pred1，2 =850MΩ. This indicates that the current operating state is similar to a partial insulation fault state.

[0237] (3) Non-contact measurement

[0238] The insulation resistance R of the core plate gap is obtained in real time through the measurement of the non-contact electric target electrode. pred2 The results show that at some locations on the core plate, the insulation resistance is lower than normal, with the lowest value being around 700MΩ.

[0239] (4) Comprehensive diagnostic results

[0240] The first predicted insulation resistance R pred1，2 =850MΩ and the second predicted insulation resistance R pred2 (t) Perform weighted fusion to obtain the comprehensive predicted insulation resistance R pred (t). Since most of the predicted values ​​are lower than the threshold R in the normal state th =900MΩ, so it is determined that there is an insulation fault and a fault alarm is issued.

[0241] Further analysis revealed that the fault was primarily concentrated in the lower area of ​​the transformer, likely due to poor heat dissipation in this area, which accelerated the aging of the insulation. Therefore, it was recommended that maintenance personnel conduct an on-site inspection and address the problem as soon as possible to prevent further expansion of the fault.

[0242] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. A transformer core plate / clamp insulation monitoring system, characterized in that: It includes a high-voltage power supply, a movable non-contact electric target, a measuring circuit, a temperature sensor, a vibration sensor, a current sensor, a data acquisition module, a communication module and a host computer. The host computer is provided with a control module for analyzing the measurement data and obtaining the insulation monitoring results between the transformer core plates / clamps; wherein the high-voltage power supply is used to provide the high-voltage charge required for monitoring to the non-contact electric target; the non-contact electric target is installed at the aggregation of the side edges of the transformer core laminations or the segmented connection of multiple clamps; the measurement circuit is used to measure the charge induced by the non-contact electric target; the temperature sensor, vibration sensor and current sensor are respectively used to detect the temperature, vibration and load current data of the transformer during operation; the data acquisition module is electrically connected to the temperature sensor, vibration sensor, current sensor and measurement circuit, and is used to collect data detected by these sensors and measurement circuit; the communication module is connected to the data acquisition module, and is used to send the collected data to the host computer; The control module is used to perform the following steps: S10, establish the operating characteristics of the transformer core, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance array under normal state and inter-plate insulation fault state; S20, receiving the data acquisition module to collect the core temperature matrix, core vibration matrix, core ground current, as real-time data; S30, preprocessing the received real-time data and extracting features of the real-time data, recorded as the first feature; S40, matching the first feature with historical features in a preset historical feature database to obtain multiple historical features with the highest matching degree; S50, clustering the multiple historical features to obtain cluster features, and using the insulation resistance array of the cluster features as a first predicted insulation resistance array; S60, controlling the non-contact electric target to move along a straight line at both ends of the transformer core, with the movement direction being parallel to the normal direction of the core plates, controlling the high-voltage power supply to load a high-voltage charge on the non-contact electric target, reading the induced charge value of the non-contact electric target measured by the measurement circuit, and calculating the predicted insulation resistance of the core plate gap based on the induced charge value as a second predicted insulation resistance array; the two ends of the transformer core are laminations of any two cores that form a loop for charge movement; S70, fusing the first predicted insulation resistance array and the second predicted insulation resistance array to obtain a comprehensive predicted insulation resistance array; S80. Compare the comprehensive predicted insulation resistance array with the insulation resistance threshold of the iron core in a normal state to determine whether there is an insulation fault. If the predicted value is lower than the insulation resistance threshold, it is determined to be an insulation fault and an alarm signal is issued.

2. The insulation monitoring system between transformer core plates / clamps according to claim 1, characterized in that: The specific steps of S10 include: when the transformer is operating normally, collecting the temperature, vibration and grounding current data of the core through the temperature sensor, vibration sensor and current sensor, and organizing them into the core temperature matrix, core vibration matrix and core grounding current; at the same time, by measuring the insulation resistance between the core plates / clamps, an insulation resistance array under normal conditions is established; simulating the insulation fault state between the plates, and collecting the corresponding temperature, vibration and grounding current data, and establishing a feature database under the insulation fault state between the plates, including the core temperature matrix, core vibration matrix, core grounding current and insulation resistance array under the fault condition.

3. The insulation monitoring system between transformer core plates / clamps according to claim 1, characterized in that: The specific steps of S20 include: real-time collection of monitoring data from the temperature sensor, vibration sensor and current sensor to form a core temperature matrix, a core vibration matrix and a core grounding current, respectively, wherein the sampling frequency of the temperature sensor is set to once per minute, and the sampling time is 5 minutes; the sampling frequency of the vibration sensor is set to once per 10 milliseconds, and the sampling time is 1 minute; the sampling frequency of the current sensor is set to once per second, and the sampling time is 10 minutes.

4. The insulation monitoring system between transformer core plates / clamps according to claim 1, characterized in that: The methods for preprocessing the received real-time data include denoising, smoothing and standardization.

5. The insulation monitoring system between transformer core plates / clamps according to claim 1, characterized in that: The specific steps of S40 include: establishing a historical feature database containing features in normal operating conditions and inter-board insulation fault conditions; using Euclidean distance or cosine similarity method to compare the first feature with all features in the historical feature database one by one, and selecting the top 3-5 historical features with the highest similarity as matching results.

6. The transformer core plate / clamp insulation monitoring system according to claim 1, characterized in that: The specific steps of S50 include: using the k-means clustering algorithm to perform cluster analysis on the multiple historical features obtained in step S40, and dividing them into several similar clusters; for each cluster, extracting the average value or median of the insulation resistance array therein as the first predicted insulation resistance array of the cluster feature.

7. The transformer core plate / clamp insulation monitoring system according to claim 1, characterized in that: The specific steps of S60 include: controlling the non-contact electric target to move back and forth at the side convergence of the transformer core laminations or the segmented connection of the multi-segment clamps, with the movement direction parallel to the normal direction of the core plate; using the high-voltage power supply to apply a high-voltage charge of several thousand volts to tens of thousands of volts to the electric target; the measurement circuit detects the charge value sensed by the electric target in real time, and calculates the predicted insulation resistance of the core plate gap according to the formula R=V / I to form a second predicted insulation resistance array.

8. The transformer core plate / clamp insulation monitoring system according to claim 1, characterized in that: The method for fusing the first predicted insulation resistance array and the second predicted insulation resistance array is a weighted average method.

Images