Error state evaluation method and device of current transformer

Abstract

The invention provides an error state evaluation method and device for a current transformer, and belongs to the technical field of electric power metering online monitoring, and the method comprises the steps: obtaining the secondary current data of the current transformer, and determining the index parameters of the current transformer based on the secondary current data; the index parameters comprise a zero-sequence imbalance parameter, a negative-sequence imbalance parameter and a fluctuation imbalance parameter; constructing sample groups in different states; based on the index parameter, calculating the sample group to obtain a time sequence data set; performing feature enhancement on the time sequence data set to obtain an enhanced data set; and performing feature extraction on the enhanced data set to obtain a feature vector, and inputting the feature vector into a naive Bayes classifier to obtain an error state evaluation result. According to the error state evaluation method of the current transformer, the zero-sequence imbalance parameter, the negative-sequence imbalance parameter and the fluctuation imbalance parameter are introduced, and the time sequence data set is subjected to feature enhancement, so that the problem of inaccurate error state evaluation of the error state evaluation method of the current transformer is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of online monitoring of electric power metering, and in particular to a method and a device for evaluating the error state of a current transformer, which are specifically used for error evaluation of a current transformer. Background Art

[0002] With the continuous development of power systems and the improvement of their intelligence level, higher requirements are placed on the measurement accuracy and error control of current transformers. As a key measuring device in power systems, the error state of current transformers directly affects the operational safety, accuracy of electric energy metering and economic benefits of power systems. Therefore, accurate evaluation of the error state of current transformers is an important basis for ensuring the stable operation of power systems.

[0003] The error sources of current transformers are complex and diverse, including manufacturing process, material properties, working environment, electromagnetic interference and other factors. The interaction of these factors makes it difficult to accurately predict and control the error state of current transformers. Therefore, conducting research on the error state evaluation of current transformers will help to gain a deeper understanding of the mechanism of error generation and provide a scientific basis for error control. Summary of the invention

[0004] In view of this, it is necessary to provide a method and device for evaluating the error state of a current transformer, so as to solve the technical problem of inaccurate error state evaluation in the existing error state evaluation method of the current transformer.

[0005] In order to solve the above problems, on the one hand, the present invention provides an error state evaluation method of a current transformer, comprising:

[0006] Acquire secondary current data of the current transformer, and determine index parameters of the current transformer based on the secondary current data; the index parameters include: zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter;

[0007] Determine an additional error introduced by the current transformer, and construct sample groups of different states based on the additional error;

[0008] Based on the indicator parameters, the sample group is calculated to obtain a time series data set;

[0009] Performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset;

[0010] Feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0011] In a possible implementation, determining an index parameter of a current transformer based on the secondary current data includes:

[0012] Determining a zero-sequence unbalance parameter and a negative-sequence unbalance parameter based on three different A-phase zero-sequence components in the secondary current data;

[0013] Based on the fluctuation component caused by the three-phase load in the secondary current data, a fluctuation unbalance parameter is determined.

[0014] In a possible implementation, determining the fluctuation unbalance parameter based on the fluctuation component caused by the three-phase load in the secondary current data includes:

[0015] Determine the fluctuation component caused by the A-phase load based on the A-phase current difference at adjacent moments in the secondary current data;

[0016] Determine the fluctuation component caused by the B-phase load based on the B-phase current difference at adjacent moments in the secondary current data;

[0017] Determine the fluctuation component caused by the C-phase load based on the C-phase current difference at adjacent moments in the secondary current data;

[0018] The fluctuation component caused by the A-phase load, the fluctuation component caused by the B-phase load and the fluctuation component caused by the C-phase load are added together to obtain the fluctuation unbalance parameter.

[0019] In a possible implementation, constructing sample groups in different states based on the additional error includes:

[0020] Different additional errors are added with 1 and then multiplied with the secondary current data to obtain sample groups in different states.

[0021] In a possible implementation, feature enhancement is performed on the time series dataset based on an attention mechanism to obtain an enhanced dataset, including:

[0022] Based on the attention mechanism, the time series data set is spatially fused to obtain a spatial fused data set;

[0023] Channel fusion is performed on the spatial fusion data set to obtain an enhanced data set.

