Method and device for monitoring operation state of dry-type transformer
The method improves dry-type transformer monitoring by installing backup sensors and using adaptive filtering and LSTM networks for real-time electromagnetic interference analysis, ensuring early fault detection and enhanced safety.
Patent Information
- Application Number
- CN202510308593.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-17
- Publication Date
- 2025-07-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing dry transformer operating status monitoring methods lack real-time and comprehensiveness, and cannot effectively reflect the electromagnetic interference of the transformer under different working conditions, resulting in the failure to detect chronic faults in a timely manner. Traditional fault diagnosis algorithms lack deep learning capabilities and are prone to ignore low-probability and high-risk faults.
Install multiple electromagnetic interference sensors at key locations of the dry transformer, configure backup sensors, and use adaptive filtering algorithms and cyclic neural network model built by LSTM units to monitor and evaluate the operating status of the transformer in real time through frequency and time domain analysis, abnormal detection and classification algorithms of electromagnetic interference data.
Real-time and comprehensive monitoring of electromagnetic interference is achieved, data accuracy and reliability are improved, dynamic change capture capability of electromagnetic interference is enhanced, and the safety and reliability of transformers are ensured, and data loss and blind spots in traditional methods are avoided.
Smart Images
Figure CN120316664A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power equipment monitoring, and more particularly to a method and device for monitoring the operating state of a dry-type transformer. Background Art
[0002] With the rapid development of the power system, transformers, as important power equipment, are widely used in various power projects. Due to their advantages such as pollution-free, fire-resistant, and low-noise, dry-type transformers have gradually become one of the main options in the power system. To ensure the stable operation and extend the service life of transformers, it is usually necessary to monitor them in real time.
[0003] The existing technologies have the following deficiencies: The existing methods for monitoring the operating state of dry-type transformers mainly rely on regular manual inspections or basic electrical parameter monitoring. Although they can provide certain operating state information, they often lack real-time and comprehensiveness, and cannot effectively reflect the electromagnetic interference situation of transformers under different working conditions, resulting in monitoring blind spots. Most current monitoring methods still focus on post-fault responses, lacking dynamic analysis of the long-term operating state of transformers. For those chronic and progressive faults, their development trends cannot be detected in advance, leading to failure to intervene and repair in time before the occurrence of faults, seriously affecting the operating safety of transformers. The data processing and analysis functions are still relatively simple, unable to effectively integrate different types of monitoring data for multi-dimensional analysis. The fault diagnosis algorithms of existing monitoring systems are relatively traditional, mainly relying on simple threshold setting and comparison methods, lacking the ability of deep learning and autonomous judgment of abnormal events, making some low-probability and high-risk faults easily overlooked.
[0004] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention
[0005] The object of the present invention is to provide a method and device for monitoring the operating state of a dry-type transformer to solve the problems in the above background art.
[0006] To achieve the above object, the present invention provides the following technical solution: A method for monitoring the operating state of a dry-type transformer, specifically including the following steps: Step 1: Independently install a plurality of electromagnetic interference sensors at different positions of the dry-type transformer, and configure backup and communication protocols to monitor and integrate the electromagnetic interference data in each area in real time; Step 2: Use an adaptive filtering algorithm to dynamically filter electromagnetic interference according to the electromagnetic interference data, and perform frequency-domain and time-domain analysis on the electromagnetic interference data; Step 3: Apply a classification algorithm to perform anomaly detection on the electromagnetic interference data that has undergone adaptive dynamic filtering, and identify the characteristic differences between normal signals and electromagnetic interference data; Step 4: Through a recurrent neural network model built by LSTM units, evaluate the overall operating state of the dry-type transformer based on the electromagnetic interference data and its category labels at the current position; Preferably, in Step 1, add the top position label of the dry-type transformer, the side position label near the terminal, and the ground connection point position label. Independently install multiple electromagnetic interference sensors at the position labels to monitor the electromagnetic interference data in real time. Among them, the electromagnetic interference sensor at the top position label is used to monitor the electromagnetic interference situation in the top area, the electromagnetic interference sensor at the side position label near the terminal is used to monitor the electromagnetic interference situation in the wiring connection area, and the electromagnetic interference sensor at the ground connection point position label is used to monitor the electromagnetic interference situation in the grounding area. Configure a communication protocol for multi-channel electromagnetic interference data transmission. Set backup sensors for each position label to ensure the integrity of the electromagnetic interference data even when the electromagnetic interference sensor at a certain position fails. Send the electromagnetic interference data to the control center every five minutes, and distinguish the monitoring timestamps and ID fields of the electromagnetic interference data at different label positions through data integration.
