Voltage transformer measurement state identification method, system, device and storage medium

Abstract

The present invention relates to a method, system, device and storage medium for identifying the measurement state of a voltage transformer, wherein the method comprises the following steps: collecting the secondary side voltage signal of a voltage transformer sample and preprocessing it to obtain a sample feature vector; inputting the feature vectors of positive and negative samples into a support vector machine model, and adjusting the penalty coefficient according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, and iterating to obtain an optimized support vector machine; collecting the secondary side voltage of a target voltage transformer to be identified and preprocessing it to obtain a feature vector of the target voltage transformer, and inputting the feature vector of the target voltage transformer into the optimized support vector machine to obtain a measurement state identification result; the preprocessing method is specifically as follows: processing the secondary side voltage signal to obtain segmented time domain features and frequency domain features; performing cumulative transformation processing on the segmented time domain features and then performing feature fusion with the frequency domain features to obtain a feature vector.

Description

Technical Field

[0001] The invention relates to a voltage transformer measurement state identification method, system, equipment and storage medium, belonging to the technical field of transformer state identification. Background Art

[0002] With the continuous expansion of the scale of power systems and the continuous improvement of voltage levels, the demand for high-precision and high-stability voltage measurement is becoming more and more urgent. As an important device in the power system, the performance and reliability of the voltage transformer for metering are directly related to the safe, stable and economical operation of the power system. Therefore, accurate and rapid detection and identification of voltage transformer anomalies are of great significance to ensure the normal operation of the power system. At present, the detection and identification of anomalies of voltage transformers for metering are mainly determined by detecting the electrical performance parameters of the voltage transformer, such as excitation current, error curve, etc. This method requires regular testing and is affected by many factors, such as voltage level and load conditions. The vibration sensor is used to detect the vibration signal of the voltage transformer during operation, and the abnormality is determined by analyzing the characteristics of the vibration signal. This method has the advantages of non-invasiveness and remote detection, but it has high requirements on the accuracy and stability of the sensor, and still has some limitations and shortcomings. Therefore, further research and exploration of new detection methods and technologies to improve the accuracy and reliability of voltage transformer anomaly identification is an important research direction at present.

[0003] A Chinese invention patent with publication number CN114818817A discloses a method for identifying weak faults of capacitive voltage transformers. The method uses CVT operation SCADA three-phase voltage data as the analysis object, constructs a time domain jump convolution to extract the spatiotemporal features of the SCADA three-phase voltage and converts it into a one-dimensional spatiotemporal feature sequence, and uses a feature enhancement mechanism to adaptively enhance the spatiotemporal fault features. The Bi-LSTM network is then used to fully extract and analyze the time domain features from both positive and negative directions, and the hidden layer feature vector is obtained and input into the fully connected layer to obtain the next data prediction value; then the NMSE value between the predicted value and the actual value is calculated, and the adaptive threshold is set by the exponentially weighted moving average method. Finally, the CVT fault is judged by comparing the NMSE value with the threshold.

[0004] The above existing methods can fully extract the spatiotemporal characteristics of CVT operation SCADA data, capture the changing trend of CVT operation voltage data, have strong recognition ability for weak faults, and achieve accurate prediction of CVT faults. However, the above existing methods do not take into account that CVT feature data contains a large number of features. Only feature enhancement processing cannot eliminate the influence of low-correlation features on model recognition accuracy. How to construct appropriate input features to train machine learning models is an urgent problem to be solved. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes a voltage transformer measurement state identification method, system, device and storage medium.

[0006] The technical solution of the present invention is as follows:

[0007] On the one hand, the present invention provides a method for identifying a voltage transformer measurement state, comprising the following steps:

[0008] Collecting the secondary side voltage signal of the voltage transformer sample and preprocessing it to obtain the sample feature vector;

[0009] The voltage transformer samples in normal and abnormal measurement states are taken as positive and negative samples respectively, and the feature vectors of the positive and negative samples are input into the support vector machine model. The penalty coefficient of the support vector machine is adjusted according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, and the model is iterated to obtain the optimized support vector machine.

[0010] The secondary side voltage of the target voltage transformer to be identified is collected and preprocessed to obtain a feature vector of the target voltage transformer, and the feature vector of the target voltage transformer is input into the optimized support vector machine to obtain a measurement state identification result;

[0011] The specific preprocessing method is as follows:

[0012] Process the secondary side voltage signal to obtain the time domain characteristics and frequency domain characteristics after segmentation;

[0013] The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector.