[0024] In a possible implementation, spatial fusion is performed on the time series dataset based on an attention mechanism to obtain a spatial fusion dataset, including:

[0025] Normalizing the time series data set to obtain a first normalized data set;

[0026] Inputting the first normalized data set into a multi-scale convolutional model to obtain a first multi-scale feature;

[0027] Determine the attention value of the first multi-scale feature at three linear projection layers;

[0028] After weighted calculation of two attention values ​​of the first multi-scale feature in the three linear projection layers, data fusion is performed on the weighted calculation result and the other attention value, and the fusion result is added to the time series data in the time series data set to obtain a spatial fusion data set.

[0029] In a possible implementation, performing channel fusion on the spatial fusion data set to obtain an enhanced data set includes:

[0030] Normalizing the data of different channels in the spatial fusion data set to obtain a second normalized data set;

[0031] Inputting the second normalized data set into a multi-scale convolutional model to obtain a second multi-scale feature;

[0032] Determine the attention value of the second multi-scale feature at two linear projection layers;

[0033] Based on the attention values ​​of the second multi-scale features in the two linear projection layers, data fusion is performed on different channel data in the spatial fusion data set, and the fusion result is added to the channel data to obtain an enhanced data set.

[0034] In a possible implementation, feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer, including:

[0035] Normalizing the enhanced data set to obtain a third normalized data set;

[0036] Performing a dropout operation on the third normalized data set to obtain a down-fitting data set;

[0037] Inputting the third normalized data set and the down-fit data set into a preset Transformer model for feature extraction to obtain a first feature vector and a second feature vector;

[0038] Fusing the first feature vector and the second feature vector to obtain a fused feature vector;

[0039] Inputting the fused feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer;

[0040] The Transformer model only includes a decoder structure.

[0041] In a possible implementation, the error state evaluation method of the current transformer further includes:

[0042] When it is determined that an abnormal current transformer exists among the multiple current transformers based on the error state evaluation results of the multiple current transformers, a three-phase contribution index of each current transformer is determined based on the three-phase current amplitudes of the same time sequence in the secondary current data of each current transformer;

[0043] It is determined that the current transformer with the largest three-phase contribution index among multiple current transformers is the out-of-tolerance current transformer.

[0044] On the other hand, the present invention further provides an error state evaluation device for a current transformer, comprising:

[0045] An acquisition module, used for acquiring secondary current data of a current transformer, and determining index parameters of the current transformer based on the secondary current data; the index parameters include: a zero-sequence unbalance parameter, a negative-sequence unbalance parameter, and a fluctuation unbalance parameter;

[0046] A sample construction module, used to determine an additional error introduced by the current transformer, and to construct sample groups of different states based on the additional error;

[0047] A sample calculation module, used to calculate the sample group based on the indicator parameter to obtain a time series data set;

[0048] A feature enhancement module, used for performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset;

[0049] The error evaluation module is used to extract features from the enhanced data set to obtain a feature vector, and input the feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0050] The beneficial effects of adopting the above implementation method are:

[0051] The error state evaluation method and device of the current transformer provided by the present invention obtains the secondary current data of the current transformer, and determines the zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter of the current transformer based on the secondary current data; under normal circumstances, the negative-sequence unbalance, zero-sequence unbalance and current fluctuation unbalance parameters of the secondary current of the current transformer are relatively stable. When the current transformer metering error changes, the data distribution of its negative-sequence, zero-sequence unbalance and fluctuation unbalance parameters will change, and no longer meet the distribution characteristics under normal operating conditions. The present invention introduces zero-sequence unbalance parameters, negative-sequence unbalance parameters and fluctuation unbalance parameters, which can more comprehensively evaluate the error state of the current transformer and improve the accuracy of error state evaluation.

[0052] In addition, the present invention introduces additional errors to construct sample groups in different states, which may include normal states, alarm states or abnormal states, so that the obtained time series data set is closer to the actual situation; based on the attention mechanism, the feature enhancement of the time series data set can improve the data features; the feature extraction of the enhanced data set is performed to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain the error state evaluation result of the current transformer.

