Power transformer state monitoring method based on multi-modal data analysis
Through multimodal data analysis, multiple data of power transformers are collected and correlated, and feature fusion images are constructed, which solves the problem of isolation of transformer monitoring data, realizes more accurate and extensive condition monitoring, predicts faults and reduces power outages for maintenance.
Patent Information
- Application Number
- CN202510699021.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-28
- Publication Date
- 2025-09-09
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing online monitoring system cannot effectively utilize the relationship between multiple data, resulting in insufficient accuracy in transformer monitoring and an inability to deeply reflect the operating status of the transformer.
Through multimodal data analysis, the mechanical vibration, infrared imaging, partial discharge and oil detection data of the power transformer are collected, a multimodal feature fusion image is constructed, feature point selection and data association are performed, and comprehensive monitoring is achieved.
It improves the accuracy and coverage of transformer monitoring, can predict fault causes, reduce the number of fixed-point inspections and repairs, and minimize the impact of power outages.
Smart Images
Figure CN120611213A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of transformer monitoring, and in particular to a power transformer state monitoring method based on multimodal data analysis. Background Art
[0002] Transformers are essential equipment for power transmission and distribution, widely used in industries such as industry, agriculture, transportation, and urban communities. The stable operation of transformers plays an important role in power transmission. However, due to the complex structure, variable internal parameters, and harsh operating environments of transformers, online monitoring and status analysis of transformers are of great significance and importance.
[0003] The power industry has relatively mature detection and maintenance methods and work procedures for faults that cause transformers to fail to operate normally, such as overall faults caused by moisture, overheating, and aging, and local faults caused by arc discharge and partial discharge.
[0004] Online monitoring is gaining more and more attention. Without affecting the normal operation of the monitored transformer, online monitoring can analyze the operating status of the transformer through continuous sampling of the status quantity reflecting the transformer's operating status, and infer the changes or evolution trends of possible faults that have occurred or are occurring inside the transformer. Therefore, in a sense, online monitoring and existing laboratory fixed-point testing complement each other. The former provides timely intelligence for the latter, and the latter confirms and supplements the test results of the former. Full utilization of online monitoring technology can, on the one hand, timely discover and identify transformer faults. On the other hand, the high degree of utilization and wide application of monitoring data can reduce the number of fixed-point inspections and maintenance of transformers, reduce operating costs, and reduce the impact of power outages and maintenance on power transmission and transformation operations.
[0005] At present, the existing monitoring system still has deficiencies. The existing online monitoring system generally uses sensors to collect, process, and judge data, thereby realizing the judgment of a certain parameter of the transformer and thus judging the operating status of the transformer. Therefore, this method causes isolation between multiple data, and cannot reflect the operating status of the transformer more deeply through the relationship between multiple data, thereby reducing the accuracy of transformer monitoring;
[0006] In response to the above technical problems, this application proposes a solution. Summary of the Invention
[0007] In the present invention, multimodal data collection is performed on the power transformer, and the collected data results are processed to construct a multimodal feature fusion image. Feature points are selected in the feature fusion image according to a random distribution method. The operating data corresponding to the feature points are verified to achieve comprehensive monitoring of a large amount of operating data during the operation of the power transformer. The problem that the monitoring data of the power transformer monitoring system is isolated and cannot dynamically reflect the operating status of the transformer through the correlation between multiple data is solved, and a power transformer status monitoring method based on multimodal data analysis is proposed.
[0008] The purpose of the present invention can be achieved through the following technical solutions:
[0009] The power transformer condition monitoring method based on multimodal data analysis includes the following steps:
[0010] Step 1: Collect multi-source data on the operation data of the power transformer, obtain the mechanical vibration data, infrared imaging data, partial discharge data, and oil detection data of the power transformer, and organize the collected data to create a data array;
[0011] Step 2: Perform amplitude-time analysis on the mechanical vibration in the data array, perform temperature gradient analysis on the infrared imaging data, perform electric field-time analysis on the partial discharge data, and perform solubility-time analysis on the oil test data to obtain data analysis results;
[0012] Step 3: Create correlations for the data analysis results, obtain the power transformer monitoring image results, and output them;
[0013] Step 4: Randomly select points from the power transformer monitoring image results, and obtain the point analysis results of the power transformer based on the selected data;
[0014] Step 5: Statistically organize the multiple acquired point analysis results of the power transformer to obtain an evaluation of the power transformer's operating status.
