Mutual inductor state evaluation method and system based on cloud edge collaboration
By performing transformer data preprocessing and slicing processing at edge computing and analyzing it in cloud computing power, the problem of large data transmission volume and time delay in transformer state evaluation is solved, real-time monitoring and accurate fault detection of transformers are realized.
Patent Information
- Application Number
- CN202411213038.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-30
- Publication Date
- 2025-01-03
AI Technical Summary
In the prior art, the transformer state evaluation method is difficult to meet the real-time monitoring needs due to the large data transmission volume and the time delay, resulting in insufficient timeliness of fault discovery and processing.
The transformer state evaluation method based on cloud-edge collaboration is adopted to perform data preprocessing and slicing processing through edge computing to reduce the amount of data transmission, and use cloud computing power for analysis to achieve real-time monitoring. The specific steps include collecting mutual inductive data for preprocessing, judging fluctuation values, performing data slicing and transmitting to the cloud, and using pre-trained models for state analysis.
It improves the timeliness of fault detection and processing, reduces data transmission delay, enhances monitoring accuracy, meets real-time monitoring needs, and avoids hidden dangers caused by detection delay.
Smart Images
Figure CN119250556A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of mutual inductor technology, and specifically to a mutual inductor status evaluation method and system based on cloud-edge collaboration. Background Art
[0002] The transformer is an important device for measuring voltage and current in the power system. Its accuracy is directly related to the reliability of relay protection, measuring instruments and power metering. In order to evaluate the status of the transformer, a distribution network transformer operation status evaluation method based on cloud-edge collaboration, such as patent number CN115759799A, includes obtaining the field test data of the distribution network transformer; obtaining the historical verification data of the distribution network transformer from the remote server side; using the target state evaluation model to comprehensively analyze the field test data and the historical verification data to obtain the evaluation result of the operation status of the distribution network transformer. Through the above scheme, the operation status of the distribution network transformer can be intelligently evaluated, thereby realizing the effective control of the distribution network transformer and greatly saving human resources. Although cloud computing has powerful computing power and massive data processing capabilities, and can perform complex analysis and long-term data storage, it has the problems of large data transmission volume and extended time, which is difficult to meet the real-time monitoring needs. Summary of the invention
[0003] In view of the problem that the existing technology has large data transmission volume and long time, which makes it difficult to meet the real-time monitoring needs, the present invention provides a transformer status evaluation method and system based on cloud-edge collaboration, which makes full use of the localized data processing characteristics of edge computing, pushes data processing and calculation to the edge of the network, so that the data can be preliminarily processed locally, reducing the data transmission volume, and uses data slicing technology to slice the data, and transmits the sliced data in parallel to the cloud, jointly solving the problems of large data transmission volume and long time faced by the cloud, and using the computing power of the cloud to make up for the problem of low edge computing computing power, and finally realizes the real-time monitoring needs of the transformer. It solves the problem that it is difficult to meet the real-time monitoring needs due to large data transmission volume and long time. It improves the timeliness of fault detection and processing, avoids the hidden dangers caused by detection delays, and thus improves the monitoring accuracy.
[0004] In order to solve the above technical problems, the present invention provides a method for evaluating the state of a transformer based on cloud-edge collaboration, comprising the following steps: S1: Collect mutual inductance data, pre-process the mutual inductance data on the side to obtain first target data, determine the fluctuation value of the first target data in each first preset period, and discard the first target data if the fluctuation value is less than the preset value, otherwise execute S2; S2: Slice the first target data based on data characteristics and network characteristics of the first target data to obtain second target data, and send the second target data to the cloud server using private communication; S3: Input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
[0005] After adopting the above technical solution, the present invention has the following advantages: The first target data is specially selected for the status evaluation of the transformer, and can directly or indirectly evaluate and predict the operating status, performance or potential faults of the transformer. Therefore, the first target data is obtained. When the transformer operates normally, the first target data will be maintained in a relatively stable range. However, once the transformer encounters a fault, the corresponding first target data will fluctuate. In order to identify these fault signs in time, a preset value is set. When the actual fluctuation value is lower than this preset value, it can be judged that the transformer is not currently faulty. Therefore, if the fluctuation value is less than the preset value, the first target data is discarded. By processing the data multiple times, the amount of data is reduced, and the data transmission delay is indirectly reduced; In the process of data transmission, the network bandwidth determines the data transmission time to a certain extent. Therefore, the data is sliced based on the network characteristics. The first target data includes voltage and current data. In order to ensure the correct reorganization of the voltage and current data in the subsequent data reorganization process, identifiers are added to the voltage data and the current data respectively when the sliced data is transmitted. The identifiers will affect the number and size of the slices. Therefore, the first target data is sliced based on the data characteristics of the first target data. The data is sliced reasonably through the data characteristics of the first target data and the network characteristics. At the same time, after the data is sliced, when a certain data is abnormal, the sliced data can be directly retransmitted to prevent the retransmission of the entire data, further significantly reducing the data transmission delay; It improves data processing efficiency and transmission efficiency, improves the timeliness of fault discovery and processing, avoids hidden dangers caused by detection delays, thereby improving monitoring accuracy and solving the problem of difficulty in meeting real-time monitoring needs due to large data transmission volume and long time.