[0053] In summary, the present invention improves the accuracy of error state assessment by introducing zero-sequence unbalance parameters, negative-sequence unbalance parameters and fluctuating unbalance parameters, and enhances the features of the time series data set based on the attention mechanism, which can improve data features, thereby solving the technical problem of inaccurate error state assessment in existing error state assessment methods of current transformers. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0055] Figure 1 A flow chart of an embodiment of the error state evaluation method of a current transformer provided by the present invention;

[0056] Figure 2 A flow chart of another embodiment of the error state evaluation method of a current transformer provided by the present invention;

[0057] Figure 3 A flow chart of spatial fusion provided by the present invention;

[0058] Figure 4 A flow chart of channel fusion provided by the present invention;

[0059] Figure 5 A schematic diagram of the feature extraction model provided by the present invention;

[0060] Figure 6 A principle block diagram of an embodiment of the error state evaluation device for a current transformer provided by the present invention

[0061] Figure 7 A schematic structural diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0063] In the description of the embodiments of the present application, unless otherwise specified, “plurality” means two or more.

[0064] The terms "including" and "having" and any variations thereof in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or equipment comprising a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules not clearly listed or inherent to these processes, methods, products or equipment.

[0065] The naming or numbering of the steps in the embodiments of the present invention does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The execution order of the named or numbered process steps can be changed according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved.

[0066] Reference to an "embodiment" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiment may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0067] The present invention provides a method and device for evaluating the error state of a current transformer, which are respectively described below.

[0068] Embodiment 1:

[0069] like Figure 1 As shown, the present invention provides a method for evaluating the error state of a current transformer, comprising:

[0070] S101, obtaining secondary current data of a current transformer, and determining index parameters of the current transformer based on the secondary current data; the index parameters include: a zero-sequence unbalance parameter, a negative-sequence unbalance parameter, and a fluctuation unbalance parameter;

[0071] S102, determining an additional error introduced by the current transformer, and constructing sample groups in different states based on the additional error;

[0072] S103, calculating the sample group based on the indicator parameter to obtain a time series data set;

[0073] S104, performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset;

[0074] S105 , extracting features from the enhanced data set to obtain a feature vector, and inputting the feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0075] It can be understood that the secondary current data collected from the current transformer can be used to calculate index parameters such as zero-sequence unbalance, negative-sequence unbalance, and fluctuation unbalance; the introduction of fluctuation unbalance parameters improves the accuracy of the model;

[0076] An additional error is introduced into a mutual sensor in the three-phase system to construct a training set, which includes normal, alarm and abnormal data.

[0077] The attention mechanism is used to enhance the features of the training data set. First, the indicator data are spatially fused separately, and then the spatially fused indicator data are channel-fused. The data is enhanced to improve the data features.

[0078] The data after feature enhancement is sent to the improved Transformer model for feature extraction, and the naive Bayes algorithm is used to construct an error assessment model for the extracted features;

[0079] Further calculations are performed on the current transformer group that is evaluated to be abnormal to locate the abnormal transformer.

[0080] The error state evaluation method of the current transformer provided by the present invention obtains the secondary current data of the current transformer, and determines the zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter of the current transformer based on the secondary current data; under normal circumstances, the negative-sequence unbalance, zero-sequence unbalance and current fluctuation unbalance parameters of the secondary current of the current transformer are relatively stable. When the current transformer metering error changes, the data distribution of its negative-sequence, zero-sequence unbalance and fluctuation unbalance parameters will change, and no longer meet the distribution characteristics under normal operating conditions. The present invention introduces zero-sequence unbalance parameters, negative-sequence unbalance parameters and fluctuation unbalance parameters, and can more comprehensively evaluate the error state of the current transformer, thereby improving the accuracy of error state evaluation.

[0081] In addition, the present invention introduces additional errors to construct sample groups in different states, which may include normal states, alarm states or abnormal states, so that the obtained time series data set is closer to the actual situation; based on the attention mechanism, the feature enhancement of the time series data set can improve the data features; the feature extraction of the enhanced data set is performed to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain the error state evaluation result of the current transformer.