[0014] As a preferred implementation, the method for performing cumulative transformation processing on the segmented time domain features is specifically as follows:

[0015] The features of the same type at different times in the segmented time domain features are cumulatively transformed, specifically:

[0016] ;

[0017] ;

[0018] ;

[0019] Where k is the number of signal segments, N is the total length of the signal, and b is the length of the segment. is the correction factor, N 0 means the set standard signal length. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation;

[0020] The initial cumulative transformation result after the cumulative transformation is cumulatively transformed again to obtain the final cumulative transformation result:

[0021] ;

[0022] in, It is the final cumulative transformation result of the j-th feature when the signal length is divided into k segments.

[0023] As a preferred implementation, the method of fusing the cumulative transformation processing results with the frequency domain features to obtain the feature vector is specifically as follows:

[0024] Get the cumulative transformation processing results and frequency domain features as the input set:

[0025] ;

[0026] in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform;

[0027] After normalizing the input set, the normalized results are sequentially convolved and pooled to obtain a fused feature set;

[0028] Compare the similarities of each fusion feature in the fusion feature set at the same measurement state at different times to determine the weight of each fusion feature:

[0029] ;

[0030] in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features.

[0031] Feature screening is performed based on the weights of each fusion feature:

[0032]

[0033] in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold;

[0034] Perform feature fusion to obtain the feature vector:

[0035] ;

[0036] in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

[0037] As a preferred implementation, the steps of adjusting the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy and performing model iteration are specifically as follows:

[0038] The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept;

[0039] Get the geometric distance from the sample to the hyperplane:

[0040] ;

[0041] ;

[0042] in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples;

[0043] Divide the distance levels of positive and negative samples to the hyperplane;

[0044] ;

[0045] in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively;

[0046] Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, adjust the penalty coefficient:

[0047] ;

[0048] in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample.

[0049] On the other hand, the present invention also provides a voltage transformer measurement state identification system, comprising:

[0050] A preprocessing module, used for collecting the secondary side voltage signal of the voltage transformer sample and performing preprocessing to obtain the sample feature vector;

[0051] The model training module is used to take voltage transformer samples in normal and abnormal measurement states as positive and negative samples respectively, input the feature vectors of the positive and negative samples into the support vector machine model, and adjust the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, perform model iteration, and obtain the optimized support vector machine;

[0052] A state recognition module is used to collect the secondary side voltage of the target voltage transformer to be identified and perform preprocessing to obtain the feature vector of the target voltage transformer, input the feature vector of the target voltage transformer into the optimized support vector machine, and obtain the measurement state recognition result;

[0053] Among them, the method for preprocessing in the preprocessing module is specifically as follows:

[0054] Process the secondary side voltage signal to obtain the time domain characteristics and frequency domain characteristics after segmentation;

[0055] The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector.

[0056] As a preferred implementation, the method in which the preprocessing module performs cumulative transformation processing on the segmented time domain features is specifically as follows:

[0057] The features of the same type at different times in the segmented time domain features are cumulatively transformed, specifically:

[0058] ;

[0059] ;

[0060] ;

[0061] Where k is the number of signal segments, N is the total length of the signal, and b is the length of the segment. is the correction factor, N 0 means the set standard signal length. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation;

[0062] The initial cumulative transformation result after the cumulative transformation is cumulatively transformed again to obtain the final cumulative transformation result:

[0063] ;

[0064] in, It is the final cumulative transformation result of the j-th feature when the signal length is divided into k segments.

[0065] As a preferred implementation, the method of fusing the cumulative transformation processing results with the frequency domain features to obtain the feature vector is specifically as follows:

[0066] Get the cumulative transformation processing results and frequency domain features as the input set:

[0067] ;

[0068] in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform;

[0069] After normalizing the input set, the normalized results are sequentially convolved and pooled to obtain a fused feature set;

[0070] Compare the similarities of each fusion feature in the fusion feature set at the same measurement state at different times to determine the weight of each fusion feature:

[0071] ;

[0072] in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features.

