Voltage transformer abnormity prediction method and system
Through improved convolutional neural network and attention mechanism, combined with daily, weekly, monthly and annual average phase voltage data, a voltage transformer anomaly prediction model is constructed, which solves the insufficient utilization of historical information and low processing efficiency of large data volumes in the existing technology. The problem of this achieves more efficient and accurate anomaly prediction.
Patent Information
- Application Number
- CN202411185265.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-27
- Publication Date
- 2024-12-13
AI Technical Summary
The existing voltage transformer anomaly prediction methods cannot effectively utilize historical information, and are inefficient when processing large data volumes. It is difficult for existing neural networks to process multiple types of data at the same time, resulting in low abnormal prediction efficiency and high calculation cost.
Adopt an improved convolutional neural network and introduce attention mechanism to build a voltage transformer anomaly prediction model. Through preprocessing and feature extraction of voltage historical data, the comprehensiveness and accuracy of anomaly prediction will be improved, and through daily, weekly and month , annual average phase voltage data processing large data volume.
It improves the accuracy and efficiency of abnormal prediction of voltage transformer, can effectively process large data volume, reduce calculation costs, and enhances the feature extraction ability of abnormal information.
Smart Images
Figure CN119128751A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power transmission and distribution equipment, and in particular relates to a method and system for predicting abnormality of a voltage transformer. Background Art
[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.
[0003] In recent years, the country has strongly supported the development of traditional substations towards intelligence and digitalization. Smart substations should be supported by advanced detection technology to ensure that substations can operate reliably, efficiently, economically and safely. Among them, voltage transformers are the source of voltage information for the protection, control, measurement, and metering systems of smart substations. They are the basic link for the stable operation of substations and an important component of the power transmission and power supply system. They can ensure the stable operation of equipment and are closely related to the normal operation of the power system. However, due to long-term operation and interference from other external factors, the voltage transformers of substations are prone to abnormalities, which directly affect the normal operation of substation equipment and even threaten the safety of the entire power grid system and personnel.
[0004] At present, the existing voltage transformer abnormality prediction cluster analysis method, neural network method, etc., but there are still the following shortcomings:
[0005] (1) It relies only on current information and cannot effectively utilize historical information, which means it is not global and is not suitable for situations with large amounts of data.
[0006] (2) The existing neural network used for voltage transformer anomaly prediction cannot process multiple types of data at the same time, resulting in low anomaly prediction efficiency and high computational cost. Summary of the invention
[0007] In order to overcome the shortcomings of the above-mentioned prior art, the present invention provides a voltage transformer abnormality prediction method and system, which can construct a voltage transformer abnormality prediction model by integrating an improved convolutional neural network and introducing an attention mechanism, thereby improving the comprehensiveness and accuracy of voltage transformer abnormality prediction and improving the abnormality prediction efficiency.
[0008] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:
[0009] A first aspect of the present invention provides a voltage transformer abnormality prediction method, comprising:
[0010] Obtaining historical data of voltage transformer secondary side voltage;
[0011] Preprocessing the acquired voltage history data to obtain preprocessed voltage history data;
[0012] Based on the preprocessed voltage history data, a voltage transformer abnormality prediction model is constructed;
[0013] The voltage transformer abnormality prediction model is constructed by introducing a channel attention mechanism into an improved convolutional neural network containing several parallel channels, and the voltage mutual inductance abnormality prediction model is trained to obtain a trained voltage mutual inductance abnormality prediction model;
[0014] The secondary side voltage of the voltage transformer to be detected is obtained, and the voltage to be detected is input into the trained voltage transformer abnormality prediction model to obtain abnormal information of the voltage transformer.
[0015] The acquired abnormal information of the voltage transformer is evaluated.
[0016] As a further technical solution, the voltage history data includes normal data and abnormal data, mainly including: daily average phase voltage data, weekly average phase voltage data, monthly average phase voltage data, and annual average phase voltage data.
[0017] As a further technical solution, the process of obtaining the historical data of the secondary side voltage of the voltage transformer is as follows: discrete sampling is performed on the secondary voltage output by the voltage transformer to obtain the three-phase voltage of the secondary voltage, and the daily average phase voltage, weekly average phase voltage, monthly average phase voltage, and annual average phase voltage are calculated to respectively construct a daily average phase voltage data set, a weekly average phase voltage data set, a monthly average phase voltage data set, and an annual average phase voltage data set.
