A partial discharge detection method based on UV radiation
Through UV radiation sensors and machine learning models, identifying and predicting the local discharge location of power system equipment, the problems of blind spots and misjudgment in the prior art are solved, and higher detection accuracy and real-time performance are achieved.
Patent Information
- Application Number
- CN202311686402.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-08
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2043-12-08
AI Technical Summary
The prior art has blind spots and misjudgment problems when detecting local discharges, especially when concealed parts or weak signals, it is difficult to accurately identify discharge locations and events.
UV radiation sensors are used to obtain UV radiation changes in discharge positions in power system equipment, and discharge events are identified and predicted through DBSCAN clustering analysis and machine learning model training.
It improves the detection ability of hidden parts, reduces the misjudgment rate, achieves higher recognition accuracy and real-time, and has stronger scalability.
Smart Images

Figure CN117907759B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the field of partial discharge detection, and in particular to a partial discharge detection method based on UV radiation. Background Art
[0002] Partial discharge refers to the phenomenon that high-voltage electrical equipment such as transformers discharge locally under the action of an electric field. Partial discharge can cause heating of electrical equipment parts and components, accelerating the aging process. The products generated by the heating of electrical equipment parts and components may also cause secondary damage to the electrical equipment. In addition, long-term partial discharge may even cause electrical equipment parts to be broken down. Therefore, effective detection of partial discharge is of great significance.
[0003] Existing technologies all use parameters such as voltage, current, and temperature to detect partial discharges, and identify partial discharge phenomena by monitoring the electrical and thermal characteristics of the equipment. For example, voltage and current monitoring is to monitor the voltage and current waveforms of the equipment to find characteristic signals generated by partial discharges, such as pulse discharges and periodic discharges. Temperature monitoring is because partial discharges are accompanied by local temperature rises, so the occurrence of partial discharges can be identified by monitoring the temperature changes on the surface or local areas of the equipment. In addition to voltage, current, and temperature, other parameters such as sound waves and infrared radiation can also be monitored to identify the occurrence of partial discharges. However, this technical solution has certain limitations and may not cover all partial discharge situations, especially for hidden areas or situations where the partial discharge signal is weak. At the same time, since there may be other factors in the operation of the equipment that cause fluctuations in parameters such as voltage, current, and temperature, it may lead to misjudgment of partial discharges. Summary of the invention
[0004] In order to overcome the shortcomings of the prior art, the present invention provides a partial discharge detection method based on UV radiation. The method uses a UV radiation sensor to obtain UV radiation change information at the discharge position in the electromechanical equipment of the power system, and realizes the recognition and prediction of discharge events through cluster analysis and machine learning model training.
[0005] The technical solution of the present invention is as follows:
[0006] The present invention provides a partial discharge detection method based on UV radiation, comprising: screening out discharge detection equipment in electromechanical equipment of an electric power system, and installing a UV radiation sensor to obtain change information of UV radiation at a discharge position; clustering UV radiation data using DBSCAN based on a discharge detection equipment model and a corresponding UV radiation sensor position, identifying a UV radiation area, and constructing a UV radiation area-power value-UV radiation value relationship map; collecting UV radiation data around a possible discharge position determined by cluster analysis as a training data set, and recording labels of discharge events, i.e., normal and abnormal states, according to the relationship map; extracting features of UV radiation data from the training data set using a time series feature method as input, training a machine learning model, and using label information to guide the model to learn UV radiation features of normal and abnormal states; and after evaluating the model using cross-validation, using the trained machine learning model to predict UV radiation data collected in real time, and identifying discharge positions and events.
[0007] Further, in some possible implementations, the S200 includes:
[0008] S210: Establish a discharge detection equipment model based on the characteristics of the power system and the type of equipment;
[0009] S220: Using DBSCAN to identify UV radiation areas according to the density of UV radiation data points.
[0010] Further, in some possible implementations, the S300 includes:
[0011] S310: determining a possible discharge position according to the UV radiation area obtained by clustering the UV radiation data using DBSCAN in S200;
[0012] S320: using the constructed UV radiation area-power value-UV radiation value relationship map, the collected UV radiation data points are associated with the corresponding UV radiation areas, and the labels of the discharge events are marked.