[0073] Feature screening is performed based on the weights of each fusion feature:

[0074]

[0075] in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold;

[0076] Perform feature fusion to obtain the feature vector:

[0077] ;

[0078] in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

[0079] As a preferred implementation, the model training module adjusts the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, and the steps of performing model iteration are specifically as follows:

[0080] The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept;

[0081] Get the geometric distance from the sample to the hyperplane:

[0082] ;

[0083] ;

[0084] in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples;

[0085] Divide the distance levels of positive and negative samples to the hyperplane;

[0086] ;

[0087] in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively;

[0088] Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, adjust the penalty coefficient:

[0089] ;

[0090] in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample.

[0091] On the other hand, the present invention further proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for identifying the measurement state of a voltage transformer as described in any embodiment of the present invention is implemented.

[0092] On the other hand, the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the voltage transformer measurement state identification method as described in any embodiment of the present invention.

[0093] The present invention has the following beneficial effects:

[0094] The present invention can distinguish different measurement states of the transformer more prominently by cumulatively transforming the time domain characteristics of the output signal on the secondary side of the voltage transformer, and at the same time fuse it with the frequency domain characteristics so that it can contain more comprehensive information, providing a data basis for the recognition model. The penalty coefficient of the support vector machine model is adjusted adaptively to suppress the hyperplane offset phenomenon caused by the imbalance of the number of positive and negative samples, thereby improving the recognition accuracy of the support vector machine model. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] Figure 1 The figure is a schematic diagram of the method flow of the first embodiment of the present invention. DETAILED DESCRIPTION

[0096] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0097] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0098] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0099] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0100] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0101] Embodiment 1:

[0102] See also Figure 1 This embodiment proposes a voltage transformer measurement state identification method, comprising the following steps:

[0103] S100, collecting the secondary side voltage signal of the voltage transformer sample and preprocessing it to obtain the sample feature vector;

[0104] S200, taking voltage transformer samples in normal and abnormal measurement states as positive and negative samples respectively, inputting feature vectors of the positive and negative samples into a support vector machine model, and adjusting the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, continuously iterating the model to achieve model convergence, and obtaining an optimized support vector machine;

[0105] S300, collecting the secondary side voltage of the target voltage transformer to be identified and performing preprocessing to obtain a feature vector of the target voltage transformer, inputting the feature vector of the target voltage transformer into an optimized support vector machine, and obtaining a measurement state identification result;

[0106] The specific preprocessing method is as follows:

[0107] The secondary side voltage signal is processed to obtain the time domain characteristics and frequency domain characteristics after segmentation, specifically, the time domain statistical characteristics mean, standard deviation, waveform factor, variance, maximum value, minimum value, etc. of the acquired signal are segmented and the frequency domain characteristics after Fourier transformation are used as the initial characteristics of the signal;

[0108] The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector.

[0109] Most of the abnormal voltage transformer measurement is caused by equipment aging. Therefore, enhancing the cumulative characteristics of the transformer secondary circuit signal is crucial to distinguish the abnormal state of the voltage transformer. It is necessary to cumulatively change the traditional time domain characteristics and cumulatively transform the characteristics of the same type at different times. The cumulative transformation formula is as follows:

[0110] ;

[0111] ;

[0112] Where k represents the number of signal segments, N represents the total signal length of the voltage signal, and b represents the length of each segment of the signal after segmentation. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation.

[0113] According to the above formula, cumulative transformation actually accumulates features in the time domain. The transformed result is related to the original time domain features and the length of the signal. It is a time fusion method. Therefore, it is necessary to ensure that the sample lengths in the data set are consistent to avoid large differences in feature values ​​caused by sample length, which is not conducive to model training.

[0114] In order to avoid the influence of sample length on cumulative features, this embodiment proposes the following method for processing:

[0115] ;

[0116] ;

[0117] in, is the correction factor, N 0 indicates the preset standard signal length.

[0118] In order to strengthen the data accumulation feature and reduce the impact of data length on the feature threshold, the initial accumulation transformation result after the accumulation transformation is accumulated again to obtain the final accumulation transformation result:

[0119] ;

[0120] in, It is the final cumulative transformation result of the j-th feature when the signal length is divided into k segments.

[0121] As a preferred implementation of this embodiment, the method of fusing the cumulative transformation processing results with the frequency domain features to obtain the feature vector is specifically as follows:

[0122] Get the cumulative transformation processing results and frequency domain features as the input set:

[0123] ;

[0124] in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform;

[0125] After the input set is normalized, the normalized results are sequentially subjected to convolution and pooling operations to obtain a fused feature set.