[0015] As a preferred embodiment of the present invention, it also includes a power transformer status monitoring system, which specifically includes a front-end data acquisition module, a data array creation module, a data analysis module, a random monitoring module and a statistical processing module;
[0016] The front-end data acquisition module is used to collect the operating data of the power transformer to obtain mechanical vibration data, infrared imaging data, partial discharge data and oil detection data;
[0017] The data analysis module performs data shielding on the mechanical vibration data to obtain vibration net value data and data acquisition time, performs screen segmentation and temperature recognition on the infrared imaging data to obtain temperature characterization values and temperature distribution areas, performs data sorting on the partial discharge data to obtain discharge intensity and discharge duration, and calculates the oil detection data and acquisition time to obtain the oil dissolution amount change rate;
[0018] The data array creation module uses time as the horizontal axis, vibration intensity, discharge intensity, oil dissolution amount and temperature characterization values as the multivariate vertical axis, and the power transformer area as the normal axis to create a multi-axis image, and records the analysis results of the data analysis module in the multi-axis image;
[0019] The random monitoring module is used to select points in the multi-axis image, and compare the selected point data with the set safe operation standards to obtain the point operation results;
[0020] The statistical collating module integrates the point operation results to obtain a comprehensive operation result, and records it as the power transformer operation status evaluation.
[0021] As a preferred embodiment of the present invention, the front-end data acquisition module includes a vibration acquisition module, an infrared acquisition module, a discharge monitoring module and a spectrum analysis module;
[0022] The vibration acquisition module is used to collect vibration data, wherein the vibration data includes vibration intensity, vibration frequency and acquisition time;
[0023] The infrared monitoring module collects infrared images, the discharge monitoring module is used to collect partial discharges, and the spectrum analysis module performs spectrum analysis on the oil through the oil observation window, and obtains oil detection data based on the spectrum analysis results. The oil detection data is the amount of gas dissolved in the oil.
[0024] As a preferred embodiment of the present invention, when the data analysis module performs data shielding on mechanical vibration data, it selects vibration data and compares the vibration frequency in the vibration data with a preset frequency range, eliminates vibration data outside the preset frequency range, and then selects vibration intensity, eliminates vibration data with vibration intensity less than a set intensity threshold, records the vibration intensity in the remaining vibration data as vibration net value data, and retains the acquisition time.
[0025] As a preferred embodiment of the present invention, the data analysis module processes the infrared image through an algorithm to eliminate spatial noise. The data analysis module divides the transformer into several wet temperature zones and dry temperature zones according to set conditions. The data analysis module performs weighted average on the temperatures of the wet temperature zones and the dry temperature zones to obtain temperature characterization values of the wet temperature zones and the dry temperature zones.
[0026] When the data analysis module calculates the temperature characterization value, a wet temperature zone or a dry temperature zone is selected, and the area of the selected area is calculated. At the same time, the corresponding areas of different temperatures in the selected area are calculated with a gradient of 1°C, and the temperatures in the selected area are weighted averaged with the corresponding areas as weights to obtain the temperature characterization value.
[0027] As a preferred embodiment of the present invention, when the data creation array is recorded in a multi-axis image, a vibration intensity change curve is drawn through the time horizontal axis and the vibration intensity vertical axis, a discharge intensity curve is drawn through the time horizontal axis and the discharge intensity vertical axis, a solubility change curve is drawn through the time horizontal axis and the oil dissolution amount vertical axis, and a regional temperature change curve is drawn with the temperature characterization value as the vertical axis, time as the horizontal axis, and the area as the normal axis.
[0028] As a preferred embodiment of the present invention, the random monitoring module randomly selects a point in the multi-axis image, obtains the coordinates of the horizontal axis and the normal axis corresponding to the point, and obtains the corresponding vibration intensity, discharge intensity, oil dissolution amount, temperature zone and its corresponding temperature characterization value based on the horizontal axis coordinate;
[0029] The random monitoring module compares the vibration intensity, discharge intensity, temperature characterization value and oil dissolution amount corresponding to the acquisition point with the set standard range. If the comparison results are all within the standard range, it is recorded as a normal operating point. If any comparison result is not within the standard range, it is recorded as a faulty operating point.
[0030] As a preferred embodiment of the present invention, the statistical arrangement module records the points where the vibration intensity and the oil dissolution amount are both outside the standard range as vibration interference points, and records the points where the temperature characterization value and the discharge intensity are both outside the standard range as temperature interference points;
[0031] The statistical collation module calculates the proportion of faulty operation points, vibration interference points and temperature interference points in the total number of points, and performs threshold judgment on the calculation results. It outputs an abnormal operation signal or a normal operation signal based on the judgment results, and outputs vibration interference abnormality and temperature interference abnormality signals at the same time.