[0013] Further, in some possible implementations, the S400 includes:
[0014] S410: extracting features of UV radiation data from the training data set as input;
[0015] S420: Associating UV radiation data in the training data set with discharge event labels;
[0016] S430: using the extracted UV radiation features as input and corresponding label information to train a machine learning model;
[0017] S440: When training the machine learning model, use the label information recorded in step S300 to guide the model to learn the UV radiation characteristics of normal and abnormal states.
[0018] Further, in some possible implementations, the S500 includes:
[0019] S510: Cross-validate the trained machine learning model;
[0020] S520: using the trained machine learning model to predict UV radiation data collected in real time;
[0021] S530: Based on the prediction results of the model, possible discharge locations and events are identified.
[0022] The beneficial effects of the present invention are:
[0023] 1) Stronger detection capability for hidden parts: UV radiation can penetrate insulating materials of a certain thickness, so it can detect discharge conditions in hidden parts, avoiding the blind spots that may exist when using parameters such as voltage, current, and temperature for partial discharge detection.
[0024] 2) Higher recognition accuracy: This method can accurately identify discharge locations and events through cluster analysis and machine learning model training, avoiding possible misjudgments when using parameters such as voltage, current, and temperature for partial discharge detection.
[0025] 3) Higher real-time performance: This method has relatively low real-time requirements because the UV radiation sensor can collect and store data in real time for subsequent processing and analysis.
[0026] 4) Stronger scalability: This method can establish relationship graphs and machine learning models based on the UV radiation characteristics of different devices, and has stronger scalability and applicability. BRIEF DESCRIPTION OF THE DRAWINGS
[0027] Figure 1 It is a schematic diagram of the flow of the partial discharge detection method based on UV radiation;
[0028] Figure 2 It is a schematic diagram of the process of partial discharge detection method S200 based on UV radiation;
[0029] Figure 3 It is a schematic diagram of the process of partial discharge detection method S300 based on UV radiation;
[0030] Figure 4 It is a schematic diagram of the process of partial discharge detection method S400 based on UV radiation;
[0031] Figure 5It is a schematic diagram of the process of partial discharge detection method S500 based on UV radiation. DETAILED DESCRIPTION
[0032] The technical solution of the present invention is further described below with reference to the accompanying drawings and through specific embodiments:
[0033] Example 1
[0034] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0035] The following are exemplary UV radiation based discharge detection methods:
[0036] The overall steps are as follows Figure 1 Shown: Included
[0037] S100: Screen out the electromechanical equipment in the power system that needs to be tested for discharge, and install UV radiation sensors to obtain information on changes in UV radiation at the discharge location.
[0038] In this step, by installing a UV radiation sensor, the change information of UV radiation at the discharge position can be accurately obtained, which helps to accurately locate the discharge position. The UV radiation sensor can provide real-time radiation data, so that the monitoring of the discharge event can be responded to in a timely manner.
[0039] First, the electromechanical equipment in the power system is screened to determine which equipment needs to be discharged. This can be based on factors such as equipment type, historical fault conditions, and importance.
[0040] Install UV radiation sensors around the equipment that needs to be detected for discharge. These sensors can be installed on the surface of the equipment or near it to capture the changes in UV radiation at the discharge location.
[0041] For example, a discharge test needs to be performed on a generator. First, the generator is screened to determine whether it needs to be tested for discharge. Then, UV radiation sensors are installed around the generator to obtain information about changes in UV radiation at the discharge location. Once a discharge event occurs, the UV radiation sensor can capture the corresponding UV radiation changes and help accurately locate the discharge location.
[0042] S200: Based on the discharge detection equipment model and the corresponding UV radiation sensor position, DBSCAN is used to cluster the UV radiation data, identify the UV radiation area, and construct a UV radiation area-power value-UV radiation value relationship map.