[0006] Preferably, in S1, the preprocessing of the mutual induction data on the side to obtain the first target data includes: Filtering the mutual inductance data based on the cutoff frequency and the filter order to obtain first valid data; Eliminate invalid data from the first valid data based on three times the standard deviation to obtain second valid data; Extract the first target data from the second valid data based on preset rules. Identify and eliminate data caused by measurement errors or extreme situations to avoid distorting the overall distribution of the data and affecting the accuracy of subsequent analysis. By eliminating these invalid data, the amount of data is reduced. Extract the first target data from the valid data based on preset rules, further reducing the amount of data and significantly improving the quality and reliability of the data set.
[0007] Preferably, S2 includes: S21: Acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain a target time period; S22: Slicing the first target data based on data features and a target time period of the first target data to obtain second target data.
[0008] The demand for network bandwidth may vary significantly in different time periods. By analyzing historical bandwidth data, we can identify periods of high bandwidth demand and periods of low bandwidth demand. When slicing data in periods of high bandwidth demand and low bandwidth demand, the slicing strategies are different. Therefore, slicing the first target data based on the target time period helps to slice data in different time periods in a targeted manner, so as to minimize transmission delays.
[0009] Preferably, in S21, dividing the preset time period based on the historical bandwidth data to obtain the target time period includes: If the number of adjacent historical bandwidth data greater than the preset bandwidth data is greater than the preset number, the time period corresponding to the adjacent historical bandwidth data greater than the preset bandwidth data is the first target time period; If the number of adjacent historical bandwidth data less than the preset bandwidth data is greater than or equal to the preset number, the time period corresponding to the adjacent historical bandwidth data less than or equal to the preset bandwidth data is the second target time period. By identifying the high bandwidth demand time period and the low bandwidth demand time period, it is helpful to slice the data in different time periods in a targeted manner, so as to minimize the transmission delay.
[0010] Preferably, the S22 includes: S221: classifying the first target data based on data features of the first target data to obtain a classification result, and setting an identifier based on the classification result; S222: Obtain an average bandwidth of the first target period, and slice the first target data corresponding to the first target period based on the average bandwidth and the overhead of the identifier; S223: Obtain the first target data volume corresponding to the second target time period, and slice the first target data corresponding to the second target time period based on the first target data volume and the overhead of the identifier.
[0011] The first target period is a period of high bandwidth demand. In order to reduce transmission delay, the slices in high bandwidth demand periods are generally larger. In order to avoid network congestion caused by excessively large slices, which in turn causes longer transmission delays, slicing is performed based on the average bandwidth, so that each transmission can carry more data while avoiding network congestion, thereby reducing the total number of transmissions and reducing the overall transmission delay. In periods of low bandwidth demand, network congestion is prone to occur. In order to avoid network congestion, slicing is performed based on the first target data volume. Incorporating identifier overhead into the slicing strategy can prevent slices from being too large and reducing transmission efficiency. By identifying periods of high bandwidth demand and low bandwidth demand, it is helpful to slice data in different time periods in a targeted manner to minimize transmission delays.
[0012] Preferably, S3 includes: S31: inputting the second target data into the pre-trained transformer state evaluation model, grouping the second target data based on data features of the second target data, matching the second target data of different groups in each second preset period with the corresponding benchmark in the transformer state evaluation model, and thereby obtaining a matching result; S32: Output the state of the transformer based on the matching result. Group processing can analyze the state of the transformer more finely and reduce misjudgment caused by data mixing. The benchmark is the normal voltage and current law of the transformer. By matching the obtained current and voltage with the corresponding benchmark, the state of the transformer is output based on the matching result, avoiding the use of a single data to judge the state of the transformer, thereby improving the robustness of the model and the monitoring accuracy.