[0007] Preferably, in Step 2, the control center detects the missing values of the electromagnetic interference data and deletes the corresponding missing value records. Select the electromagnetic interference data that exceeds the mean plus or minus 2 times the standard deviation as an outlier and replace it with the mean. Detect and delete the duplicate records at the corresponding label positions of the electromagnetic interference data. Use normalization to shrink and map the electromagnetic interference data to the same scale. Initialize the weight vector and covariance matrix of the recursive least squares adaptive filter, input the electromagnetic interference data and reference data for learning and updating. Create a column vector containing the current and past input data, and output the electromagnetic interference data that has undergone adaptive dynamic filtering based on the dot product of the weight vector and the input electromagnetic interference data vector. The specific formula is:
[0008] where, represents the electromagnetic interference data that has undergone adaptive dynamic filtering, represents the transpose of the weight vector at time step n−1, represents the input electromagnetic interference data. Divide the electromagnetic interference data that has undergone adaptive dynamic filtering into short time segments through a Hann window, slide the Hann window, and apply the Fourier transform to each window to obtain the spectral characteristics of the electromagnetic interference data that has undergone adaptive dynamic filtering. Combine the spectral characteristics of each time window to form a time-frequency diagram that reflects both time and frequency information.
[0009] Preferably, in the third step, the Euclidean method is used to calculate the distance between the electromagnetic interference data at the current position and the nearest target point representing the determined data category. The specific formula is as follows:
[0010] where represents the coordinates of the electromagnetic interference data at the current position, represents the coordinates of the nearest target point, represents the distance between the electromagnetic interference data at the current position and the nearest target point. The distance between the electromagnetic interference data at the current position and the nearest target point and the number of point positions are used as eigenvalue. According to the eigenvalue, the K nearest neighbor electromagnetic interference data to the target point are selected. According to the category labels of the K neighbors, the category of the nearest target point representing the determined data category is selected by means of majority voting. The specific voting formula is as follows:
[0011] where represents the category of the nearest target point representing the determined data category, represents all possible categories of the nearest target point representing the determined data category, represents the indicator function, specifically the category label of the i-th neighbor, is equal to the current category , then the value of this indicator function is 1, otherwise it is 0, and the category label of the electromagnetic interference data at the current position after voting is returned.
[0012] Preferably, in the fourth step, an LSTM unit is selected to build a recurrent neural network model. According to the electromagnetic interference data at the current position, the corresponding category label and the hidden state of the previous time window, the activation value of the input layer is calculated through the Sigmoid activation function. The specific formula is as follows:
[0013] where represents the activation value of the input layer, represents the Sigmoid activation function, represents the weight matrix, represents the hidden state of the previous time window, represents the electromagnetic interference data at the current position and the corresponding category label, represents the bias term. The output layer selects a fully connected layer and calculates the activation value of the output layer through the Sigmoid activation function according to the electromagnetic interference data at the current position, the corresponding category label, the hidden state of the previous time stamp and the cell state of the electromagnetic interference data at the current position. The specific formula is as follows:
[0014] where represents the activation value of the forgetting layer, represents the Sigmoid activation function, represents the weight matrix, represents the hidden state of the previous time window, represents the electromagnetic interference data at the current position, represents the bias term, and outputs the overall operating state of the current dry-type transformer through the fully connected layer.
