A method, device, equipment and medium for online quantitative evaluation of error state of magnetic valve type current transformer

Abstract

The present invention provides a method, device, equipment and medium for online quantitative evaluation of the error state of a magnetic valve type current transformer, and relates to the technical field of error state evaluation of a magnetic valve type current transformer. The present invention combines deep learning with traditional statistical analysis techniques to collect secondary output data and environmental data of the current transformer in real time, predict noise using a deep convolutional neural network, analyze environmental errors using a multivariate linear regression model, and use a deep autoencoder to extract features to facilitate subsequent signal analysis, thereby achieving accurate evaluation of the error state of the current transformer. This method can not only monitor the performance of the current transformer in real time, but also distinguish the specific effects of noise and environmental factors on the error, providing a scientific basis for the maintenance and adjustment of the power system. It can not only improve the measurement accuracy of the current transformer, but also reduce power system failures caused by errors, and improve the operating efficiency and reliability of the power system.

Description

Technical Field

[0001] The present invention relates to the technical field of magnetic valve type current transformer error state evaluation, and in particular to an online quantitative evaluation method, device, equipment and medium for the error state of a magnetic valve type current transformer. Background Art

[0002] In the power system, the current transformer is a key measuring device, and its accuracy directly affects the stable operation of the power system and the accuracy of electric energy metering. However, in actual operation, the current transformer is often affected by many factors, such as ambient temperature, humidity changes, and internal noise, which can cause measurement errors and thus affect the performance of the power system. Traditional error assessment methods usually rely on offline calibration and regular maintenance, which cannot reflect the error state of the current transformer in real time, and it is difficult to distinguish the specific impact of different factors on the error. Therefore, developing a method that can quantitatively evaluate the error state of the current transformer online is of great significance to improving the operating efficiency and reliability of the power system.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0004] The object of the present invention is to provide an online quantitative evaluation method, device, equipment and medium for the error state of a magnetic valve type current transformer, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An online quantitative evaluation method for the error state of a magnetic valve type current transformer, the specific steps comprising:

[0007] S1: Collect historical data of current transformers, then build a noise prediction model for noise analysis and a multivariate linear regression model for environmental analysis, and use historical data to optimize the noise prediction model and the multivariate linear regression model respectively;

[0008] S2: Take the preset standard data as input, use the deep autoencoder to extract its features, obtain the corresponding feature vector and calculate its reconstruction error, and optimize the model parameters of the deep autoencoder by minimizing the reconstruction error;

[0009] S3: Collect real-time data of the current transformer, process the data through the noise prediction model and the multivariate linear regression model in turn, and then use the deep autoencoder to extract features, and at the same time determine the working status and error type of the current transformer.

[0010] Preferably, the historical data includes historical secondary output data and historical environmental data, the historical secondary output data includes historical noisy data under actual conditions and historical noise-free data for reference, the historical secondary output data and historical environmental data are used as input data for standardized processing, and the standardized historical secondary output data and historical environmental data are divided into a first training set and a first verification set, and the first training set and the first verification set are used to train the noise prediction model.

[0011] Preferably, the corresponding environmental error is calculated based on the historical noise-free data and the preset standard data, the historical noise-free data and the historical environmental data are divided into an independent variable set, and the independent variable set is divided into a second training set and a second validation set, and the second training set and the second validation set are used to optimize the multivariate linear regression model.

[0012] Preferably, the real-time data includes real-time secondary output data and real-time environmental data. The real-time secondary output data and real-time environmental data are standardized as input data, and then substituted into a noise prediction model to calculate and eliminate their noise signals to obtain real-time noise-free data. Subsequently, the real-time noise-free data and real-time environmental data are substituted into a multivariate linear regression model to generate environmental errors and eliminate them to obtain theoretical data. Finally, a deep autoencoder is used to extract features from the theoretical data to obtain corresponding feature vectors for subsequent signal analysis.

[0013] Preferably, the historical noisy data is calibrated as , , the historical noise-free data is calibrated as , , the actual noise signal is calibrated as , The relationship among the three is expressed as:

[0014] ;

[0015] ;

[0016] In the formula , , Respectively represent the elements in the corresponding data set, subscript Indicates the data number. Represents the number of elements in a data set;

[0017] Then the real-time secondary output data is calibrated as , , real-time noise-free data calibration is , , the relationship between the two is expressed as:

[0018] ;

[0019] ;

[0020] In the formula represents the noise signal predicted by the noise prediction model, .

