A main transformer defect identification method based on intelligent management

Abstract

The application relates to a main transformer defect identification method based on intelligent management, and through a collaborative technical path of'small model edge preliminary screening + large model cloud precise judgment', a light-weight main transformer defect identification model is constructed on the edge side, real-time preliminary judgment of defects such as clamp insulation deterioration and core multi-point grounding is realized based on multi-dimensional data such as ground current full waveform and partial discharge atlas, and high-risk data is uploaded to the cloud; the cloud integrates dynamic and static data such as defect history library, load working condition and maintenance record, constructs a cloud multi-modal diagnosis model, then outputs a defect cause diagnosis report through correlation analysis and deep mining, generates a'man-machine-material-method' four-in-one operation and maintenance strategy, forms a'monitoring-diagnosis-decision-execution' closed-loop management chain, and the application has the advantages of intelligence, convenience, high precision and high efficiency.

Description

Technical Field

[0001] This invention belongs to the field of transformer defect detection technology, specifically relating to a method for identifying defects in main transformers based on intelligent management. Background Technology

[0002] As a core piece of equipment in the power grid, the main transformer's core clamps bear the dual functions of magnetic circuit conduction and mechanical support. If defects such as insulation deterioration and multiple grounding points are not identified and addressed in a timely manner, they will directly threaten the safety of the transformer itself and may even lead to large-scale power outages or equipment damage. According to statistics, there have been many transformer failures caused by core clamp defects in the past five years, with far-reaching social impacts. Especially during peak load periods or extreme weather conditions, such defects can quickly evolve into cascading failures such as equipment breakdown and abnormal oil temperature, seriously affecting the reliability of power grid supply. Therefore, achieving accurate identification and rapid handling of core clamp defects is a key link in ensuring the safe operation of main transformers and improving the resilience of the power grid. Currently, power grid companies face the following problems in the identification and maintenance management of defects in iron core clamps: 1. Traditional monitoring technologies rely on single-dimensional parameters (such as ground current threshold alarms and partial discharge peak monitoring), only issuing threshold alarms for specific parameters. They lack multi-parameter correlation analysis and comprehensive status assessment, resulting in a high misjudgment rate and frequent occurrences of "over-alarms" or "missed reports and oversights." 2. Traditional manual inspections rely on the experience and judgment of maintenance personnel, leading to problems such as strong subjectivity, low efficiency, and delayed defect location. 3. Maintenance decisions are mostly based on isolated data and experience, making it difficult to make scientific decisions by combining information such as equipment historical maintenance records and load conditions. This can easily lead to misjudgments of defect risk levels and insufficient maintenance (such as inadequate protection for high-risk equipment and excessive attention to low-risk equipment). Therefore, to solve the above problems, it is necessary to develop an intelligent, convenient, accurate, and efficient method for identifying defects in main transformers. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent, convenient, accurate and efficient method for identifying defects in main transformers.

[0004] The objective of this invention is achieved as follows: a method for identifying defects in main transformers based on intelligent management, comprising the following steps: Step 1: Collect grounding current and partial discharge data of transformer core clamps using a high-precision current sensor and partial discharge monitoring device, and preprocess and integrate the collected data to form the original database; Step 2: Based on the data within the original database, key features of current harmonic distortion rate and partial discharge pulse concentration are extracted through wavelet transform, and a lightweight main transformer defect identification model is constructed simultaneously. Step 3: By using a lightweight main transformer defect identification model and integrating feature data such as grounding current and partial discharge, the insulation defect risk of the iron core clamps is identified in real time. When an abnormal defect is identified, an alarm signal is issued. At the same time, the data for that time period is stored and backed up, and a defect identification report is generated. Then, the defect identification report and the corresponding abnormal data are sent to the IoT management platform via the MQTT protocol, laying a good data foundation for accurate cloud-based analysis. Step 4: Acquire heterogeneous data such as the defect history database, equipment dynamic operating conditions, and maintenance records. Clean and standardize the acquired historical data to achieve time series alignment and unit unification. Then, based on the processed historical data, construct a cloud-based multimodal diagnostic model. Next, use the cloud-based multimodal diagnostic model to mine the correlation of abnormal data uploaded to the IoT management platform, and input historical defect images to verify the accuracy of defect diagnosis and the rationality of defect diagnosis reports. Finally, output the final defect cause diagnosis report. Step 5: Utilize the cloud-based multimodal diagnostic model again and, based on the final defect cause diagnosis report, coordinate information across four dimensions: personnel, equipment, spare parts, and maintenance procedures. Output a four-in-one maintenance strategy encompassing "people, machines, materials, and methods," while verifying the rationality of this strategy. This will enable a closed-loop management chain of "monitoring, diagnosis, decision-making, and execution."

[0005] Furthermore, in step 1, the grounding current data of the transformer core clamp includes the dynamic waveforms of the total current and capacitive current, and the partial discharge data includes the discharge peak value and phase distribution spectrum.

[0006] Furthermore, the defect identification report generated in step 3 includes defect location, risk level, and defect cause analysis.