[0015] Preferably, a monitoring device for the operating state of a dry-type transformer specifically includes: 1. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the top area; 2. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the wiring connection area; 3. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the grounding area; 4. A backup sensor for ensuring the integrity of electromagnetic interference data even when an electromagnetic interference sensor at a certain position fails; 5. A control center for receiving and integrating and distinguishing electromagnetic interference data.
[0016] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. By installing multiple electromagnetic interference sensors at key positions of the dry-type transformer, the electromagnetic interference situation in different areas can be monitored in real time, effectively improving the comprehensiveness and accuracy of the data, making the evaluation of electromagnetic interference more scientific and reliable. By configuring backup sensors at each position, even if a certain sensor fails, the integrity of the data can be ensured, avoiding data loss caused by single-point failures. The electromagnetic interference data is sent to the control center every five minutes, ensuring the real-time nature of the data, facilitating quick response and handling of potential electromagnetic interference problems. By detecting missing values and outliers through the control center, the data quality is ensured, abnormal values are replaced by the mean and duplicate records are deleted. Through recursive least squares adaptive filtering, the weights can be learned and updated in real time, dynamically adjusting the response to electromagnetic interference data, thereby improving the reliability and accuracy of the data.
[0017] 2. Calculate the distance between the current position electromagnetic interference data and the nearest target point using the Euclidean distance. Through the K-nearest neighbor algorithm for classification decision-making, it can effectively utilize the information of neighboring data to improve the classification accuracy. Through the majority voting mechanism, it can maintain high robustness in the face of noise and outliers, ensuring the reliability of the final decision. Select LSTM units to build a recurrent neural network model. By leveraging the characteristics of time series data, it can better capture the dynamic changes of electromagnetic interference data, can remember long-term historical information, and avoid the gradient vanishing problem of traditional RNNs in processing long sequence data. By calculating the activation values of the input layer and output layer through the Sigmoid activation function and combining with the hidden state of the previous time window, the system can accurately evaluate the overall operating state of the dry-type transformer, not only improving the monitoring accuracy of electromagnetic interference but also enhancing the safety and reliability of the dry-type transformer. Brief Description of the Drawings
[0018] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.
[0019] Figure 1 It is a method flow chart of a method for monitoring the operating state of a dry-type transformer according to the present invention. Detailed Embodiments
[0020] Now, the exemplary embodiments will be described more comprehensively with reference to the drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art.
[0021] The present invention provides a method for monitoring the operating state of a dry-type transformer as shown in Figure 1 and specifically includes the following steps: Step 1: Independently install multiple electromagnetic interference sensors at different positions of the dry-type transformer and configure backup and communication protocols to monitor and integrate the electromagnetic interference data in each area in real time; Add top position labels, side position labels near the terminal, and ground point position labels to the dry-type transformer. Independently install multiple electromagnetic interference sensors at the position labels to monitor electromagnetic interference data in real time. Among them, the electromagnetic interference sensors at the top position label are used to monitor the electromagnetic interference in the top area, the electromagnetic interference sensors at the side position label near the terminal are used to monitor the electromagnetic interference in the wiring connection area, and the electromagnetic interference sensors at the ground point position label are used to monitor the electromagnetic interference in the grounding area. Configure a communication protocol for multi-channel electromagnetic interference data transmission. Set up backup sensors for each position label to ensure the integrity of electromagnetic interference data when the electromagnetic interference sensor at a certain position fails. Send the electromagnetic interference data to the control center every five minutes, and distinguish the monitoring timestamps and ID fields of electromagnetic interference data at different label positions through data integration.