[0021] Preferably, the logic of the standardization process is:

[0022] Calculate the mean and standard deviation of the input data as follows:

[0023] ;

[0024] ;

[0025] Then the input data is standardized and the calculation method is:

[0026] ;

[0027] In the formula Represents input data, represents the mean of the input data, represents the standard deviation of the input data, Represents the output data after normalization.

[0028] Preferably, the logic for optimizing the noise prediction model is:

[0029] A noise prediction model is constructed based on building a deep convolutional neural network, using random values ​​as the initial model parameters;

[0030] The historical noisy data in the first training set is used as input data for standardization to generate the corresponding output data And input it into the noise prediction model to calculate the corresponding noise signal;

[0031] Construct a loss function to measure the difference between the noise signal output by the noise prediction model and the actual noise signal. The calculation method is:

[0032] ;

[0033] Then calculate the gradient according to the loss function, and use the back propagation algorithm to update the model parameters of the noise prediction model. The calculation method is:

[0034] ;

[0035] In the formula represents the model parameters of the noise prediction model, Represents the learning rate of the noise prediction model;

[0036] The model parameters of the noise prediction model are iterated multiple times using the first training set. When the loss function is less than or equal to a preset function threshold, the optimization is stopped and the noise prediction model is considered to be optimized.

[0037] Preferably, the historical environmental data includes historical temperature data and historical humidity data, and the multivariate linear regression model for environmental error is expressed as:

[0038] ;

[0039] In the formula , , They represent the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multivariate linear regression model respectively. represents the adjustment constant, Indicates environmental error;

[0040] The environmental error is obtained by comparing the historical noise-free data with the preset standard data, and the calculation method is:

[0041] ;

[0042] In the formula Represents preset standard data, all standard data together constitute a set , ;

[0043] Then substitute the second training set into the multiple linear regression model, use the nonlinear least squares method to solve the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multiple linear regression model, and adjust the constant to minimize the cumulative error of the multiple linear regression model. The cumulative error is calculated as follows:

[0044] ;

[0045] In the formula represents the cumulative error of the multivariate linear regression model, Represents the environmental error predicted by the multivariate linear regression model.

[0046] Preferably, the reconstruction error is calculated as follows:

[0047] ;

[0048] In the formula represents the reconstruction error, represents the model parameters of the deep autoencoder, Represents the use of deep autoencoder Perform feature extraction.

[0049] Preferably, when the real-time secondary output data of the current transformer and the real-time environmental data are processed, the total error rate is calculated according to the corresponding noise signal and the predicted environmental error. and noise ratio , the calculation methods are:

[0050] ;

[0051] ;

[0052] By time Determine as interval and calculate time The average total error rate within and average noise ratio , the calculation method is:

[0053] ;

[0054] ;

[0055] In the formula Indicates time The number of samples within ;

[0056] If in time The average total error rate , it is considered that the current transformer has a small error and works normally;

[0057] If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by the environment, and has an abnormal working state;

[0058] If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by noise, and has an abnormal working state;

[0059] In the formula represents the preset total error threshold, Indicates the preset noise ratio threshold.

[0060] An online quantitative evaluation device for the error state of a magnetic valve type current transformer, the evaluation device adopts the above-mentioned evaluation method, comprising:

[0061] A data acquisition module, which is a logic circuit electrically connected to the data analysis module and is used to collect historical secondary output data and real-time secondary output data measured by the current transformer and send them to the data analysis module;

[0062] A sensor module, which is electrically connected to the data analysis module and includes a temperature sensor and a humidity sensor, and is used to collect historical environmental data and real-time environmental data of the environment where the current transformer is located and send the data analysis module;

[0063] A data analysis module, the data analysis module is electrically connected to the feature extraction module, the error analysis module, and the data storage module, and is used to construct a noise prediction model and a multiple linear regression model according to historical secondary output data and historical environmental data, and substitute the real-time secondary output data and real-time environmental data into the noise prediction model and the multiple linear regression model to generate theoretical data, and then send the calculated noise signal and environmental error to the error analysis module, and also send the historical secondary output data, historical environmental data, real-time secondary output data, and real-time environmental data to the data storage module for storage;

[0064] An error analysis module, wherein the error analysis module is used to determine the working state and error type of the current transformer according to the noise signal and the environmental error;

[0065] A feature extraction module, which is electrically connected to the signal transmission module and is used to optimize the model parameters of the deep autoencoder according to preset standard data, and use the optimized deep autoencoder to extract features from the theoretical data and send the features to the signal transmission module;

[0066] A signal sending module is used to send characteristic data extracted from theoretical data to subsequent ports for signal analysis.