[0007] Furthermore, the equipment dynamic operating conditions in step 4 include real-time load rate, oil temperature and vibration data, and the operation and maintenance files include maintenance records, spare parts inventory and maintenance personnel information and qualifications.

[0008] Furthermore, the data preprocessing and integration operations in step 1 are specifically as follows: First, noise reduction and filtering are performed on the information data to remove abnormal information and error information generated during the transition state when the operating mode changes. Then, the information data is integrated, registered, and structured to increase the compactness and density of the data and reduce the amount of data transmitted.

[0009] The beneficial effects of this invention are as follows: This invention employs a collaborative technical approach of "small-scale edge screening + large-scale cloud-based precise judgment." At the edge, a lightweight main transformer defect identification model is constructed. Based on multi-dimensional data such as the full waveform of grounding current and partial discharge spectra, it achieves real-time preliminary judgment of defects such as clamp insulation degradation and multi-point grounding of the iron core, and uploads high-risk data to the cloud. The cloud then integrates dynamic and static data such as defect history databases, load conditions, and maintenance records to construct a cloud-based multimodal diagnostic model. Subsequently, through correlation analysis and deep mining, a defect cause diagnosis report is output, and a model is generated. The "Man-Machine-Material-Method" four-in-one operation and maintenance strategy forms a closed-loop management chain of "monitoring-diagnosis-decision-execution". This invention adopts this structure, using side-by-side models for real-time early warning and cloud-based precise diagnosis to reduce the false judgment rate, thereby enabling intelligent decision-making and shortening the defect handling cycle. Furthermore, by utilizing the "Man-Machine-Material-Method" four-in-one operation and maintenance strategy, personnel, equipment, and spare parts can be rationally allocated, improving resource utilization efficiency and avoiding resource idleness and waste, further achieving the goals of cost savings and increased revenue. In summary, this invention has the advantages of being intelligent, convenient, accurate, and efficient. Detailed Implementation

[0010] The present invention will now be further described.

[0011] Example: A method for identifying defects in a main transformer based on intelligent management, comprising the following steps: Step 1: Using a high-precision current sensor and partial discharge monitoring device, the grounding current and partial discharge data of the transformer core clamp are collected. The collected data is then preprocessed and integrated to form a raw database. The preprocessing and integration operations are as follows: First, the information data is filtered for noise reduction while removing abnormal information and error information generated during the transition state when the operating mode changes. Then, the information data is integrated, registered, and structured to increase the compactness and density of the data and reduce the amount of data transmission. The grounding current data of the transformer core clamp includes the dynamic waveforms of the total current and capacitive current, and the partial discharge data includes the discharge peak value and phase distribution spectrum. Step 2: Based on the data within the original database, key features of current harmonic distortion rate and partial discharge pulse concentration are extracted through wavelet transform, and a lightweight main transformer defect identification model is constructed simultaneously. Step 3: Using a lightweight main transformer defect identification model and integrating feature data such as grounding current and partial discharge, the insulation defect risk of the iron core clamps is identified in real time. When an abnormal defect is identified, an alarm signal is issued. The data for that time period is stored and backed up, and a defect identification report is generated. Then, the defect identification report and the corresponding abnormal data are sent to the IoT management platform via the MQTT protocol, laying a good data foundation for accurate cloud-based analysis. During this process, the defect identification report includes defect location, risk level, and defect cause analysis. Step 4: Acquire heterogeneous data such as the defect history database, equipment dynamic operating conditions, and maintenance records. Clean and standardize the acquired historical data to achieve time series alignment and unit unification. Equipment dynamic operating conditions include real-time load rate, oil temperature, and vibration data. Maintenance records include maintenance records, spare parts inventory, and maintenance personnel information and qualifications. Then, based on the processed historical data, construct a cloud-based multimodal diagnostic model. Next, use the cloud-based multimodal diagnostic model to mine the correlation of abnormal data uploaded to the IoT management platform, and input historical defect images to verify the accuracy of defect diagnosis and the rationality of defect diagnosis reports. Finally, output the final defect cause diagnosis report. Step 5: Utilize the cloud-based multimodal diagnostic model again and, based on the final defect cause diagnosis report, coordinate information across four dimensions: personnel, equipment, spare parts, and maintenance procedures. Output a four-in-one maintenance strategy encompassing "people, machines, materials, and methods," while verifying the rationality of this strategy. This will enable a closed-loop management chain of "monitoring, diagnosis, decision-making, and execution."