[0022] Step 2: Dynamically filter electromagnetic interference according to the electromagnetic interference data using an adaptive filtering algorithm, and perform frequency-domain and time-domain analysis on the electromagnetic interference data; Detect missing values of the electromagnetic interference data through the control center, and delete the corresponding missing value records. Select the electromagnetic interference data that exceeds the mean plus or minus 2 standard deviations as outlier anomalies and replace them with the mean. Detect and delete duplicate records at the corresponding label positions of the electromagnetic interference data. Use normalization to shrink and map the electromagnetic interference data to the same scale. Initialize the weight vector and covariance matrix of the recursive least squares adaptive filter. Input the electromagnetic interference data and reference data for learning and updating. Create a column vector containing the current and past input data. Output the adaptively dynamically filtered electromagnetic interference data based on the dot product of the weight vector and the input electromagnetic interference data vector. The specific formula is:
[0023] Among them, represents the adaptively dynamically filtered electromagnetic interference data, represents the transpose of the weight vector at time step n−1, represents the input electromagnetic interference data. Divide the adaptively dynamically filtered electromagnetic interference data into short-time segments through a Hann window. Slide the Hann window and apply the Fourier transform to each window to obtain the spectral characteristics of the adaptively dynamically filtered electromagnetic interference data. Combine the spectral characteristics of each time window to form a time-frequency diagram that reflects both time and frequency information.
[0024] Step 3: Apply a classification algorithm to perform anomaly detection on the adaptively dynamically filtered electromagnetic interference data and identify the characteristic differences between normal signals and electromagnetic interference data; Use Euclidean distance to calculate the distance between the current position electromagnetic interference data and the nearest target point representing the determined data class. The specific formula is:
[0025] Among them, represents the coordinates of the electromagnetic interference data at the current position, represents the coordinates of the nearest target point, represents the distance between the electromagnetic interference data at the current position and the nearest target point. The distance between the electromagnetic interference data at the current position and the nearest target point and the number of point positions are used as feature values. According to the feature values, K nearest neighbor electromagnetic interference data closest to the target point are selected. According to the class labels of the K neighbors, the majority voting method is used to select the class of the nearest target point representing the decision of the data class. The specific voting formula is:
[0026] Among them, represents the class of the nearest target point that determines the data class, represents all possible classes of the nearest target point that determines the data class, represents the indicator function, specifically the class label of the i-th neighbor is equal to the current class , then the value of this indicator function is 1, otherwise it is 0, and the class label of the electromagnetic interference data at the current position after voting is returned.
[0027] Step 4: Through the recurrent neural network model built by LSTM units, evaluate the overall operating state of the dry-type transformer according to the electromagnetic interference data and its class label at the current position; Select the recurrent neural network model built by LSTM units. According to the electromagnetic interference data and the corresponding class label at the current position and the hidden state of the previous time window, calculate the activation value of the input layer through the Sigmoid activation function. The specific formula is:
[0028] Among them, represents the activation value of the input layer, represents the Sigmoid activation function, represents the weight matrix, represents the hidden state of the previous time window, represents the electromagnetic interference data and the corresponding class label at the current position, represents the bias term. The output layer selects a fully connected layer and calculates the activation value of the output layer through the Sigmoid activation function according to the electromagnetic interference data and the corresponding class label at the current position, the hidden state of the previous time stamp, and the cell state of the electromagnetic interference data at the current position. The specific formula is:
[0029] Among them, Represents the activation value of the forgetting layer, Represents the Sigmoid activation function, Represents the weight matrix, Represents the hidden state of the previous time window, Represents the electromagnetic interference data at the current position, Represents the bias term, and outputs the overall operating state of the current dry-type transformer through the fully connected layer.
[0030] The present invention provides a monitoring device for the operating state of a dry-type transformer, which specifically includes: 1. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the top area; 2. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the wiring connection area; 3. An electromagnetic interference sensor for monitoring the electromagnetic interference situation in the grounding area; 4. A backup sensor for ensuring the integrity of electromagnetic interference data even when an electromagnetic interference sensor at a certain position fails; 5. A control center for receiving and integrating and differentiating electromagnetic interference data.