[0018] As a further technical solution, a method for preprocessing the acquired voltage history data is: using an inverse tangent function conversion method for normalization processing.
[0019] As a further technical solution, the method for obtaining abnormal information of a voltage transformer is specifically as follows:
[0020] Acquiring voltage data of a voltage transformer to be predicted, and preprocessing the voltage data;
[0021] The preprocessed voltage data is input into the voltage transformer abnormality prediction model to output the abnormal information of the voltage transformer.
[0022] As a further technical solution, the method for evaluating the acquired abnormal information of the voltage transformer is specifically: using the root mean square error and the determination coefficient index to evaluate the modeling accuracy, and the calculation formula is as follows:
[0023]
[0024] Among them, r 2 is the coefficient of determination, RMSE is the root mean square, u iis the historical voltage data, is the value to be predicted, is the voltage mean of historical data, and n is the number of samples.
[0025] As a further technical solution, the voltage transformer abnormality prediction model includes a daily average phase voltage convolution network channel, a weekly average phase voltage convolution network channel, a monthly average phase voltage convolution network channel, and an annual average phase voltage convolution network channel.
[0026] A second aspect of the present invention provides a voltage transformer abnormality prediction system, comprising:
[0027] The historical data acquisition module is configured to: acquire the secondary side voltage historical data of the voltage transformer;
[0028] The preprocessing module is configured to: preprocess the acquired historical data;
[0029] The model prediction module is configured to: construct a voltage transformer abnormality prediction model based on the preprocessed voltage historical data, and perform training, obtain the voltage data to be predicted, input it into the voltage transformer abnormality prediction model, and output the voltage transformer abnormality information.
[0030] The evaluation module is configured to: evaluate the abnormality prediction of the voltage transformer.
[0031] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in a method for predicting abnormality of a voltage transformer as described in the first aspect of the present invention.
[0032] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in a method for predicting an abnormality of a voltage transformer as described in the first aspect of the present invention are implemented.
[0033] One or more of the above technical solutions have the following beneficial effects:
[0034] (1) The embodiment of the present invention is based on an improved parallel multi-channel convolutional neural network and introduces an attention mechanism, which can make full use of the historical data of the voltage transformer, extract more comprehensive abnormal information features, and improve the accuracy of abnormality prediction.
[0035] (2) The embodiment of the present invention uses the daily average phase voltage convolution network channel, the weekly average phase voltage convolution network channel, the monthly average phase voltage convolution network channel, and the annual average phase voltage convolution network channel, which is not affected by the amount of data and realizes abnormal prediction of large amounts of data. At the same time, it reduces the amount of calculation and improves the efficiency of abnormal prediction.
[0036] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0038] Figure 1 This is a flow chart of a voltage transformer abnormality prediction method according to an embodiment of the present invention. DETAILED DESCRIPTION
[0039] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.
[0040] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.
[0041] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.
[0042] This embodiment discloses a method for predicting abnormality of a voltage transformer.
[0043] like Figure 1 As shown, a voltage transformer abnormality prediction method includes:
[0044] Step S1: Acquire the voltage history data of the secondary side of the voltage transformer;
[0045] Step S2: preprocessing the acquired voltage history data to obtain preprocessed voltage history data;
[0046] Step S3: constructing a voltage transformer abnormality prediction model based on the preprocessed voltage history data and performing training;
[0047] Step S4: obtaining voltage data of the voltage transformer to be predicted, and inputting the voltage to be detected into the trained voltage transformer abnormality prediction model to obtain abnormal information of the voltage transformer;
[0048] Among them, the voltage transformer abnormality prediction model is constructed by introducing a channel attention mechanism into an improved convolutional neural network containing several parallel channels, and the voltage mutual inductance abnormality prediction model is trained to obtain a trained voltage mutual inductance abnormality prediction model;
[0049] Step S5: Evaluate the acquired abnormal information of the voltage transformer.
[0050] In step S1, the process of obtaining the historical data of the secondary voltage of the voltage transformer is to perform discrete sampling on the secondary voltage output by the voltage transformer.