[0067] An online quantitative evaluation device for the error state of a magnetic valve type current transformer, the evaluation device comprising:

[0068] Storage media for storing computer programs;

[0069] A processor is used to execute the computer program to implement the above-mentioned evaluation method.

[0070] A medium for online quantitative evaluation of the error state of a magnetic valve type current transformer, the medium being used to store a computer program, and the computer program implementing the above-mentioned evaluation method when executed by a processor.

[0071] Compared with the prior art, the present invention has the following beneficial effects:

[0072] By combining deep learning and traditional statistical analysis techniques, the present invention can collect the secondary output data and environmental data of the current transformer in real time, use a deep convolutional neural network to predict noise, analyze environmental errors through a multivariate linear regression model, and use a deep autoencoder to extract features to facilitate subsequent signal analysis, thereby accurately evaluating the error state of the current transformer. This method can not only monitor the performance of the current transformer in real time, but also distinguish the specific impact of noise and environmental factors on the error, providing a scientific basis for the maintenance and adjustment of the power system. It can not only improve the measurement accuracy of the current transformer, but also reduce power system failures caused by errors, and improve the operating efficiency and reliability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0074] Figure 2 It is a schematic diagram of the module structure of the present invention. DETAILED DESCRIPTION

[0075] 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 specific embodiments.

[0076] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0077] Example:

[0078] See also Figure 1 , the present invention provides a technical solution:

[0079] An online quantitative evaluation method for the error state of a magnetic valve type current transformer, the specific steps comprising:

[0080] S1: Collect historical secondary output data and historical environmental data of the current transformer, where the historical secondary output data includes historical noisy data under actual conditions and historical noise-free data for reference. The historical secondary output data and historical environmental data are used as input data for standardization, and then a noise prediction model is constructed based on a deep convolutional neural network. The standardized historical secondary output data is divided into a first training set and a first validation set, and the first training set and the first validation set are used to train and optimize the noise prediction model.

[0081] In this step, the historical noisy data can be measured by connecting the current transformer to the actual working circuit, while the historical noise-free data for reference can be obtained by connecting the current transformer to the test circuit under laboratory conditions to minimize the interference of noise. The noise signal mainly comes from the electromagnetic interference of the power line, the adjacent equipment, or due to the instability of the internal components of the current transformer. The deep convolutional neural network has a strong feature extraction capability and can capture the subtle changes of noise from complex data. Compared with traditional noise processing methods, such as simple filtering technology, it can provide more accurate noise prediction. In addition, the CNN model can automatically adjust parameters during the training process to adapt to the noise characteristics under different conditions, which makes the model have strong generalization ability and can cope with a variety of different noise environments. By removing noise, the quality of the secondary output data is improved, providing a purer data basis for subsequent environmental error analysis and feature extraction. Accurate noise prediction helps to distinguish the influence of noise and environmental factors on the current transformer error, providing a clearer perspective for error analysis, and helping to more accurately judge the working state and error source of the current transformer.

[0082] The historical noisy data is calibrated as , , the historical noise-free data is calibrated as , , the actual noise signal is calibrated as , The relationship among the three is expressed as:

[0083] ;

[0084] ;

[0085] In the formula , , Respectively represent the elements in the corresponding data set, subscript Indicates the data number. Represents the number of elements in a data set;

[0086] Then the real-time secondary output data is calibrated as , , real-time noise-free data calibration is , , the relationship between the two is expressed as:

[0087] ;

[0088] ;

[0089] In the formula represents the noise signal predicted by the noise prediction model, .

[0090] The logic for optimizing the noise prediction model is:

[0091] A noise prediction model is constructed based on building a deep convolutional neural network, using random values ​​as the initial model parameters;

[0092] The historical noisy data in the first training set is used as input data for standardization to generate the corresponding output data And input it into the noise prediction model to calculate the corresponding noise signal;

[0093] Construct a loss function to measure the difference between the noise signal output by the noise prediction model and the actual noise signal. The calculation method is:

[0094] ;

[0095] Then calculate the gradient according to the loss function, and use the back propagation algorithm to update the model parameters of the noise prediction model. The calculation method is:

[0096] ;

[0097] In the formula represents the model parameters of the noise prediction model, Represents the learning rate of the noise prediction model;

[0098] The model parameters of the noise prediction model are iterated multiple times using the first training set. When the loss function is less than or equal to a preset function threshold, the optimization is stopped and the noise prediction model is considered to be optimized.