[0126] In order to enhance the adaptability of the features, the fused features need to be similar under the same measurement state of the mutual inductor. This embodiment determines the weight of each fused feature by comparing the similarity of each fused feature in the fused feature set under the same measurement state at different times:

[0127] ;

[0128] in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features.

[0129] Feature screening is performed based on the weights of each fusion feature:

[0130]

[0131] in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold;

[0132] Based on the above feature screening formula, the features with correlation coefficients lower than the threshold are selected with the coefficients set to zero to achieve the screening of high-quality features.

[0133] The signal of the mutual inductor for a certain period of time is extracted to construct the feature set. By screening high-quality features and fusing the features through the fully connected layer, the final feature vector is obtained:

[0134] ;

[0135] in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

[0136] Considering that when the samples are unbalanced, the hyperplane of SVM will be biased towards the area with a small number of samples, resulting in more errors in identifying this type of samples. This embodiment adjusts the SVM penalty coefficient to suppress the deviation of the hyperplane and optimize the recognition ability of the model when the samples are unbalanced.

[0137] As a preferred implementation of this embodiment, the penalty coefficient of the support vector machine is adjusted according to the distance of each sample to the support vector machine hyperplane and the recognition accuracy, and the steps of performing model iteration are specifically as follows:

[0138] The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept;

[0139] Get the geometric distance from the sample to the hyperplane:

[0140] ;

[0141] ;

[0142] in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples;

[0143] This embodiment proposes a method of automatically setting the penalty coefficient of SVM according to the distance level of positive and negative samples to the hyperplane, so that the mean distances of the two types of samples to the hyperplane remain roughly consistent, suppressing the deviation of the hyperplane. First, the distance level of positive and negative samples to the hyperplane needs to be divided, and the expression formula is:

[0144] ;

[0145] in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively;

[0146] Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, the classification results of positive and negative samples and the distance to the hyperplane are analyzed, and the attention of the training model to the sample is adaptively adjusted. The adjustment formula of the penalty coefficient is:

[0147] ;

[0148] in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample. According to the above formula, the higher the distance level, the greater the penalty coefficient, which makes the model pay more attention to the training of this type of samples, thereby adjusting the offset of the hyperplane.

[0149] Finally, the data collected from the secondary side of the voltage transformer are divided into a training set and a test set, and the improved support vector machine model is trained using the training set.

[0150] First, the penalty coefficient of each sample is initialized and set, and the recognition performance of the support vector machine model of the training set in the test set is calculated, such as recognition accuracy, recall rate and other indicators for judgment.

[0151] By analyzing the results of the test set, if the recognition accuracy is lower than a certain threshold, the penalty coefficient of the sample is adjusted and increased, and then the model recognition accuracy is increased to above the threshold;

[0152] During the training process, the accuracy and recall of model recognition are continuously analyzed. If the variation range of these two indicators is less than 2%, it indicates that the model has converged and the iteration is stopped. Then the trained parameters are input into the model, the characteristic vector of each sample is identified, and the measurement status of the voltage transformer is recognized.

[0153] Embodiment 2:

[0154] This embodiment provides a voltage transformer measurement state identification system, including:

[0155] A preprocessing module, used to collect the secondary side voltage signal of the voltage transformer sample and preprocess it to obtain the sample feature vector; this module is used to implement the function of step S100 in the first embodiment, which will not be described in detail here;

[0156] A model training module is used to take voltage transformer samples in normal and abnormal measurement states as positive and negative samples, respectively, input the feature vectors of the positive and negative samples into the support vector machine model, and adjust the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, perform model iteration, and obtain an optimized support vector machine; this module is used to implement the function of step S200 in embodiment 1, which will not be repeated here;

[0157] A state recognition module is used to collect the secondary side voltage of the target voltage transformer to be identified and perform preprocessing to obtain the feature vector of the target voltage transformer, and input the feature vector of the target voltage transformer into the optimized support vector machine to obtain the measurement state recognition result; this module is used to implement the function of step S300 in the first embodiment, and will not be repeated here;

[0158] Among them, the method for preprocessing in the preprocessing module is specifically as follows:

[0159] Process the secondary side voltage signal to obtain the time domain characteristics and frequency domain characteristics after segmentation;

[0160] The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector.