[0013] Preferably, the S32 includes: If any two or more second target data of different groups in any second preset period do not match the corresponding benchmark in the pre-trained transformer state evaluation model, a transformer fault is output; If there are no two or more different groups of second target data that do not match the corresponding benchmark in the pre-trained transformer state evaluation model in each second preset period, the number of groups of second target data in different groups in each second preset period that do not match the corresponding benchmark in the pre-trained transformer state evaluation model is obtained, and the total amount is calculated based on the number of groups. If the total amount is greater than the preset total amount, a transformer fault is output. Through group matching judgment, the abnormalities of the transformer in multiple aspects can be quickly captured, which improves the sensitivity of fault detection. The method of accumulating the number of mismatched groups can capture those abnormalities that are not obvious but persist, further enhancing the comprehensiveness of transformer fault detection.
[0014] Beneficial effects of this program: The first target data is specially selected for the status evaluation of the transformer, and can directly or indirectly evaluate and predict the operating status, performance or potential faults of the transformer. Therefore, the first target data is obtained. When the transformer is operating normally, the first target data will be maintained in a relatively stable range. However, once the transformer encounters a fault, the corresponding first target data will fluctuate significantly. In order to identify these fault signs in time, a preset value is set. When the actual fluctuation value is lower than this preset value, it can be judged that the transformer is not currently faulty. Therefore, if the fluctuation value is less than the preset value, the first target data is discarded. By processing the data multiple times, the amount of data is reduced, thereby reducing the data transmission delay; In the process of data transmission, the network bandwidth determines the data transmission time to a certain extent. Therefore, the data is sliced based on the network characteristics. The first target data includes voltage and current data. In order to ensure the correct reorganization of the voltage and current data in the subsequent data reorganization process, identifiers are added to the voltage data and the current data respectively when the sliced data is transmitted. The identifiers will affect the number and size of the slices. Therefore, the first target data is sliced based on the data characteristics of the first target data. The data is sliced reasonably based on the data characteristics of the first target data and the network characteristics. At the same time, after the data is sliced, when a certain data is abnormal, the sliced data can be directly retransmitted to prevent the retransmission of the entire data, further significantly reducing the data transmission delay; Through group matching judgment, the abnormalities of the transformer in multiple aspects can be quickly captured, which improves the sensitivity of fault detection. The judgment method of accumulating the number of mismatched groups can capture those abnormalities that are not obvious but persist, further enhancing the comprehensiveness of transformer fault detection. It solves the problem of difficulty in meeting real-time monitoring needs due to large data transmission volume and long transmission time.
[0015] The present invention also provides a transformer status evaluation system based on cloud-edge collaboration, including a sensor module, a data processing module, a data transmission module, and a data analysis module; The sensor module is used to collect mutual induction data; The data processing module is used to pre-process the mutual induction data on the side to obtain the first target data; The data transmission module is used for slicing the first target data based on the data characteristics and network characteristics of the first target data to obtain the second target data; The data analysis module is used to input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
[0016] Preferably, the data transmission module includes a target time period acquisition module, and the target time period acquisition module is used to acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain the target time period.
[0017] Preferably, the data analysis module includes a group matching module, which is used to group the second target data based on data characteristics of the second target data, and match the second target data of different groups in each second preset period with corresponding benchmarks in the transformer state evaluation model.
[0018] Beneficial effects of this program: The first target data is specially selected for the status evaluation of the transformer, and can directly or indirectly evaluate and predict the operating status, performance or potential faults of the transformer. When the transformer is operating normally, the first target data will be maintained in a relatively stable range. However, once the transformer encounters a fault, the corresponding first target data will fluctuate. In order to identify these fault signs in time, a preset value is set. When the actual fluctuation value is lower than this preset value, it can be judged that the transformer is not currently faulty. Therefore, if the fluctuation value is less than the preset value, the first target data is discarded. By processing the data multiple times, the amount of data is reduced, thereby reducing the data transmission delay; In the process of data transmission, the network bandwidth determines the data transmission time to a certain extent. Therefore, the data is sliced based on the network characteristics. The first target data includes voltage and current data. In order to ensure the correct reorganization of voltage and current data in the subsequent data reorganization process, identifiers are added to the voltage data and current data respectively when the sliced data is transmitted. The identifiers will affect the number and size of slices. Therefore, the first target data is sliced based on the data characteristics of the first target data. The data is sliced reasonably through the data characteristics of the first target data and the network characteristics. At the same time, after the data is sliced, when a certain data is abnormal, the sliced data can be directly retransmitted to prevent the retransmission of the entire data, further reducing the data transmission delay; Through multi-group matching judgment, the abnormalities of the mutual inductor in multiple aspects can be quickly captured, which improves the sensitivity of fault detection and further enhances the comprehensiveness of fault detection; It solves the problem of difficulty in meeting real-time monitoring needs due to large data transmission volume and long transmission time.