[0031] By installing multiple electromagnetic interference sensors at key positions of the dry-type transformer, the electromagnetic interference situation in different areas can be monitored in real time, effectively improving the comprehensiveness and accuracy of the data, making the evaluation of electromagnetic interference more scientific and reliable. By configuring backup sensors at each position, even if a sensor fails, the integrity of the data can be guaranteed, avoiding data loss caused by single-point failures. The electromagnetic interference data is sent to the control center every five minutes, ensuring the real-time nature of the data, facilitating quick response and handling of potential electromagnetic interference problems. By the control center detecting missing values and outliers, the data quality is ensured, using the mean to replace outliers and deleting duplicate records. Through recursive least squares adaptive filtering, the weights can be learned and updated in real time, dynamically adjusting the response to electromagnetic interference data, thereby improving the reliability and accuracy of the data.
[0032] The Euclidean distance is used to calculate the distance between the current position electromagnetic interference data and the nearest target point. Through the K-nearest neighbor algorithm for classification decision-making, the information of neighboring data can be effectively utilized to improve the classification accuracy. Through the majority voting mechanism, it can maintain high robustness in the face of noise and outliers, ensuring the reliability of the final decision. An LSTM unit is selected to build a recurrent neural network model. Utilizing the characteristics of time series data, it can better capture the dynamic changes of electromagnetic interference data, can remember long-term historical information, and avoids the problem of gradient disappearance in the traditional RNN when dealing with long sequence data. By calculating the activation values of the input layer and the output layer through the Sigmoid activation function and combining the hidden state of the previous time window, the system can accurately evaluate the overall operating state of the dry-type transformer, not only improving the monitoring accuracy of electromagnetic interference, but also enhancing the safety and reliability of the dry-type transformer.
[0033] Only some exemplary embodiments of the present invention have been described above by way of illustration. Undoubtedly, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A method for monitoring the operating state of a dry-type transformer, characterized in that, It includes the following steps: Step 1: Independently install multiple electromagnetic interference sensors at different positions of the dry-type transformer, and configure backup and communication protocols to monitor and integrate the electromagnetic interference data in each area in real time; Step 2: Use an adaptive filtering algorithm to dynamically filter electromagnetic interference according to the electromagnetic interference data, and perform frequency-domain and time-domain analysis on the electromagnetic interference data; Step 3: Apply a classification algorithm to perform anomaly detection on the electromagnetic interference data that has been adaptively dynamically filtered, and identify the characteristic differences between normal signals and electromagnetic interference data; Step 4: Through a recurrent neural network model built by an LSTM unit, evaluate the overall operating state of the dry-type transformer according to the electromagnetic interference data and its class label at the current position.
2. The operating state monitoring method of a dry-type transformer according to claim 1, characterized in that: In the said Step 1, independently installing multiple electromagnetic interference sensors at different positions of the dry-type transformer specifically includes: the electromagnetic interference sensor at the top position label is used to monitor the electromagnetic interference situation in the top area, the electromagnetic interference sensor at the side position label near the terminal is used to monitor the electromagnetic interference situation in the wiring connection area, and the electromagnetic interference sensor at the ground connection point position label is used to monitor the electromagnetic interference situation in the grounding area.
3. The operating state monitoring method of a dry-type transformer according to claim 1, characterized in that: In the said Step 2, the specific steps of using an adaptive filtering algorithm to dynamically filter electromagnetic interference according to the electromagnetic interference data include: initializing the weight vector and covariance matrix of the recursive least squares adaptive filter, inputting the electromagnetic interference data and reference data for learning and updating, creating a column vector containing the current and past input data, and outputting the electromagnetic interference data that has been adaptively dynamically filtered based on the dot product of the weight vector and the input electromagnetic interference data vector. The specific formula is: ； Among them, represents the electromagnetic interference data after adaptive dynamic filtering, represents the transpose of the weight vector at time step n−1, represents the input electromagnetic interference data.