[0032] Compared with the prior art, the present invention has the following beneficial effects:
[0033] 1. In the present invention, multimodal data collection is performed on the power transformer, and the collected data results are processed to construct a multimodal feature fusion image. Feature points are selected in the feature fusion image according to a random distribution method. The operating data corresponding to the points are verified to achieve comprehensive monitoring of a large amount of operating data during the operation of the power transformer. The random distribution is used to fully realize the instantaneous state detection during the verification process. The results of the random distribution are statistically analyzed to verify the long-term operating status of the power transformer, thereby expanding the coverage of the power transformer monitoring operation.
[0034] 2. In the present invention, by establishing a correlation between transformer oil data and vibration data, and a correlation between temperature characterization values and partial discharge data, the correlated fault data in the power transformer is monitored, and the cause of the power transformer failure is predicted and analyzed based on the proportion of the correlated fault data in the overall failure, thereby improving the functionality of the power transformer monitoring system.
[0035] 3. In the present invention, by dividing the power transformer into dry and wet areas and selecting temperature characterization values based on the temperature gradient and temperature distribution area, the overall flexibility and data pertinence of transformer temperature monitoring are greatly improved, thereby providing better data support for the operation monitoring of the power transformer. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.
[0037] Figure 1 is a system block diagram of the present invention;
[0038] Figure 2 It is a system flow chart of the present invention. DETAILED DESCRIPTION
[0039] The following is a clear and complete description of the technical solutions of the present invention in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0040] Example 1:
[0041] See also Figure 1 - Figure 2 As shown, the power transformer condition monitoring method based on multimodal data analysis includes the following steps:
[0042] Step 1: Collect multi-source data on the operation data of the power transformer, obtain the mechanical vibration data, infrared imaging data, partial discharge data, and oil detection data of the power transformer, and organize the collected data to create a data array;
[0043] Step 2: Perform amplitude-time analysis on the mechanical vibration in the data array, perform temperature gradient analysis on the infrared imaging data, perform electric field-time analysis on the partial discharge data, and perform solubility-time analysis on the oil test data to obtain data analysis results;
[0044] Step 3: Create a correlation between the data analysis results, so that the results of multiple data analyses are created in the same three-dimensional coordinate system, obtain the power transformer monitoring image results, and output them;
[0045] Step 4: Randomly select points from the power transformer monitoring image results, and obtain the point analysis results of the power transformer based on the selected data, so as to determine whether the operation of the power transformer corresponding to the point meets the standard requirements;
[0046] Step 5: Statistically organize the multiple acquired point analysis results of the power transformer to obtain an evaluation of the power transformer's operating status, including the overall normal operation of the power transformer, the interference of the power transformer's vibration on the oil and gas content, and the interference of partial discharge on the internal environment of the power transformer.
[0047] Example 2:
[0048] See also Figure 1 - Figure 2 As shown, it also includes a power transformer status monitoring system, which specifically includes a front-end data acquisition module, a data array creation module, a data analysis module, a random monitoring module and a statistical processing module;
[0049] The front-end data acquisition module includes vibration acquisition module, infrared acquisition module, discharge monitoring module and spectrum analysis module;
[0050] The vibration acquisition module is used to collect vibration data to obtain mechanical vibration data, where the vibration data includes vibration intensity, vibration frequency and acquisition time;
[0051] The infrared monitoring module collects infrared images to obtain infrared imaging data. The discharge monitoring module is used to collect partial discharge and obtain partial discharge data. The spectrum analysis module performs spectrum analysis on the oil through the oil observation window and obtains oil detection data based on the spectrum analysis results. The oil detection data is the amount of gas dissolved in the oil.
[0052] The data analysis module performs data shielding on the mechanical vibration data to obtain the vibration net value data and data acquisition time. The method for the data analysis module to perform data shielding on the mechanical vibration data is as follows:
[0053] S1: Select vibration data and compare the vibration frequency in the vibration data with the preset frequency range, and eliminate the vibration data outside the preset frequency range.
[0054] S2: Select the vibration intensity and remove the vibration data with the vibration intensity less than the set intensity threshold;
[0055] S3: Record the vibration intensity in the remaining vibration data as vibration net value data, and retain the collection time;
[0056] The partial discharge data is collated to obtain the discharge intensity and duration, and the oil monitoring data and collection time are calculated to obtain the rate of change of the oil dissolution amount;
[0057] The data analysis module divides the infrared imaging data into images and identifies the temperature to obtain the temperature characterization value and temperature distribution area as follows:
[0058] A1: The data analysis module processes the infrared image using the non-local means algorithm to eliminate spatial noise;
[0059] A2: The data analysis module divides the transformer into several wet temperature zones and dry temperature zones according to the set conditions;
[0060] A3: The data analysis module performs weighted averages on the temperatures of the wet temperature zone and the dry temperature zone to obtain temperature representation values of the wet temperature zone and the dry temperature zone.