[0099] By adopting the back-propagation algorithm and gradient descent optimization, the model can automatically adjust parameters to adapt to different noise characteristics, which makes the model have strong generalization ability and can cope with a variety of different noise environments, which helps to enhance the accuracy of the noise prediction model and make the entire error evaluation process more accurate and reliable.

[0100] The logic of standardization processing is:

[0101] Calculate the mean and standard deviation of the input data as follows:

[0102] ;

[0103] ;

[0104] Then the input data is standardized and the calculation method is:

[0105] ;

[0106] In the formula Represents input data, represents the mean of the input data, represents the standard deviation of the input data, Represents the output data after normalization.

[0107] In the process of standardization, the input data is not a fixed set or data, but represents the sum of all data that need to be standardized, which can be historical secondary output data, historical environmental data, real-time secondary output data or real-time environmental data. By converting the input data into a standard form with zero mean and unit variance, it helps to eliminate the dimensional differences between the data, so that data with different characteristics can be compared and analyzed on the same scale. This can not only speed up the convergence of the machine learning model and improve the training efficiency, but also better capture the patterns in the data, reduce the model deviation caused by different data scales, and thus improve the generalization ability of the model.

[0108] S2: Calculate the corresponding environmental error based on the historical noise-free data and the preset standard data, divide the historical noise-free data and the historical environmental data into an independent variable set, construct a multivariate linear regression model about the environmental error based on the independent variable set, and divide the independent variable set into a second training set and a second validation set, and use the second training set and the second validation set to optimize the multivariate linear regression model.

[0109] In this step, since the environmental error is affected by many factors, a multivariate linear regression model is used for analysis. Compared with the single variable analysis method, it can provide a more comprehensive and accurate environmental error analysis. Fluctuations in ambient temperature can affect the resistance and magnetism of the internal materials of the current transformer, thereby affecting its performance. High humidity environments may cause corrosion inside the current transformer or at the terminals, affecting the stability of the electrical connection. Therefore, the environmental error is generated differently from the noise signal, and needs to be separated and analyzed in different ways. The preset standard data is theoretical data for reference. In the initial process of optimizing the model parameters, it can be obtained through simulation or actual measurement using a high-precision current meter. In subsequent actual use, this value cannot be obtained directly, and can only be restored as much as possible through the algorithm to obtain the value closest to the theoretical calculation, which is the theoretical data mentioned later.

[0110] The multivariate linear regression model for environmental error is expressed as:

[0111] ;

[0112] In the formula , , They represent the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multivariate linear regression model respectively. represents the adjustment constant, Indicates environmental error;

[0113] The environmental error is obtained by comparing the historical noise-free data with the preset standard data, and the calculation method is:

[0114] ;

[0115] In the formula Represents preset standard data, all standard data together constitute a set , , the same as historical secondary output data, historical environmental data, real-time secondary output data, real-time environmental data, and theoretical data, standard data is also a data set that reflects the data changes within a sampling time period, and the elements represent the data collected at different sampling time points within the period. It should be noted that the sampling time and sampling interval of each group of different data should be set to be consistent, so that the number of elements in different data sets is equal and corresponds one by one according to the data number.

[0116] Then substitute the second training set into the multiple linear regression model, use the nonlinear least squares method to solve the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multiple linear regression model, and adjust the constant to minimize the cumulative error of the multiple linear regression model. The cumulative error is calculated as follows:

[0117] ;

[0118] In the formula represents the cumulative error of the multivariate linear regression model, Represents the environmental error predicted by the multivariate linear regression model.

[0119] The model parameters are optimized by nonlinear least squares method to ensure that the difference between the environmental error predicted by the model and the actual error is minimized, thereby improving the prediction accuracy and generalization ability of the model.

[0120] S3: Take the preset standard data as input, use the deep autoencoder to extract its features, obtain the corresponding feature vector and calculate its reconstruction error, and optimize the model parameters of the deep autoencoder by minimizing the reconstruction error.