[0012] In use, this invention first collects dynamic waveforms of grounding current and partial discharge phase distribution data of transformer core clamps using a high-precision current sensor and a partial discharge monitoring device. The collected data is then preprocessed and integrated to form a raw database. Specifically, the preprocessing and integration involves: first, noise reduction and filtering of the data to remove abnormal information and errors generated during transitions in operating modes; then, data integration, registration, and structuring to increase data compactness and density while reducing data transmission volume; finally, based on the data within the raw database, wavelet transform is used to extract key features of current harmonic distortion rate and partial discharge pulse concentration. Simultaneously, a lightweight main transformer defect identification model is constructed. Then, using this model and integrating characteristic data such as grounding current and partial discharge, the insulation defect risk of the core clamps is identified in real time. When an abnormal defect is identified, an alarm signal is issued. Data for that time period is stored and backed up, and a defect identification report is generated. Subsequently, the defect identification report and corresponding abnormal data are sent to the IoT management platform via the MQTT protocol, laying a solid data foundation for accurate cloud-based analysis. During this process, the defect identification report includes defect location, risk level, and defect cause analysis. Finally, heterogeneous data such as the defect history database, equipment dynamic operating conditions, and maintenance records are acquired and analyzed. Historical data undergoes cleaning and standardization to achieve time-series alignment and unit unification. Equipment dynamic operating conditions include real-time load rate, oil temperature, and vibration data; maintenance records include repair logs, spare parts inventory, and maintenance personnel information and qualifications. After these operations, a cloud-based multimodal diagnostic model is constructed based on the processed historical data. This model is then used to analyze the correlations of abnormal data uploaded to the IoT management platform. Historical defect images are input to verify the accuracy of defect diagnosis and the rationality of the defect diagnosis report. Finally, a final defect cause diagnosis report is output. Finally, the cloud-based multimodal diagnostic model is used again, and based on the final defect cause diagnosis report, personnel, equipment, spare parts, and maintenance procedures are coordinated. This invention provides four-dimensional information, outputting a four-in-one operation and maintenance strategy encompassing "human-machine-material-method," while simultaneously verifying the rationality of this strategy. This achieves a closed-loop management chain of "monitoring-diagnosis-decision-execution." Through a collaborative technical approach of "small-model edge screening + large-model cloud-based precise judgment," the invention utilizes real-time early warning from edge models and precise cloud-based diagnosis to reduce misjudgment rates, thereby enabling intelligent decision-making and shortening the defect handling cycle. Furthermore, the four-in-one operation and maintenance strategy allows for the rational allocation of personnel, equipment, and spare parts, improving resource utilization efficiency, avoiding resource idleness and waste, and further achieving cost savings and increased revenue. Overall, this invention possesses the advantages of being intelligent, convenient, accurate, and efficient.

[0013] The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying defects in main transformers based on intelligent management, characterized in that, Includes the following steps: Step 1: Collect grounding current and partial discharge data of transformer core clamps using a high-precision current sensor and partial discharge monitoring device, and preprocess and integrate the collected data to form the original database; Step 2: Based on the data within the original database, key features of current harmonic distortion rate and partial discharge pulse concentration are extracted through wavelet transform, and a lightweight main transformer defect identification model is constructed simultaneously. Step 3: By using a lightweight main transformer defect identification model and integrating feature data such as grounding current and partial discharge, the insulation defect risk of the iron core clamps is identified in real time. When an abnormal defect is identified, an alarm signal is issued. At the same time, the data for that time period is stored and backed up, and a defect identification report is generated. Then, the defect identification report and the corresponding abnormal data are sent to the IoT management platform via the MQTT protocol, laying a good data foundation for accurate cloud-based analysis. Step 4: Acquire heterogeneous data such as the defect history database, equipment dynamic operating conditions, and maintenance records. Clean and standardize the acquired historical data to achieve time series alignment and unit unification. Then, based on the processed historical data, construct a cloud-based multimodal diagnostic model. Next, use the cloud-based multimodal diagnostic model to mine the correlation of abnormal data uploaded to the IoT management platform, and input historical defect images to verify the accuracy of defect diagnosis and the rationality of defect diagnosis reports. Finally, output the final defect cause diagnosis report. Step 5: Utilize the cloud-based multimodal diagnostic model again and, based on the final defect cause diagnosis report, coordinate information across four dimensions: personnel, equipment, spare parts, and maintenance procedures. Output a four-in-one maintenance strategy encompassing "people, machines, materials, and methods," while verifying the rationality of this strategy. This will enable a closed-loop management chain of "monitoring, diagnosis, decision-making, and execution." 2. The main transformer defect identification method based on intelligent management as described in claim 1, characterized in that: In step 1, the grounding current data of the transformer core clamp includes the dynamic waveforms of the total current and capacitive current, and the partial discharge data includes the discharge peak value and phase distribution spectrum.

3. The main transformer defect identification method based on intelligent management as described in claim 1, characterized in that: The defect identification report generated in step 3 includes defect location, risk level, and defect cause analysis.

4. The main transformer defect identification method based on intelligent management as described in claim 1, characterized in that: The equipment dynamic operating conditions in step 4 include real-time load rate, oil temperature and vibration data, and the operation and maintenance files include maintenance records, spare parts inventory and maintenance personnel information and qualifications.

5. The main transformer defect identification method based on intelligent management as described in claim 1, characterized in that: The data preprocessing and integration operations in step 1 are as follows: First, the information data is filtered to reduce noise, while removing abnormal information and error information generated during the transition state when the operating mode changes. Then, the information data is integrated, registered, and structured to increase the compactness and density of the data and reduce the amount of data transmitted.