[0061] Specifically, in step A3, when the data analysis module calculates the temperature characterization value, it selects any wet temperature zone or dry temperature zone in step A2 and calculates the area of the selected area to obtain the total area S of the area. At the same time, the corresponding areas of different temperatures in the selected area are calculated with a gradient of 1°C to obtain the area Si corresponding to each temperature gradient, where i = 1, 2, 3, ... n. The temperature in the selected area is weighted and averaged using the corresponding area as the weight to obtain the temperature characterization value T. Where ti is the temperature gradient of area Si;
[0062] The data array creation module uses time as the horizontal axis, vibration intensity, discharge intensity, oil dissolution amount and temperature characterization values as the multivariate vertical axis, and the power transformer area as the normal axis to create a multi-axis image, and records the analysis results of the data analysis module in the multi-axis image;
[0063] When the data creation array is recorded in a multi-axis image, a vibration intensity change curve is plotted using time as the horizontal axis and vibration intensity as the vertical axis; a discharge intensity curve is plotted using time as the horizontal axis and discharge intensity as the vertical axis; a solubility change curve is plotted using time as the horizontal axis and oil dissolution amount as the vertical axis; and a regional temperature change curve is plotted using temperature representation value as the vertical axis, time as the horizontal axis, and region as the normal axis.
[0064] The random monitoring module randomly selects a point in the multi-axis image, obtains the coordinates of the horizontal axis and normal axis corresponding to the point, and obtains the corresponding vibration intensity, discharge intensity, oil dissolution amount, temperature zone and its corresponding temperature characterization value based on the horizontal axis coordinate;
[0065] The random monitoring module compares the vibration intensity, discharge intensity, temperature characterization value and oil dissolution amount corresponding to the acquired point with the set standard range. If the comparison results are all within the standard range, it is recorded as a normal operating point. If any comparison result is not within the standard range, it is recorded as a faulty operating point, thus obtaining the point operation result;
[0066] The statistical processing module records the points where the vibration intensity and oil dissolution amount are both outside the standard range as vibration interference points, and records the points where the temperature characterization value and discharge intensity are both outside the standard range as temperature interference points;
[0067] The statistical processing module calculates the proportion of faulty operation points, vibration interference points, and temperature interference points in the total number of points, obtains the proportion of fault points, the proportion of vibration interference points, and the proportion of temperature interference points, and performs threshold judgment on the calculation results. If the proportion of fault points is greater than the set threshold, an abnormal operation signal is generated. If the proportion of fault points is not greater than the set threshold, a normal operation signal is generated.
[0068] The statistical processing module calculates the ratio of the vibration interference point ratio to the fault point ratio. If the ratio is greater than the set threshold, a vibration interference anomaly is generated. If the ratio is not greater than the set threshold, no response is taken.
[0069] The statistical processing module calculates the ratio of the temperature interference point ratio to the fault point ratio. If the ratio is greater than the set threshold, a temperature interference anomaly is generated. If the ratio is not greater than the set threshold, no response is taken.
[0070] Thresholds, preset values, and preset ranges are set for comparative analysis of results to determine quality. The values are determined based on a combination of large-scale model analysis of sample data and manual experience, and can also be adjusted appropriately based on seasonal or common-sense factors.
[0071] The settings of weight ratio coefficients, influencing factors, etc. are assigned specific values according to the influence of each parameter on the result, which ultimately reflects the impact on the result. They are also set and entered into storage through a combination of large-scale model analysis of sample data and manual experience, and can also be appropriately adjusted based on seasonal or common-sense influencing conditions.
[0072] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.
Claims
1. A method for monitoring the condition of a power transformer based on multimodal data analysis, characterized in that: The following steps are involved: Step 1: Collect multi-source data on the operation data of the power transformer, obtain the mechanical vibration data, infrared imaging data, partial discharge data, and oil detection data of the power transformer, and organize the collected data to create a data array; Step 2: Perform amplitude-time analysis on the mechanical vibration in the data array, perform temperature gradient analysis on the infrared imaging data, perform electric field-time analysis on the partial discharge data, and perform solubility-time analysis on the oil test data to obtain data analysis results; Step 3: Create correlations for the data analysis results, obtain the power transformer monitoring image results, and output them; Step 4: Randomly select points from the power transformer monitoring image results, and obtain the point analysis results of the power transformer based on the selected data; Step 5: Statistically organize the multiple acquired point analysis results of the power transformer to obtain an evaluation of the power transformer's operating status.
2. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 1, characterized in that: It also includes a power transformer condition monitoring system, which specifically includes a front-end data acquisition module, a data array creation module, a data parsing module, a random monitoring module, and a statistical collation module; The front-end data acquisition module is used to collect the operating data of the power transformer to obtain mechanical vibration data, infrared imaging data, partial discharge data and oil detection data; The data analysis module performs data shielding on the mechanical vibration data to obtain vibration net value data and data acquisition time, performs screen segmentation and temperature recognition on the infrared imaging data to obtain temperature characterization values and temperature distribution areas, performs data sorting on the partial discharge data to obtain discharge intensity and discharge duration, and calculates the oil detection data and acquisition time to obtain the oil dissolution amount change rate; The data array creation module uses time as the horizontal axis, vibration intensity, discharge intensity, oil dissolution amount and temperature characterization values as the multivariate vertical axis, and the power transformer area as the normal axis to create a multi-axis image, and records the analysis results of the data analysis module in the multi-axis image; The random monitoring module is used to select points in the multi-axis image, and compare the selected point data with the set safe operation standards to obtain the point operation results; The statistical collating module integrates the point operation results to obtain a comprehensive operation result, and records it as the power transformer operation status evaluation.
3. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, characterized in that: The front-end data acquisition module includes a vibration acquisition module, an infrared acquisition module, a discharge monitoring module and a spectrum analysis module; The vibration acquisition module is used to collect vibration data, wherein the vibration data includes vibration intensity, vibration frequency and acquisition time; The infrared monitoring module collects infrared images, the discharge monitoring module is used to collect partial discharges, and the spectrum analysis module performs spectrum analysis on the oil through the oil observation window, and obtains oil detection data based on the spectrum analysis results. The oil detection data is the amount of gas dissolved in the oil.
4. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, wherein: When the data analysis module performs data shielding on mechanical vibration data, it selects vibration data and compares the vibration frequency in the vibration data with a preset frequency range, eliminates vibration data outside the preset frequency range, and then selects vibration intensity, eliminates vibration data with a vibration intensity less than a set intensity threshold, and records the vibration intensity in the remaining vibration data as vibration net value data, while retaining the collection time.
5. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, characterized in that: The data analysis module processes the infrared image through an algorithm to eliminate spatial noise. The data analysis module divides the transformer into several wet temperature zones and dry temperature zones according to the set conditions. The data analysis module performs weighted average on the temperatures of the wet temperature zones and the dry temperature zones to obtain temperature characterization values of the wet temperature zones and the dry temperature zones; When the data analysis module calculates the temperature characterization value, a wet temperature zone or a dry temperature zone is selected, and the area of the selected area is calculated. At the same time, the corresponding areas of different temperatures in the selected area are calculated with a gradient of 1°C, and the temperatures in the selected area are weighted averaged with the corresponding areas as weights to obtain the temperature characterization value.
6. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, characterized in that: When the data creation array is recorded in a multi-axis image, a vibration intensity change curve is drawn through the time horizontal axis and the vibration intensity vertical axis, a discharge intensity curve is drawn through the time horizontal axis and the discharge intensity vertical axis, a solubility change curve is drawn through the time horizontal axis and the oil dissolution amount vertical axis, and a regional temperature change curve is drawn through the temperature characterization value as the vertical axis, time as the horizontal axis, and the region as the normal axis.
7. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, characterized in that: The random monitoring module randomly selects a point in the multi-axis image, obtains the coordinates of the horizontal axis and the normal axis corresponding to the point, and obtains the corresponding vibration intensity, discharge intensity, oil dissolution amount, temperature zone and its corresponding temperature characterization value based on the horizontal axis coordinate; The random monitoring module compares the vibration intensity, discharge intensity, temperature characterization value and oil dissolution amount corresponding to the acquisition point with the set standard range. If the comparison results are all within the standard range, it is recorded as a normal operating point. If any comparison result is not within the standard range, it is recorded as a faulty operating point.
8. The method for monitoring the state of a power transformer based on multimodal data analysis according to claim 2, characterized in that: The statistical arrangement module records the points where the vibration intensity and the oil dissolution amount are both outside the standard range as vibration interference points, and records the points where the temperature characterization value and the discharge intensity are both outside the standard range as temperature interference points; The statistical collation module calculates the proportion of faulty operation points, vibration interference points and temperature interference points in the total number of points, and performs threshold judgment on the calculation results. It outputs an abnormal operation signal or a normal operation signal based on the judgment results, and outputs vibration interference abnormality and temperature interference abnormality signals at the same time.