[0121] The reconstruction error is calculated as:

[0122] ;

[0123] In the formula represents the reconstruction error, represents the model parameters of the deep autoencoder, Represents the use of deep autoencoder Perform feature extraction.

[0124] By using deep autoencoders for feature extraction, high-dimensional current data can be compressed into low-dimensional feature representations while retaining the main information of the data. This helps reduce the complexity and processing difficulty of the data, makes subsequent signal analysis more convenient, and improves response time and analysis speed.

[0125] S4: Collect the real-time secondary output data and real-time environmental data of the current transformer, standardize the real-time secondary output data and real-time environmental data as input data, and then substitute them into the noise prediction model to calculate and eliminate their noise signals to obtain real-time noise-free data. Then substitute the real-time noise-free data and real-time environmental data into the multivariate linear regression model, generate environmental errors and eliminate them to obtain theoretical data. Finally, use the deep autoencoder to extract features from the theoretical data to obtain the corresponding feature vector for subsequent signal analysis.

[0126] In this step, the noise signal and environmental error in the real-time secondary data are eliminated by using the optimized noise prediction model and multivariate linear regression model to obtain purer theoretical data. The deep autoencoder is then used to extract features from the theoretical data to extract key information from the data. These feature vectors are of great value for signal analysis and pattern recognition. By integrating multiple groups of models, the entire process from raw data to the final feature vector can be processed and optimized, improving the consistency and efficiency of data processing.

[0127] When processing the real-time secondary output data of the current transformer and the real-time environmental data, the total error rate is calculated based on the corresponding noise signal and the predicted environmental error. and noise ratio , the calculation methods are:

[0128] ;

[0129] ;

[0130] By time Determine as interval and calculate time The average total error rate within and average noise ratio , the calculation method is:

[0131] ;

[0132] ;

[0133] In the formula Indicates time The number of samples within .

[0134] If in time The average total error rate , it is considered that the current transformer has a small error and works normally;

[0135] If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by the environment, and has an abnormal working state;

[0136] If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by noise, and has an abnormal working state;

[0137] In the formula represents the preset total error threshold, Indicates the preset noise ratio threshold.

[0138] In this step, by calculating the total error rate and the noise ratio, the working status of the current transformer can be accurately evaluated, and the source of the error can be distinguished as environmental factors or noise influence. The cause of the current transformer failure with large error can be diagnosed more accurately, which is convenient for users to adjust the maintenance strategy in a targeted manner, such as taking environmental control measures when the environmental impact is large, and performing noise suppression when the noise impact is large, thereby improving maintenance efficiency and reducing costs.

[0139] See also Figure 2 The present invention also provides an online quantitative evaluation device for the error state of a magnetic valve type current transformer, wherein the evaluation device adopts the above-mentioned evaluation method and comprises:

[0140] A data acquisition module, which is a logic circuit electrically connected to the data analysis module and is used to collect historical secondary output data and real-time secondary output data measured by the current transformer and send them to the data analysis module;

[0141] A sensor module, which is electrically connected to the data analysis module and includes a temperature sensor and a humidity sensor, and is used to collect historical environmental data and real-time environmental data of the environment where the current transformer is located and send the data analysis module;

[0142] A data analysis module, the data analysis module is electrically connected to the feature extraction module, the error analysis module, and the data storage module, and is used to construct a noise prediction model and a multiple linear regression model according to historical secondary output data and historical environmental data, and substitute the real-time secondary output data and real-time environmental data into the noise prediction model and the multiple linear regression model to generate theoretical data, and then send the calculated noise signal and environmental error to the error analysis module, and also send the historical secondary output data, historical environmental data, real-time secondary output data, and real-time environmental data to the data storage module for storage;

[0143] An error analysis module, wherein the error analysis module is used to determine the working state and error type of the current transformer according to the noise signal and the environmental error;

[0144] A feature extraction module, which is electrically connected to the signal transmission module and is used to optimize the model parameters of the deep autoencoder according to preset standard data, and use the optimized deep autoencoder to extract features from the theoretical data and send the features to the signal transmission module;

[0145] A signal sending module is used to send characteristic data extracted from theoretical data to subsequent ports for signal analysis.

[0146] An online quantitative evaluation device for the error state of a magnetic valve type current transformer, the evaluation device comprising:

[0147] Storage media for storing computer programs;

[0148] A processor is used to execute the computer program to implement the above-mentioned evaluation method.