[0043] In this step, cluster analysis using DBSCAN can automatically identify UV radiation areas, reducing the need for manual intervention and improving identification efficiency. Building a relationship map helps to visualize and analyze the relationship between UV radiation areas and power values, providing a reliable basis for subsequent discharge event identification.
[0044] like Figure 2 As shown, this step specifically includes:
[0045] S210: Establish a discharge detection equipment model based on the characteristics of the power system and the type of equipment, and determine possible characteristics of the discharge location and the location of the UV radiation sensor;
[0046] S220: using DBSCAN to identify UV radiation areas according to the density of UV radiation data points;
[0047] Before using DBSCAN, you need to set two important parameters: neighborhood radius and minimum number of neighborhood samples. The neighborhood radius defines the neighborhood range of a data point, and the minimum number of neighborhood samples defines the number of data points in the minimum neighborhood of a core point.
[0048] The DBSCAN algorithm is performed on the UV radiation data points with the set parameters. The algorithm identifies high-density areas based on the density of the data points and treats low-density data points as noise. In the case of UV radiation data points, high-density areas may correspond to areas with higher radiation intensity, while low-density areas may correspond to areas with lower radiation intensity.
[0049] Based on the results of DBSCAN, areas with high density of UV radiation are identified and visualized.
[0050] S230: Based on the identified UV radiation area, a UV radiation area-power value-UV radiation value relationship map is constructed.
[0051] For the UV radiation area identified in step S220, the average power value and UV radiation value in the area are calculated. The associated information of the power value and UV radiation value of each area is obtained, and each UV radiation area is used as a node. The connection between the nodes represents the relationship between them. A UV radiation area-power value-UV radiation value relationship map is constructed.
[0052] S300: Collect UV radiation data around the possible discharge location determined by cluster analysis as a training data set, and record labels of discharge events, namely normal and abnormal states, according to a relationship map.
[0053] This step records the labels of discharge events through the relationship map, which can more accurately associate the collected UV radiation data points with the discharge events, thereby providing more reliable label information for the machine learning model. Using the UV radiation area-power value-UV radiation value relationship map, the collected UV radiation data points can be labeled in combination with the regional feature information, so as to better use the regional features to guide the machine learning model to learn the characteristics of the discharge event.
[0054] like Figure 3 As shown, this step specifically includes:
[0055] S310: Determine possible discharge locations based on the UV radiation areas obtained by clustering the UV radiation data using DBSCAN in S200. These locations may correspond to potential discharge activity areas in the electromechanical equipment of the power system.
[0056] S320: Using the previously constructed UV radiation area-power value-UV radiation value relationship map, the collected UV radiation data points can be associated with the corresponding UV radiation area according to the characteristics of each UV radiation area, and the label of the discharge event can be marked. This step can be achieved by matching the UV radiation data point with the nearest UV radiation area and labeling the data point with a normal or abnormal state according to the label information of the area.
[0057] S400: Extract features of UV radiation data from a training data set using a time series feature method as input, train a machine learning model, and use label information to guide the model to learn UV radiation features in normal and abnormal states.
[0058] This step can better capture the temporal regularity of UV radiation data by extracting the time series features of UV radiation data, and provide the model with richer feature information. By using label information to guide the model to learn the UV radiation features of normal and abnormal states, the model can strengthen its learning of features under different states and improve its ability to identify discharge events.
[0059] like Figure 4 , as shown, this step specifically includes:
[0060] S410: Extracting features of UV radiation data from the training data set as input, including time series features, frequency domain features, statistical features, etc. These features can reflect the changing laws and characteristics of UV radiation data.
[0061] S420: Associating the UV radiation data in the training data set with the corresponding discharge event labels to ensure that each data point has corresponding label information of a normal or abnormal state.
[0062] S430: Using the extracted UV radiation features as input and the corresponding label information, a machine learning model is trained. Various machine learning algorithms, such as support vector machines, random forests, neural networks, etc., can be used to build a classification model for discharge events.