[0161] As a preferred implementation of this embodiment, the method in which the preprocessing module performs cumulative transformation processing on the segmented time domain features is specifically as follows:

[0162] The features of the same type at different times in the segmented time domain features are cumulatively transformed, specifically:

[0163] ;

[0164] ;

[0165] ;

[0166] Where k is the number of signal segments, N is the total length of the signal, and b is the length of the segment. is the correction factor, N 0 means the set standard signal length. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation;

[0167] The initial cumulative transformation result after the cumulative transformation is cumulatively transformed again to obtain the final cumulative transformation result:

[0168] ;

[0169] in, It is the final cumulative transformation result of the j-th feature when the signal length is divided into k segments.

[0170] As a preferred implementation of this embodiment, the method of fusing the cumulative transformation processing results with the frequency domain features to obtain the feature vector is specifically as follows:

[0171] Get the cumulative transformation processing results and frequency domain features as the input set:

[0172] ;

[0173] in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform;

[0174] After normalizing the input set, the normalized results are sequentially convolved and pooled to obtain a fused feature set;

[0175] Compare the similarities of each fusion feature in the fusion feature set at the same measurement state at different times to determine the weight of each fusion feature:

[0176] ;

[0177] in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features.

[0178] Feature screening is performed based on the weights of each fusion feature:

[0179]

[0180] in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold;

[0181] Perform feature fusion to obtain the feature vector:

[0182] ;

[0183] in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

[0184] As a preferred implementation of this embodiment, the model training module adjusts the penalty coefficient of the support vector machine according to the distance of each sample to the support vector machine hyperplane and the recognition accuracy, and the steps of performing model iteration are specifically as follows:

[0185] The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept;

[0186] Get the geometric distance from the sample to the hyperplane:

[0187] ;

[0188] ;

[0189] in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples;

[0190] Divide the distance levels of positive and negative samples to the hyperplane;

[0191] ;

[0192] in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively;

[0193] Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, adjust the penalty coefficient:

[0194] ;

[0195] in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample.

[0196] Embodiment three:

[0197] This embodiment provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for identifying the measurement state of a voltage transformer as described in any embodiment of the present invention is implemented.

[0198] Embodiment 4:

[0199] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for identifying the measurement state of a voltage transformer as described in any embodiment of the present invention is implemented.

[0200] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0201] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0202] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0203] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0204] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A voltage transformer measurement state identification method, It is characterized in that The following steps are involved: Collecting the secondary side voltage signal of the voltage transformer sample and preprocessing it to obtain the sample feature vector; The voltage transformer samples in normal and abnormal measurement states are taken as positive and negative samples respectively, and the feature vectors of the positive and negative samples are input into the support vector machine model. The penalty coefficient of the support vector machine is adjusted according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, and the model is iterated to obtain the optimized support vector machine. The secondary side voltage of the target voltage transformer to be identified is collected and preprocessed to obtain a feature vector of the target voltage transformer, and the feature vector of the target voltage transformer is input into the optimized support vector machine to obtain a measurement state identification result; The specific preprocessing method is as follows: Process the secondary side voltage signal to obtain the time domain characteristics and frequency domain characteristics after segmentation; The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector; The method for performing cumulative transformation processing on the segmented time domain features is specifically as follows: The features of the same type at different times in the segmented time domain features are cumulatively transformed, specifically: ； ； ； Where k is the number of signal segments, N is the total length of the signal, and b is the length of the segment. is the correction factor, N 0 Indicates the set standard signal length. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation; The initial cumulative transformation result after the cumulative transformation is cumulatively transformed again to obtain the final cumulative transformation result: ； in, is the final cumulative transformation result of the jth feature when the signal length is divided into k segments; The method of fusing the cumulative transformation processing result with the frequency domain feature to obtain the feature vector is specifically as follows: Get the cumulative transformation processing results and frequency domain features as the input set: ； in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform; After normalizing the input set, the normalized results are sequentially convolved and pooled to obtain a fused feature set; Compare the similarities of each fusion feature in the fusion feature set at the same measurement state at different times to determine the weight of each fusion feature: ； in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features. Feature screening is performed based on the weights of each fusion feature: in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold; Perform feature fusion to obtain the feature vector: ； in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

2. A voltage transformer measurement state identification method according to claim 1, It is characterized in that The steps of adjusting the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy and performing model iteration are specifically as follows: The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept; Get the geometric distance from the sample to the hyperplane: ； ； in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples; Divide the distance levels of positive and negative samples to the hyperplane; ； in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively; Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, adjust the penalty coefficient: ； in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample.