[0149] A medium for online quantitative evaluation of the error state of a magnetic valve type current transformer, the medium being used to store a computer program, and the computer program implementing the above-mentioned evaluation method when executed by a processor.

[0150] In summary, the present invention proposes an online quantitative evaluation method for the error state of a magnetic valve type current transformer by combining deep learning and traditional statistical analysis techniques. The method can collect the secondary output data and environmental data of the current transformer in real time, predict the noise using a deep convolutional neural network, analyze the environmental error through a multivariate linear regression model, and extract features using a deep autoencoder, thereby achieving an accurate evaluation of the error state of the current transformer. This method can not only monitor the performance of the current transformer in real time, but also distinguish the specific effects of noise and environmental factors on the error, providing a scientific basis for the maintenance and adjustment of the power system. Through the present invention, the measurement accuracy of the current transformer can be significantly improved, the power system failures caused by errors can be reduced, the operating efficiency and reliability of the power system can be improved, and significant economic and social benefits can be achieved.

[0151] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0152] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0153] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0154] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. An online quantitative evaluation method for the error state of a magnetic valve current transformer, characterized in that: The specific steps include: S1: Collect historical data of current transformers, then build a noise prediction model for noise analysis and a multivariate linear regression model for environmental analysis, and use historical data to optimize the noise prediction model and the multivariate linear regression model respectively; The historical data includes historical secondary output data and historical environmental data, the historical secondary output data includes historical noisy data under actual conditions and historical noise-free data for reference, the historical secondary output data and the historical environmental data are used as input data for standardization, and the standardized historical secondary output data and the historical environmental data are divided into a first training set and a first validation set, and the first training set and the first validation set are used to train the noise prediction model; Calculate the corresponding environmental error according to the historical noise-free data and the preset standard data, divide the historical noise-free data and the historical environmental data into an independent variable set, and divide the independent variable set into a second training set and a second validation set, and use the second training set and the second validation set to optimize the multivariate linear regression model; S2: Take the preset standard data as input, use the deep autoencoder to extract its features, obtain the corresponding feature vector and calculate its reconstruction error, and optimize the model parameters of the deep autoencoder by minimizing the reconstruction error; S3: Collect real-time data of the current transformer, process the data through the noise prediction model and the multivariate linear regression model, and then use the deep autoencoder to extract features and judge the working status and error type of the current transformer; The real-time data includes real-time secondary output data and real-time environmental data. The real-time secondary output data and real-time environmental data are standardized as input data, and then substituted into a noise prediction model to calculate and remove the noise signal to obtain real-time noise-free data. Then, the real-time noise-free data and real-time environmental data are substituted into a multivariate linear regression model to generate environmental errors and remove them to obtain theoretical data. Finally, a deep autoencoder is used to extract features from the theoretical data to obtain corresponding feature vectors for subsequent signal analysis. Time All historical noisy data in are calibrated as , , the historical noise-free data is calibrated as , , the actual noise signal is calibrated as , The relationship among the three is expressed as: In the formula , , Respectively represent the elements in the corresponding data set, each element represents the time The data obtained by each sampling in Indicates the sampling number, Indicates time The number of samples within ; Then the real-time secondary output data is calibrated as , , real-time noise-free data calibration is , , the relationship between the two is expressed as: In the formula represents the noise signal predicted by the noise prediction model, ; The logic of standardization processing is: Calculate the mean and standard deviation of the input data as follows: Then the input data is standardized and the calculation method is: In the formula Represents input data, represents the mean of the input data, represents the standard deviation of the input data, It represents the output data after standardization. The input data includes historical secondary output data, historical environmental data, real-time secondary output data, and real-time environmental data. The logic for optimizing the noise prediction model is: A noise prediction model is constructed based on building a deep convolutional neural network, using random values ​​as the initial model parameters; The historical noisy data in the first training set is used as input data for standardization to generate the corresponding output data And input it into the noise prediction model to calculate the corresponding noise signal; Construct a loss function to measure the difference between the noise signal output by the noise prediction model and the actual noise signal. The calculation method is: Then calculate the gradient according to the loss function, and use the back propagation algorithm to update the model parameters of the noise prediction model. The calculation method is: In the formula represents the model parameters of the noise prediction model, including the weight and bias of the noise prediction model, Represents the learning rate of the noise prediction model; The model parameters of the noise prediction model are iterated multiple times using the first training set. When the loss function is less than or equal to a preset function threshold, the optimization is stopped and the noise prediction model is considered to be optimized.