[0051] The voltage transformer is an electrical continuous quantity, and it is inconvenient to make logical judgments. The secondary voltage output by the voltage transformer is sampled discretely, specifically,
[0052] The secondary voltages of the three-phase voltage transformers are set to U A , U B , U C , which is the secondary output of the voltage transformer. Set a sampling period. For example, collect a secondary voltage value every 1 hour. The three-phase voltages are U A1 , U B1 , U C1 , the three-phase voltages in the next sampling period are U A2 , U B2 , U C2 , and so on, all the sampled values are sent to the memory for storage, and the secondary voltage values of the three-phase voltage collected every day are averaged to obtain the daily average phase voltage data. The calculation formula is as follows:
[0053]
[0054] Among them, U At , U Bt , U Ct They are the average voltage of the three-phase voltage per day, representing the voltage value of the day, U Ai , U Bi , U Ci is the instantaneous sampling value of the three-phase voltage at each sampling time point on the day, that is, the instantaneous value of the three-phase voltage at a certain moment; n is the total number of voltage sampling points on the day, n=24.
[0055] All daily average phase voltage values are used to construct the daily average phase voltage dataset D(U A1 , U B1 , U C1 ......U At , U Bt , U Ct ).
[0056] Similarly, obtain the weekly average phase voltage data set T(U A1 , U B1 , U C1 ......U Az , U Bz , U Cz), monthly average phase voltage data set M(U A1 , U B1 , U C1 ......U Ay , U By , U Cy ), annual average phase voltage data set N(U A1 , U B1 , U C1 ......U An , U Bn , U Cn ), where t represents day, t∈(1, 2, ....i), z represents week, z∈(1, 2, ....j), y represents month, y∈(1, 2, ....k), and n represents year, n∈(1, 2, ....l).
[0057] After the above steps, the instantaneous sampling value is replaced by the average value, which can reduce the influence of sampling error and power system voltage fluctuation on the secondary voltage.
[0058] In step S2, the method for preprocessing the acquired voltage history data is: using the inverse tangent function conversion method for normalization processing. Specifically, the commonly used normalization formula is as follows:
[0059] Y=arctan(X-5)*2π(2)
[0060] Substitute the obtained daily average phase voltage dataset, weekly average phase voltage dataset, monthly average phase voltage dataset, and annual average phase voltage dataset as input X into formula (2) to obtain the normalized voltage data: daily average phase voltage dataset Y d , Weekly average phase voltage data set Y t , monthly average phase voltage data set Y m , Annual average phase voltage data set Y y .
[0061] After the above steps, normalization can prevent attribute values with a large numerical range in the initial data from covering attribute values with a small data range, providing more accurate data support for the voltage transformer abnormality prediction model.
[0062] In step S3, a voltage transformer abnormality prediction model is constructed based on the preprocessed voltage history data and trained.
[0063] In this embodiment, the voltage transformer abnormality prediction model has a specific framework as follows:
[0064] (1) Input layer
[0065] The input layer is used by the model to receive the normalized daily average phase voltage dataset Y d, Weekly average phase voltage data set Y t , monthly average phase voltage data set Y m , Annual average phase voltage data set Y y .
[0066] (2) Parallel multi-channel convolution kernels
[0067] After the input layer, a 4-channel convolution kernel is designed, and the first channel is used as the daily average phase voltage dataset Y d channel, the second channel is used as the weekly average phase voltage data set Y t Channel, the third channel is used as the monthly average phase voltage data set Y m Channel, the fourth channel is used as the annual average phase voltage data set Y y aisle.
[0068] Each channel uses a 5*5 convolution kernel, and features are extracted through convolution calculation, ultimately obtaining feature information of 24*24 size.
[0069] The result of the convolution is first calculated through a Relu activation function.
[0070] Specifically, taking the first channel as an example, the daily average phase voltage data set Y d Input to the first channel. When performing convolution calculation, add a bias coefficient e first, and the input is Y d All data in the data set, output is Y d The convolution formula is as follows:
[0071]
[0072] Among them, w d is the daily average phase voltage data weight.
[0073] In this embodiment, similarly, the second channel weekly average phase voltage Y t The convolution formula is as follows:
[0074]
[0075] Among them, w t is the weekly average phase voltage data weight.
[0076] In this embodiment, similarly, the monthly average phase voltage Y of the third channel m The convolution formula is as follows:
[0077]
[0078] Among them, w m is the weight of monthly average phase voltage data.
[0079] In this embodiment, similarly, the fourth channel annual average phase voltage Y t The convolution formula is as follows:
[0080]
[0081] Among them, w y is the annual average phase voltage data weight.
[0082] Set to w d =w t =w m =0.2, w y =0.4, e=0.1.
[0083] (3) Fusion layer and channel attention mechanism
[0084] Multiple fusion layers are set to stack and splice the feature dimensions of the four branches of daily average phase voltage data, weekly average phase voltage data, monthly average phase voltage data, and annual average phase voltage data, and the channel attention mechanism is used to obtain the importance of different feature information, enhance useful features according to the importance, and suppress interference.