3. A voltage transformer measurement state identification system, It is characterized in that include: A preprocessing module, used for collecting the secondary side voltage signal of the voltage transformer sample and performing preprocessing to obtain the sample feature vector; The model training module is used to take voltage transformer samples in normal and abnormal measurement states as positive and negative samples respectively, input the feature vectors of the positive and negative samples into the support vector machine model, and adjust the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, perform model iteration, and obtain the optimized support vector machine; A state recognition module is used to collect the secondary side voltage of the target voltage transformer to be identified and perform preprocessing to obtain the feature vector of the target voltage transformer, input the feature vector of the target voltage transformer into the optimized support vector machine, and obtain the measurement state recognition result; Among them, the method for preprocessing in the preprocessing module is specifically as follows: Process the secondary side voltage signal to obtain the time domain characteristics and frequency domain characteristics after segmentation; The segmented time domain features are cumulatively transformed, and the cumulative transformation results are fused with the frequency domain features to obtain a feature vector; The method for the preprocessing module to perform cumulative transformation processing on the segmented time domain features is specifically as follows: The features of the same type at different times in the segmented time domain features are cumulatively transformed, specifically: ； ； ； Where k is the number of signal segments, N is the total length of the signal, and b is the length of the segment. is the correction factor, N 0 Indicates the set standard signal length. It represents the initial cumulative transformation result of the j-th feature when the signal length is divided into k segments; Represents the i-th segment signal after segmentation; The initial cumulative transformation result after the cumulative transformation is cumulatively transformed again to obtain the final cumulative transformation result: ； in, is the final cumulative transformation result of the jth feature when the signal length is divided into k segments; The method of fusing the cumulative transformation processing result with the frequency domain feature to obtain the feature vector is specifically as follows: Get the cumulative transformation processing results and frequency domain features as the input set: ； in, It represents the final cumulative transformation result of the g-th feature when the signal length is divided into k segments. represents the mth frequency domain feature after Fourier transform; After normalizing the input set, the normalized results are sequentially convolved and pooled to obtain a fused feature set; Compare the similarities of each fusion feature in the fusion feature set at the same measurement state at different times to determine the weight of each fusion feature: ； in, represents the weight coefficient of the jth fusion feature, It represents the standard deviation of the eigenvalue of the jth fused feature under the same measurement state at different times. max(rms) represents the maximum standard deviation in the set of eigenvalue standard deviations of all fused features. min(rms) represents the minimum standard deviation in the set of eigenvalue standard deviations of all fused features. Feature screening is performed based on the weights of each fusion feature: in, and They represent the final cumulative transformation results of the j-th feature when the signal length is divided into x and y segments, respectively. represents the correlation coefficient of the jth feature, represents the selection coefficient of the jth feature, is the preset correlation coefficient threshold; Perform feature fusion to obtain the feature vector: ； in, represents the feature vector of the i-th sample, Represents the eigenvalue of the jth feature in the i-th sample.

4. A voltage transformer measurement state identification system according to claim 3, It is characterized in that The model training module adjusts the penalty coefficient of the support vector machine according to the distance from each sample to the support vector machine hyperplane and the recognition accuracy, and the steps of model iteration are specifically as follows: The support vector machine obtains the hyperplane H(x) through each sample, which is specifically expressed as wx+z=0, where w represents the weight vector, x represents the feature vector of the sample, and z represents the intercept; Get the geometric distance from the sample to the hyperplane: ； ； in, represents the distance from the i-th sample to the hyperplane, Represents the feature vector of the i-th sample; Represents the mean distance from the sample to the hyperplane, and M represents the number of input samples; Divide the distance levels of positive and negative samples to the hyperplane; ； in, represents the distance level from sample i to the hyperplane, , Represent the i-th positive and negative samples respectively, represents the ceiling operator; , are the maximum and minimum distances from the positive sample to the hyperplane, respectively. , are the maximum and minimum distances from negative samples to the hyperplane, respectively; Based on the distance classification results of each sample to the hyperplane and the recognition accuracy of each sample, adjust the penalty coefficient: ； in, represents the adjusted penalty coefficient, represents the initial penalty coefficient, is the recognition accuracy of the i-th sample.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the voltage transformer measurement state identification method according to any one of claims 1 to 2 is implemented.

6. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the voltage transformer measurement state identification method as described in any one of claims 1 to 2 is implemented.

Images