[0019] After adopting the above technical solution, the present invention has the following advantages: The first target data is specially selected for the status evaluation of the transformer, and can directly or indirectly evaluate and predict the operating status, performance or potential faults of the transformer. Therefore, the first target data is obtained. When the transformer is operating normally, the first target data will be maintained in a relatively stable range. However, once the transformer encounters a fault, the corresponding first target data will fluctuate significantly. In order to identify these fault signs in time, a preset value is set. When the actual fluctuation value is lower than this preset value, it can be judged that the transformer is not currently faulty. Therefore, if the fluctuation value is less than the preset value, the first target data is discarded. By processing the data multiple times, the amount of data is reduced, thereby reducing the data transmission delay; In the process of data transmission, the network bandwidth determines the data transmission time to a certain extent. Therefore, the data is sliced based on the network characteristics. The first target data includes voltage and current data. In order to ensure the correct reorganization of the voltage and current data in the subsequent data reorganization process, identifiers are added to the voltage data and the current data respectively when the sliced data is transmitted. The identifiers will affect the number and size of the slices. Therefore, the first target data is sliced based on the data characteristics of the first target data. The data is sliced reasonably based on the data characteristics of the first target data and the network characteristics. At the same time, after the data is sliced, when a certain data is abnormal, the sliced data can be directly retransmitted to prevent the retransmission of the entire data, further significantly reducing the data transmission delay; Through group matching judgment, the abnormalities of the transformer in multiple aspects can be quickly captured, which improves the sensitivity of fault detection. The judgment method of accumulating the number of mismatched groups can capture those abnormalities that are not obvious but persist, further enhancing the comprehensiveness of transformer fault detection. It solves the problem of difficulty in meeting real-time monitoring needs due to large data transmission volume and long transmission time. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Other features, objects and advantages of the present invention will become more apparent by reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are only for the purpose of illustrating preferred embodiments and are not to be considered as limiting the present invention. Also, the same reference symbols are used throughout the drawings to represent the same parts.
[0021] Figure 1 It is a method flow chart of the method for evaluating the state of a mutual inductor based on cloud-edge collaboration of the present invention. DETAILED DESCRIPTION
[0022] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific implementation method described herein is only an optimal embodiment of the present invention, which is only used to explain the present invention and does not limit the scope of protection of the present invention. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0023] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the drawings; the process can correspond to a method, function, procedure, subroutine, subprogram, etc.
[0024] Embodiment 1: like Figure 1 As shown, the transformer status evaluation method based on cloud-edge collaboration includes the following steps: S1: Collect mutual inductance data, pre-process the mutual inductance data on the side to obtain first target data, determine the fluctuation value of the first target data within each first preset period, and discard the first target data if the fluctuation value is less than the preset value, otherwise execute S2.
[0025] In S1, the preprocessing of the mutual induction data on the side to obtain the first target data includes: Filtering the mutual inductance data based on the cutoff frequency and the filter order to obtain first valid data; Eliminate invalid data from the first valid data based on three times the standard deviation to obtain second valid data; The first target data is extracted from the second valid data based on a preset rule.