[0085] The channel attention mechanism module includes a compression module, an excitation module and an attention module. The compression module consists of three maximum pooling layers that sum and compress the global spatial feature information in each channel to form their own channel features, which can reflect the global channel feature information, which is equivalent to expanding the network's perception field of view; the excitation module reduces the number of module parameters while enhancing the module's migration capability; the attention module performs weighted addition of the corresponding channel feature values of the original convolution and the obtained feature weights in each channel domain, so that the convolution channel features show different weights and extract key information representing the target.
[0086] (4) Output layer
[0087] Finally, different exception information is output through the output layer.
[0088] In this embodiment, the voltage transformer abnormality prediction model is trained, specifically, the daily average phase voltage data is input into the first channel, the weekly average phase voltage data is input into the second channel, the monthly average phase voltage data is input into the third channel, and the annual average phase voltage data is input into the fourth channel. When the amount of data is large, the annual average phase voltage data in the four channels is input first.
[0089] The annual average phase voltage dataset Y yThe training samples and test samples are divided into training samples and test samples at a ratio of 3:1. The training sample data is input into the fourth channel through the input layer. A 5*5 convolution kernel is used in the fourth channel for convolution calculation. During the convolution process, the Relu activation function is used for calculation. The annual average phase voltage output convolution information is obtained using formula (6).
[0090] At the same time, the daily average phase voltage data set Y d The data is divided into training samples and test samples at a ratio of 3:1. The training sample data is input into the first channel through the input layer. A 5*5 convolution kernel is used for convolution calculation in the first channel. During the convolution process, the Relu activation function is used for calculation. The daily average phase voltage output convolution information is obtained using formula (3).
[0091] At the same time, the weekly average phase voltage data set Y t The data is divided into training samples and test samples at a ratio of 3:1. The training sample data is input into the second channel through the input layer. A 5*5 convolution kernel is used for convolution calculation in the second channel. During the convolution process, the Relu activation function is used for calculation. The weekly average phase voltage output convolution information is obtained using formula (4).
[0092] At the same time, the monthly average phase voltage data set Y m The data is divided into training samples and test samples at a ratio of 3:1. The training sample data is input into the second channel through the input layer. A 5*5 convolution kernel is used for convolution calculation in the second channel. During the convolution process, the Relu activation function is used for calculation. The monthly average phase voltage output convolution information is obtained using formula (5).
[0093] The output convolution information obtained by formulas (3), (4), (5), and (6) is input into the fusion layer, and the feature dimensions of the four branches are stacked and spliced. The channel attention mechanism is introduced, and the global spatial feature information in each channel is summed and compressed through the three maximum pooling layers of the compression module in the channel attention module to form their own channel features. Then, the module migration capability is enhanced and the number of module parameters is reduced through the excitation module. Finally, the attention module performs weighted addition of the corresponding channel feature value of the original convolution and the obtained feature weight in each channel domain, so that the convolution channel features show different weights and extract the key information representing the target.
[0094] The extracted different abnormal information is output through the output layer.
[0095] After obtaining the trained voltage transformer abnormality prediction model, the test set data is input into the voltage transformer abnormality prediction model for testing.
[0096] After the above steps, a trained voltage transformer anomaly prediction model is obtained. At the same time, after testing, the accuracy of the voltage transformer anomaly prediction model is ensured, laying the foundation for the anomaly prediction of the data to be detected.
[0097] In step S4, the voltage data of the voltage transformer to be predicted is obtained, and the voltage to be detected is input into the trained voltage transformer abnormality prediction model to obtain abnormal information of the voltage transformer.
[0098] Specifically, the voltage transformer secondary voltage U to be predicted is input into the voltage transformer abnormality prediction model, and the corresponding voltage transformer abnormality information is obtained through the voltage transformer abnormality prediction model.
[0099] After the above steps, the abnormal information of the voltage transformer can be accurately predicted through the voltage transformer abnormality prediction model.
[0100] In this embodiment, in order to optimize the training of the voltage transformer abnormality prediction model, all the voltage data to be predicted and the voltage transformer abnormality information are stored in a storage device of historical data, and the model is optimized and trained to make the voltage transformer abnormality prediction model more accurate.