2. The method for online quantitative evaluation of the error state of a magnetic valve type current transformer according to claim 1, characterized in that: The historical environmental data includes historical temperature data and historical humidity data. The multivariate linear regression model of environmental error is expressed as: In the formula , , They represent the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multivariate linear regression model respectively. represents the adjustment constant, Indicates environmental error; The environmental error is obtained by comparing the historical noise-free data with the preset standard data, and the calculation method is: In the formula Represents preset standard data, all standard data together constitute a set , .

3. The method for online quantitative evaluation of the error state of a magnetic valve type current transformer according to claim 2, characterized in that: When optimizing the multiple linear regression model, the second training set is substituted into the multiple linear regression model, and the first equation coefficient, the second equation coefficient, and the third equation coefficient of the multiple linear regression model are solved using the nonlinear least squares method. The constants are adjusted to minimize the cumulative error of the multiple linear regression model. The cumulative error is calculated as follows: In the formula represents the cumulative error of the multivariate linear regression model, Represents the environmental error predicted by the multivariate linear regression model.

4. The method for online quantitative evaluation of the error state of a magnetic valve type current transformer according to claim 1, characterized in that: The reconstruction error is calculated as: In the formula represents the reconstruction error, represents the model parameters of the deep autoencoder, Represents the use of deep autoencoder Perform feature extraction.

5. The method for online quantitative evaluation of the error state of a magnetic valve type current transformer according to claim 3, characterized in that: When processing the real-time secondary output data of the current transformer and the real-time environmental data, the total error rate is calculated based on the corresponding noise signal and the predicted environmental error. and noise ratio , the calculation methods are: By time Determine as interval and calculate time The average total error rate within and average noise ratio , the calculation method is: In the formula Indicates time The number of samples within .

6. The method for online quantitative evaluation of the error state of a magnetic valve type current transformer according to claim 5, characterized in that: According to the average total error rate , average noise ratio The logic for judging the error state of the current transformer is: If in time The average total error rate , it is considered that the current transformer has a small error and works normally; If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by the environment, and has an abnormal working state; If in time The average total error rate , and the average noise ratio , it is considered that the current transformer has a large error, is mainly affected by noise, and has an abnormal working state; In the formula represents the preset total error threshold, Indicates the preset noise ratio threshold.

7. An online quantitative evaluation device for the error state of a magnetic valve type current transformer, characterized in that: The evaluation device adopts the evaluation method according to any one of claims 1 to 6, including: A data acquisition module, which is a logic circuit electrically connected to the data analysis module and is used to collect historical secondary output data and real-time secondary output data measured by the current transformer and send them to the data analysis module; A sensor module, which is electrically connected to the data analysis module and includes a temperature sensor and a humidity sensor, and is used to collect historical environmental data and real-time environmental data of the environment where the current transformer is located and send the data analysis module; A data analysis module, the data analysis module is electrically connected to the feature extraction module, the error analysis module, and the data storage module, and is used to construct a noise prediction model and a multiple linear regression model according to historical secondary output data and historical environmental data, and substitute the real-time secondary output data and real-time environmental data into the noise prediction model and the multiple linear regression model to generate theoretical data, and then send the calculated noise signal and environmental error to the error analysis module, and also send the historical secondary output data, historical environmental data, real-time secondary output data, and real-time environmental data to the data storage module for storage; An error analysis module, wherein the error analysis module is used to determine the working state and error type of the current transformer according to the noise signal and the environmental error; A feature extraction module, which is electrically connected to the signal transmission module and is used to optimize the model parameters of the deep autoencoder according to preset standard data, and use the optimized deep autoencoder to extract features from the theoretical data and send the features to the signal transmission module; A signal sending module is used to send characteristic data extracted based on theoretical data to subsequent ports for signal analysis.

8. An online quantitative evaluation device for the error state of a magnetic valve type current transformer, characterized in that: The evaluation equipment comprises: Storage media for storing computer programs; A processor, configured to execute the computer program to implement the evaluation method according to any one of claims 1 to 6.

9. An online quantitative evaluation medium for the error state of a magnetic valve type current transformer, characterized in that: The medium is used to store a computer program, and when the computer program is executed by a processor, the evaluation method according to any one of claims 1 to 6 is implemented.

Images