[0026] In this embodiment, a high-precision, high-stability voltage sensor is selected to collect the voltage data of the transformer, and a Hall effect sensor or a current transformer is selected to collect the current data of the transformer. The sensor is installed at a key position of the transformer to ensure the accuracy of data collection. Usually, a position close to the transformer is selected to reduce signal interference and transmission loss. At the same time, a suitable connector and shielded cable are used to reduce external electromagnetic interference. The sensor removes the DC component through a filtering circuit to ensure the stability of the signal. For weak signals, the signal strength is increased through an amplification circuit to enhance the signal-to-noise ratio, and the processed data is transmitted in the form of a digital signal. Through the above steps, stable data can be obtained, and the stable data is filtered based on a filtering algorithm to obtain the first valid data. For example, a low-pass filter can be used to filter the stable data, and the transfer function of the filter is: H(f)=1 / (1+(f / f c ) 2n ), where f c is the cutoff frequency, and n is the filter order. Invalid data is removed from the data based on filtering to ensure the validity of the data. For example, the triple standard deviation method can be used to remove invalid data. The calculation formula is: |x i -μ|>3σ, where x i is the data point, μ is the data mean, and σ is the standard deviation. The first target data is extracted from the second valid data. The first target data is the fluctuation amplitude, frequency component, etc. of the voltage and current. The fluctuation value of the first target data in each first preset period is determined. If the fluctuation value is less than the preset value, the first target data is discarded, otherwise S2 is executed. If S2 is executed, the first target data is packaged and a checksum is attached to ensure the integrity of the first target data during transmission. During the packaging process, the first target data is organized in a certain format, and necessary checksum information is added so that possible errors can be detected and corrected during the transmission process. For example, the first target data is verified using a CRC (cyclic redundancy check) checksum. By processing the data multiple times, the amount of data is reduced, thereby reducing the data transmission delay.
[0027] S2: Slice the first target data based on data characteristics and network characteristics of the first target data to obtain second target data, and send the second target data to the cloud server using private communication.
[0028] The S2 includes: S21: Acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain a target time period; S22: Slicing the first target data based on data features and a target time period of the first target data to obtain second target data.
[0029] In S21, the preset time period is divided based on the historical bandwidth data to obtain the target time period, including: If the number of adjacent historical bandwidth data greater than the preset bandwidth data is greater than the preset number, the time period corresponding to the adjacent historical bandwidth data greater than the preset bandwidth data is the first target time period; If the number of adjacent historical bandwidth data smaller than the preset bandwidth data is greater than or equal to the preset number, the time period corresponding to the adjacent historical bandwidth data smaller than or equal to the preset bandwidth data is the second target time period.
[0030] The S22 includes: S221: classifying the first target data based on data features of the first target data to obtain a classification result, and setting an identifier based on the classification result; S222: Obtain an average bandwidth of the first target period, and slice the first target data corresponding to the first target period based on the average bandwidth and the overhead of the identifier; S223: Obtain the first target data volume corresponding to the second target time period, and slice the first target data corresponding to the second target time period based on the first target data volume and the overhead of the identifier.
[0031] In this embodiment, the network bandwidth usage within one day can be obtained as historical bandwidth data. By analyzing the historical bandwidth data, if the historical network bandwidth is 60Mbps, 65Mbps, 66Mbps, 63Mbps, the preset bandwidth data is 61Mbps, the preset number is 3, and the time period corresponding to 65Mbps to 63Mbps is the first target time period. If the historical network bandwidth is 60Mbps, 65Mbps, 66Mbps, 63Mbps, the preset bandwidth data is 66Mbps, and the preset number is 2. At this time, the time period corresponding to 60Mbps to 65Mbps is the second target time period. The first target time period is a high bandwidth demand time period. In order to reduce the transmission delay, the high bandwidth demand time period generally has a larger slice. In order to avoid network congestion caused by excessive slices, slicing is performed based on the average bandwidth, so that each transmission can carry more data while avoiding network congestion, thereby reducing the total number of transmissions, thereby reducing the overall transmission delay. In the low bandwidth demand period, network congestion is prone to occur. In order to avoid network congestion, slicing is performed based on the first target data volume. Incorporating identifier overhead into the consideration scope of the slicing strategy can avoid slicing too large and reducing transmission efficiency. By identifying the periods of high and low bandwidth demand, it helps to slice the data in different time periods in a targeted manner to minimize transmission delays.
[0032] S3: Input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
[0033] The S3 includes: S31: inputting the second target data into the pre-trained transformer state evaluation model, grouping the second target data based on data features of the second target data, matching the second target data of different groups in each second preset period with the corresponding benchmark in the transformer state evaluation model, and thereby obtaining a matching result; S32: Output the state of the transformer based on the matching result.
[0034] The S32 includes: If any two or more second target data of different groups in any second preset period do not match the corresponding benchmark in the pre-trained transformer state evaluation model, a transformer fault is output; If there are no two or more different groups of second target data that do not match the corresponding benchmark in the pre-trained transformer state evaluation model in each second preset period, then the number of groups of second target data in different groups in each second preset period that do not match the corresponding benchmark in the pre-trained transformer state evaluation model is obtained, and the total amount is calculated based on the number of groups. If the total amount is greater than the preset total amount, a transformer fault is output.