[0101] In step S5, the method for evaluating the acquired abnormal information of the voltage transformer is specifically to evaluate the modeling accuracy by using the root mean square error and the determination coefficient index, and the calculation formula is as follows:
[0102]
[0103] Among them, r 2 is the coefficient of determination, RMSE is the root mean square, u i is the historical voltage data, is the value to be predicted, is the voltage mean of historical data, n is the number of samples, if r 2 >3, it indicates that the sample data point is abnormal.
[0104] After the above steps, the accuracy of the voltage transformer abnormality prediction model can be accurately evaluated.
[0105] Embodiment 2
[0106] The purpose of this embodiment is to provide a voltage transformer fault diagnosis system, including:
[0107] The historical data acquisition module is configured to: acquire the secondary side voltage historical data of the voltage transformer.
[0108] The preprocessing module is configured to: preprocess the acquired historical data.
[0109] The model prediction module is configured to: construct a voltage transformer abnormality prediction model based on the preprocessed voltage historical data, and perform training and testing, obtain the voltage data to be predicted, input it into the voltage transformer abnormality prediction model, and output the voltage transformer abnormality information.
[0110] The evaluation module is configured to: evaluate the abnormality prediction of the voltage transformer.
[0111] Embodiment 3
[0112] The purpose of this embodiment is to provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the program.
[0113] Embodiment 4
[0114] The purpose of this embodiment is to provide a computer-readable storage medium.
[0115] A computer-readable storage medium stores a computer program, which executes the steps of the above method when executed by a processor.
[0116] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.
[0117] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.
[0118] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.
Claims
1. A voltage transformer abnormality prediction method, characterized in that: include: Obtaining historical data of voltage transformer secondary side voltage; Preprocessing the acquired voltage history data to obtain preprocessed voltage history data; Based on the preprocessed voltage history data, a voltage transformer abnormality prediction model is constructed; The voltage transformer abnormality prediction model is constructed by introducing a channel attention mechanism into an improved convolutional neural network containing several parallel channels, and the voltage mutual inductance abnormality prediction model is trained to obtain a trained voltage mutual inductance abnormality prediction model; The secondary side voltage of the voltage transformer to be detected is obtained, and the voltage to be detected is input into the trained voltage transformer abnormality prediction model to obtain abnormal information of the voltage transformer. The acquired abnormal information of the voltage transformer is evaluated.
2. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The voltage history data includes normal data and abnormal data, and mainly includes: daily average phase voltage data, weekly average phase voltage data, monthly average phase voltage data, and annual average phase voltage data.
3. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The process of obtaining the historical data of the secondary side voltage of the voltage transformer is as follows: discrete sampling is performed on the secondary voltage output by the voltage transformer to obtain the three-phase voltage of the secondary voltage, and the daily average phase voltage, weekly average phase voltage, monthly average phase voltage and annual average phase voltage are calculated to respectively construct a daily average phase voltage data set, a weekly average phase voltage data set, a monthly average phase voltage data set and an annual average phase voltage data set.
4. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The method for preprocessing the acquired voltage history data is: using the inverse tangent function conversion method for normalization.
5. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The method for obtaining abnormal information of voltage transformer is specifically as follows: Acquiring voltage data of a voltage transformer to be predicted, and preprocessing the voltage data; The preprocessed voltage data is input into the voltage transformer abnormality prediction model to output the abnormal information of the voltage transformer.
6. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The method for evaluating the acquired abnormal information of the voltage transformer is specifically: using the root mean square error and the determination coefficient index to evaluate the modeling accuracy, and the calculation formula is as follows: Among them, r 2 is the coefficient of determination, RMSE is the root mean square, u i is the historical voltage data, is the value to be predicted, is the voltage mean of historical data, and n is the number of samples.
7. A voltage transformer abnormality prediction method according to claim 1, characterized in that: The voltage transformer abnormality prediction model includes a daily average phase voltage convolution network channel, a weekly average phase voltage convolution network channel, a monthly average phase voltage convolution network channel, and an annual average phase voltage convolution network channel.
8. A voltage transformer abnormality prediction system, characterized in that: include: The historical data acquisition module is configured to: acquire the secondary side voltage historical data of the voltage transformer. The preprocessing module is configured to: preprocess the acquired historical data. The model prediction module is configured to: construct a voltage transformer abnormality prediction model based on the preprocessed voltage historical data, and perform training, obtain the voltage data to be predicted, input it into the voltage transformer abnormality prediction model, and output the voltage transformer abnormality information. The evaluation module is configured to: evaluate the abnormality prediction of the voltage transformer.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are performed.