4. The operating state monitoring method of a dry-type transformer according to claim 1, characterized in that: In the said Step 2, the specific steps of performing frequency-domain and time-domain analysis on the electromagnetic interference data include: dividing the electromagnetic interference data that has been adaptively dynamically filtered into short-time segments through a Hanning window, sliding the Hanning window, applying the Fourier transform to each window to obtain the spectral characteristics of the electromagnetic interference data that has been adaptively dynamically filtered, and combining the spectral characteristics of each time window to form a time-frequency diagram that reflects both time and frequency information.
5. A method for monitoring the operating state of a dry-type transformer according to claim 1, characterized in that: In the said Step 3, the specific steps of applying a classification algorithm to perform anomaly detection on the electromagnetic interference data that has been adaptively dynamically filtered include: using Euclidean distance to calculate the distance between the electromagnetic interference data at the current position and the nearest target point representing the determined data class, taking the distance between the electromagnetic interference data at the current position and the nearest target point and the number of points as eigenvalues, selecting the K nearest neighbor electromagnetic interference data based on the eigenvalues, and according to the class labels of the K neighbors, using the majority voting method to select the class of the nearest target point representing the determined data class, and returning the class label of the electromagnetic interference data at the current position after voting.
6. A method for monitoring the operating state of a dry-type transformer according to claim 5, characterized in that: The specific formula for using Euclidean distance to calculate the distance between the electromagnetic interference data at the current position and the nearest target point representing the determined data class is: ； Among them, represents the coordinates of the electromagnetic interference data at the current position, represents the coordinates of the nearest target point, represents the distance between the electromagnetic interference data at the current position and the nearest target point.
7. A method for monitoring the operating state of a dry-type transformer according to claim 1, characterized in that: The formula for using the majority voting method to select the class of the nearest target point representing the determined data class is: ； Among them, represents the category of the nearest target point that determines the data category, represents all possible categories of the nearest target point that determines the data category, represents an indicator function, specifically the category label of the i-th neighbor is equal to the current category , then the value of this indicator function is 1; otherwise it is 0.
8. A method for monitoring the operating state of a dry-type transformer according to claim 1, characterized in that: In the fourth step, the specific steps of the recurrent neural network model built by the LSTM unit include: according to the electromagnetic interference data and corresponding category labels at the current position and the hidden state of the previous time window, calculate the activation value of the input layer through the Sigmoid activation function, select a fully connected layer for the output layer, and calculate the activation value of the output layer through the Sigmoid activation function according to the electromagnetic interference data and corresponding category labels at the current position, the hidden state of the previous time stamp, and the cell state of the electromagnetic interference data at the current position, and output the overall operating state of the current dry-type transformer through the fully connected layer.
9. A method for monitoring the operating state of a dry-type transformer according to claim 1, characterized in that: The specific formula for calculating the activation value of the input layer through the Sigmoid activation function is: ； Among them, represents the activation value of the input layer, represents the Sigmoid activation function, represents the weight matrix, represents the hidden state of the previous time window, represents the electromagnetic interference data and corresponding class labels at the current position, represents the bias term. The specific formula for calculating the activation value of the output layer through the Sigmoid activation function is: ； Among them, represents the activation value of the forgetting layer, represents the Sigmoid activation function, represents the weight matrix, represents the hidden state of the previous time window, represents the electromagnetic interference data at the current position, represents the bias term.
10. An operating state monitoring device for a dry-type transformer, which is used to implement the operating state monitoring method of a dry-type transformer described in any one of the above claims 1-9, and is characterized in that, Specifically, it includes: an electromagnetic interference sensor for monitoring the electromagnetic interference situation in the top area; an electromagnetic interference sensor for monitoring the electromagnetic interference situation in the wiring connection area; an electromagnetic interference sensor for monitoring the electromagnetic interference situation in the grounding area; a backup sensor for preventing the electromagnetic interference data from being incomplete when a certain electromagnetic interference sensor fails; and a control center for receiving and integrating and distinguishing the electromagnetic interference data.