[0035] In this embodiment, after receiving the transmitted second target data, the cloud server reorganizes and verifies the second target data, and uses the verification code to ensure that there is no error in the transmission of the second target data. After the verification is passed, the cloud server stores the data and enters the next step of the deep analysis stage to filter and normalize the second target data for subsequent processing. The data normalization formula is: Among them, x' is the normalized data, min(x) and max(x) are the minimum and maximum values of the data respectively. The normalized second target data is input into the transformer status evaluation model, the second target data is divided into voltage data, current data, frequency data, etc., the acquisition time corresponding to the second target data is divided into multiple cycles, and the cycle is the second preset cycle. The voltage data, current data, and frequency data are matched with the reference voltage data, reference current data, and reference frequency data in the model. If the voltage data and current data do not match the reference voltage data and reference current data, it indicates a transformer fault. If the voltage data, current data, and frequency data match the reference voltage data, reference current data, and reference frequency data in the model, only one If the match is not successful, the number of groups that do not match the corresponding benchmark in the transformer state evaluation model with the second target data of different groups in each preset period is obtained, and the total amount is calculated based on the number of groups. For example, if the second preset period is 4, in the first second preset period, only the voltage data does not match the reference voltage data, in the second second preset period, the current data, voltage data, and frequency data all match, in the third second preset period, only the current data does not match the reference current data, and in the fourth second preset period, only the current data does not match the reference current data. At this time, the total amount is 3. If the preset total amount is 2, the transformer fault is output. Through multi-group matching judgment, the abnormalities of the transformer in multiple aspects can be quickly captured, which improves the sensitivity of fault detection. The judgment method of accumulating the number of mismatched groups can capture those abnormalities that are not obvious but persist, further enhancing the comprehensiveness of fault detection.
[0036] In this embodiment, the second target data may be further processed. For example, frequency features may be extracted using Fourier transform, and instantaneous features may be extracted using wavelet transform. The wavelet transform formula is: Where a and b are scale and translation parameters, and ψ is the wavelet function. The support vector machine (SVM) algorithm is used to identify complex patterns in the second target data and detect potential faults. The decision function of the SVM classifier is: Among them, α i is the Lagrange multiplier, y i is the category label, K(x i , x) is the kernel function. According to the historical trend and current status of the second target data, its future operating status is predicted. Time series analysis is performed using deep learning algorithms such as long short-term memory network (LSTM). By analyzing the frequency characteristics, instantaneous characteristics, fault modes, and future operating status of the transformer with the second target data, a comprehensive evaluation and prediction of the transformer status is achieved.
[0037] Embodiment 2: The transformer status evaluation system based on cloud-edge collaboration includes sensor module, data processing module, data transmission module and data analysis module; The sensor module is used to collect mutual induction data; The data processing module is used to pre-process the mutual induction data on the side to obtain the first target data; The data transmission module is used for slicing the first target data based on the data characteristics and network characteristics of the first target data to obtain the second target data; The data analysis module is used to input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
[0038] The data transmission module includes a target time period acquisition module, which is used to acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain a target time period.
[0039] The data analysis module includes a group matching module, which is used to group the second target data based on data features of the second target data, and match the second target data of different groups in each second preset period with corresponding benchmarks in the transformer state evaluation model.