[0063] S440: When training the machine learning model, the label information recorded in step S300 is used to guide the model to learn the UV radiation characteristics of normal and abnormal states. This can be done by supervised learning, so that the model can learn to distinguish the UV radiation characteristics in normal and abnormal states.
[0064] S500: After evaluating the model using cross-validation, the trained machine learning model is used to predict the real-time collected UV radiation data to identify discharge locations and events.
[0065] This step uses the trained machine learning model to predict the real-time collected UV radiation data, which can achieve real-time identification of discharge locations and events and help take appropriate measures in a timely manner. The model is learned based on the training data collected in S300, and can better adapt to the discharge conditions of different equipment in the power system, improving the accuracy and reliability of discharge event identification.
[0066] like Figure 5 As shown, this step specifically includes:
[0067] S510: First, the trained machine learning model is cross-validated to evaluate its performance in identifying discharge locations and events. Cross-validation can help evaluate the generalization ability of the model, that is, the model's ability to adapt to new data.
[0068] S520: Use the trained machine learning model to predict the real-time collected UV radiation data. The model predicts the discharge location and possible occurrence of the event based on the input UV radiation data characteristics.
[0069] S530: Based on the prediction results of the model, possible discharge locations and events are identified. The model determines whether the real-time collected UV radiation data conforms to normal or abnormal conditions based on the learned UV radiation characteristics and label information, thereby identifying the location and type of the discharge event.
[0070] In summary, the discharge detection method uses UV radiation sensors to obtain UV radiation change information at the discharge location in the electromechanical equipment of the power system. Through cluster analysis and machine learning model training, it realizes the identification and prediction of discharge events, providing an efficient and accurate method for discharge detection in the power system.
[0071] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.
[0072] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.
[0073] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.
[0074] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features of the present invention.
[0075] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.
Claims
1. A partial discharge detection method based on UV radiation, characterized in that: include: S100, selecting the electromechanical equipment in the power system that needs to be tested for discharge, and installing UV radiation sensors to obtain information on changes in UV radiation at the discharge location; S200, based on the discharge detection equipment model and the corresponding UV radiation sensor position, using DBSCAN to cluster the UV radiation data, identify the UV radiation area, and construct a UV radiation area-power value-UV radiation value relationship map; S300, collecting UV radiation data around the possible discharge location determined by cluster analysis as a training data set, and recording labels of discharge events, i.e., normal and abnormal states, according to a relationship map; S400, extracting features of UV radiation data from a training data set using a time series feature method as input, training a machine learning model, and using label information to guide the model to learn UV radiation features in normal and abnormal states; S500, after evaluating the model using cross validation, use the trained machine learning model to predict the UV radiation data collected in real time to identify discharge locations and events.
2. The method for detecting partial discharge based on UV radiation according to claim 1, characterized in that: The S200 includes: S210: Establish a discharge detection equipment model based on the characteristics of the power system and the type of equipment; S220: Using DBSCAN to identify UV radiation areas according to the density of UV radiation data points.
3. The method for detecting partial discharge based on UV radiation according to claim 1, characterized in that: The S300 includes: S310: determining a possible discharge position according to the UV radiation area obtained by clustering the UV radiation data using DBSCAN in S200; S320: using the constructed UV radiation area-power value-UV radiation value relationship map, the collected UV radiation data points are associated with the corresponding UV radiation areas, and the labels of the discharge events are marked.
4. The method for detecting partial discharge based on UV radiation according to claim 1, characterized in that: The S400 includes: S410: extracting features of UV radiation data from the training data set as input; S420: Associating UV radiation data in the training data set with discharge event labels; S430: using the extracted UV radiation features as input and corresponding label information to train a machine learning model; S440: When training the machine learning model, use the label information recorded in step S300 to guide the model to learn the UV radiation characteristics of normal and abnormal states.
5. The method for partial discharge detection based on UV radiation according to claim 1, characterized in that: The S500 includes: S510: Cross-validate the trained machine learning model; S520: using the trained machine learning model to predict UV radiation data collected in real time; S530: Based on the prediction results of the model, possible discharge locations and events are identified.