[0040] In this embodiment, the sensor module includes a voltage sensor and a current sensor. A high-precision and high-stability voltage sensor is selected to collect the voltage data of the transformer, and a Hall effect sensor or a current transformer is selected to collect the current data of the transformer. The sensor is installed at a key position of the transformer to ensure the accuracy of data collection. Usually, a position close to the transformer is selected to reduce signal interference and transmission loss. At the same time, a suitable connector and shielded cable are used to reduce external electromagnetic interference. The sensor removes the DC component through a filtering circuit to ensure the stability of the signal. For weak signals, the signal strength is increased through an amplification circuit to enhance the signal-to-noise ratio, and the processed data is transmitted in the form of a digital signal. Through the above steps, stable data can be obtained. The data processing module processes the data as follows: the stable data is filtered based on a filtering algorithm to obtain the first valid data. For example, a low-pass filter can be used to filter the stable data. The transfer function of the filter is: H(f)=1 / (1+(f / f c ) 2n ), where f c is the cutoff frequency, and n is the filter order; invalid data is removed from the data on the basis of filtering to ensure the validity of the data. For example, the triple standard deviation method can be used to remove invalid data, and the calculation formula is: |x i -μ|>3σ, where x iis the data point, μ is the data mean, and σ is the standard deviation; extract the first target data from the second valid data, the first target data is the fluctuation amplitude, frequency component, etc. of the voltage and current, judge the fluctuation value of the first target data in each first preset period, if the fluctuation value is less than the preset value, discard the first target data, otherwise execute S2, if S2 is executed, the first target data is packaged, and a checksum is attached to ensure the integrity of the first target data during transmission. During the packaging process, the first target data is organized in a certain format, and necessary checksum information is added so that possible errors can be detected and corrected during the transmission process, for example, the first target data is checked using a CRC checksum. The data is processed multiple times by the data processing module, which reduces the amount of data and thus reduces the data transmission delay.
[0041] In this embodiment, the target time period acquisition module acquires the network bandwidth usage within one day as the historical bandwidth data. By analyzing the historical bandwidth data, if the historical network bandwidth is 60Mbps, 65Mbps, 66Mbps, 63Mbps, the preset bandwidth data is 61Mbps, the preset number is 3, and the time period corresponding to 65Mbps to 63Mbps is the first target time period. If the historical network bandwidth is 60Mbps, 65Mbps, 66Mbps, 63Mbps, the preset bandwidth data is 66Mbps, the preset number is 2, and the time period corresponding to 60Mbps to 65Mbps is the second target time period. The first target time period is a high bandwidth demand time period, and the second target time period is a low bandwidth demand time period. After the target time period acquisition module acquires the high bandwidth demand time period and the low bandwidth demand time period and transmits them to the data transmission module, the data transmission module performs targeted slicing processing on the data. By identifying the high bandwidth demand time period and the low bandwidth demand time period, it is helpful to perform targeted slicing of data in different time periods so as to minimize transmission delay.
[0042] In this embodiment, the group matching module divides the second target data into voltage data, current data, and frequency data, and divides the acquisition time corresponding to the first target data into multiple cycles, which is the second preset cycle. The voltage data, current data, and frequency data are matched with the reference voltage data, reference current data, and reference frequency data in the model. Through multi-group matching judgment, the abnormalities of the transformer in multiple aspects can be quickly captured, thereby improving the sensitivity and comprehensiveness of fault detection.
[0043] In this embodiment, the data analysis module can further process the second target data to extract key features from the second target data. For example, frequency features are extracted using Fourier transform, and instantaneous features are extracted using wavelet transform; the support vector machine algorithm is used to identify complex patterns in the second target data and detect potential faults. The decision function of the SVM classifier is: Among them, α i is the Lagrange multiplier, y i is the category label, K(x i , x) is the kernel function; according to the historical trend and current status of the second target data, its future operating status is predicted. Time series analysis is performed using deep learning algorithms such as long short-term memory networks. By analyzing the frequency characteristics, instantaneous characteristics, fault modes, and future operating status of the transformer with the second target data, a comprehensive evaluation and prediction of the transformer status is achieved.
[0044] After a comprehensive evaluation and prediction of the status of the transformer, the cloud server generates a detailed report, including the current status of the transformer, historical trends, potential faults, and recommended maintenance measures. These reports provide important decision-making support for the operation and maintenance of the power system. The cloud server feeds back the analysis results and suggestions to the edge device through the wireless network. The edge device makes corresponding adjustments and optimizations based on the suggestions fed back by the cloud. For example, the edge device can adjust the data collection and processing strategy based on the suggestions of the cloud to improve the accuracy and efficiency of subsequent data processing and monitoring. Users access the cloud server through the PC to view analysis reports, receive alarms, and perform maintenance tasks. The PC interacts with the cloud server in real time through network broadband to ensure that users can obtain the latest transformer status information and maintenance suggestions at any time. Based on the feedback information and data collected during operation, the system continuously optimizes the algorithm and processing flow to improve the reliability and efficiency of the overall system. Through the cloud-edge collaborative architecture, combined with the advantages of edge computing and cloud computing, and using wireless cloud penetration technology, online, real-time monitoring and in-depth analysis of the transformer status are realized, which solves many limitations of traditional detection methods and improves the operation reliability and fault handling efficiency of the power system.
[0045] The specific implementation described above is a preferred implementation of the transformer status evaluation method and system based on cloud-edge collaboration of the present invention, but it is not intended to limit the specific implementation scope of the present invention. The scope of the present invention includes but is not limited to this specific implementation. All equivalent changes made in accordance with the shape and structure of the present invention are within the protection scope of the present invention.
Claims
1. A transformer status evaluation method based on cloud-edge collaboration, characterized in that: The following steps are involved: S1: Collect mutual inductance data, pre-process the mutual inductance data on the side to obtain first target data, determine the fluctuation value of the first target data in each first preset period, and discard the first target data if the fluctuation value is less than the preset value, otherwise execute S2; S2: Slice the first target data based on data characteristics and network characteristics of the first target data to obtain second target data, and send the second target data to the cloud server using private communication; S3: Input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
2. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 1 is characterized in that: In S1, the preprocessing of the mutual induction data on the side to obtain the first target data includes: Filtering the mutual inductance data based on the cutoff frequency and the filter order to obtain first valid data; Eliminate invalid data from the first valid data based on three times the standard deviation to obtain second valid data; The first target data is extracted from the second valid data based on a preset rule.
3. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 1 is characterized in that: The S2 includes: S21: Acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain a target time period; S22: Slicing the first target data based on data features and a target time period of the first target data to obtain second target data.
4. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 3 is characterized in that: In S21, the preset time period is divided based on the historical bandwidth data to obtain the target time period, including: If the number of adjacent historical bandwidth data greater than the preset bandwidth data is greater than the preset number, the time period corresponding to the adjacent historical bandwidth data greater than the preset bandwidth data is the first target time period; If the number of adjacent historical bandwidth data smaller than the preset bandwidth data is greater than or equal to the preset number, the time period corresponding to the adjacent historical bandwidth data smaller than or equal to the preset bandwidth data is the second target time period.
5. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 4 is characterized in that: The S22 includes: S221: classifying the first target data based on data features of the first target data to obtain a classification result, and setting an identifier based on the classification result; S222: Obtain an average bandwidth of the first target period, and slice the first target data corresponding to the first target period based on the average bandwidth and the overhead of the identifier; S223: Obtain the first target data volume corresponding to the second target time period, and slice the first target data corresponding to the second target time period based on the first target data volume and the identifier overhead.
6. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 1 is characterized in that: The S3 includes: S31: inputting the second target data into a pre-trained transformer state evaluation model, grouping the second target data based on data features of the second target data, matching the second target data of different groups within each second preset period with corresponding benchmarks in the transformer state evaluation model, and thereby obtaining a matching result; S32: Output the state of the transformer based on the matching result.
7. The method for evaluating the state of a transformer based on cloud-edge collaboration according to claim 6 is characterized in that: The S32 includes: if any two or more second target data of different groups in any second preset period do not match the corresponding benchmark in the pre-trained transformer state evaluation model, outputting a transformer fault; If there are no two or more different groups of second target data that do not match the corresponding benchmark in the pre-trained transformer state evaluation model in each second preset period, then the number of groups of second target data in different groups in each second preset period that do not match the corresponding benchmark in the pre-trained transformer state evaluation model is obtained, and the total amount is calculated based on the number of groups. If the total amount is greater than the preset total amount, a transformer fault is output.
8. A transformer status evaluation system based on cloud-edge collaboration, applicable to the transformer status evaluation method based on cloud-edge collaboration as described in any one of claims 1 to 7, characterized in that: It includes sensor module, data processing module, data transmission module and data analysis module; The sensor module is used to collect mutual induction data; The data processing module is used to pre-process the mutual induction data on the side to obtain the first target data; The data transmission module is used for slicing the first target data based on the data characteristics and network characteristics of the first target data to obtain the second target data; The data analysis module is used to input the second target data into the pre-trained transformer state evaluation model for analysis, and then output the state of the transformer.
9. The transformer status evaluation system based on cloud-edge collaboration according to claim 8 is characterized in that: The data transmission module includes a target time period acquisition module, which is used to acquire historical bandwidth data within a preset time period, and divide the preset time period based on the historical bandwidth data to obtain a target time period.
10. The transformer status evaluation system based on cloud-edge collaboration according to claim 8, characterized in that: The data analysis module includes a group matching module, which is used to group the second target data based on data features of the second target data, and match the second target data of different groups in each second preset period with corresponding benchmarks in the transformer state evaluation model.