[0082] In summary, the present invention improves the accuracy of error state assessment by introducing zero-sequence unbalance parameters, negative-sequence unbalance parameters and fluctuating unbalance parameters, and enhances the features of the time series data set based on the attention mechanism, which can improve data features, thereby solving the technical problem of inaccurate error state assessment in existing error state assessment methods of current transformers.

[0083] In some embodiments, determining an index parameter of a current transformer based on the secondary current data includes:

[0084] Determining a zero-sequence unbalance parameter and a negative-sequence unbalance parameter based on three different A-phase zero-sequence components in the secondary current data;

[0085] Based on the fluctuation component caused by the three-phase load in the secondary current data, a fluctuation unbalance parameter is determined.

[0086] It is understandable that the secondary current data of the three-phase current transformer that has just been put into operation or just after maintenance is collected to calculate the zero-sequence unbalance and negative-sequence unbalance:

[0087]

[0088] Zero sequence unbalance:

[0089] Negative sequence imbalance:

[0090] in, is the three-phase current, That is the corresponding a-phase zero-sequence component.

[0091] In some embodiments, determining the fluctuation unbalance parameter based on the fluctuation component caused by the three-phase load in the secondary current data includes:

[0092] Determine the fluctuation component caused by the A-phase load based on the A-phase current difference at adjacent moments in the secondary current data;

[0093] Determine the fluctuation component caused by the B-phase load based on the B-phase current difference at adjacent moments in the secondary current data;

[0094] Determine the fluctuation component caused by the C-phase load based on the C-phase current difference at adjacent moments in the secondary current data;

[0095] The fluctuation component caused by the A-phase load, the fluctuation component caused by the B-phase load and the fluctuation component caused by the C-phase load are added together to obtain the fluctuation unbalance parameter.

[0096] It can be understood that under normal operation, the secondary current fluctuation component of the three-phase current transformer is relatively balanced, and the fluctuation parameter m3 is defined as:

[0097]

[0098] in, They are the fluctuation components caused by the loads of phase A, phase B, and phase C just after commissioning or maintenance, respectively. Represents the difference between the current at the current moment and the current at the previous moment, t=1, 2, ..., T.

[0099] In some embodiments, constructing sample groups of different states based on the additional error includes:

[0100] Different additional errors are added with 1 and then multiplied with the secondary current data to obtain sample groups in different states.

[0101] It can be understood that for the current transformer, different additional errors are introduced according to formula (5) to simulate the current transformer in different states (normal, alarm, abnormal):

[0102] I i ′=(1+Δf i )I i (5)

[0103] Among them, I i is the secondary current amplitude of the ith transformer, Δf i is the additional ratio error introduced by the i-th transformer, I i ′ is the current amplitude after the additional ratio error is introduced.

[0104] In some embodiments, feature enhancement is performed on the time series dataset based on an attention mechanism to obtain an enhanced dataset, including:

[0105] Based on the attention mechanism, the time series data set is spatially fused to obtain a spatial fused data set;

[0106] Channel fusion is performed on the spatial fusion data set to obtain an enhanced data set.

[0107] In some embodiments, spatially fusing the time series dataset based on an attention mechanism to obtain a spatially fused dataset includes:

[0108] Normalizing the time series data set to obtain a first normalized data set;

[0109] Inputting the first normalized data set into a multi-scale convolutional model to obtain a first multi-scale feature;

[0110] Determine the attention value of the first multi-scale feature at three linear projection layers;

[0111] After weighted calculation of two attention values ​​of the first multi-scale feature in the three linear projection layers, data fusion is performed on the weighted calculation result and the other attention value, and the fusion result is added to the time series data in the time series data set to obtain a spatial fusion data set.

[0112] In some embodiments, performing channel fusion on the spatial fusion data set to obtain an enhanced data set includes:

[0113] Normalizing the data of different channels in the spatial fusion data set to obtain a second normalized data set;

[0114] Inputting the second normalized data set into a multi-scale convolutional model to obtain a second multi-scale feature;

[0115] Determine the attention value of the second multi-scale feature at two linear projection layers;

[0116] Based on the attention values ​​of the second multi-scale features in the two linear projection layers, data fusion is performed on different channel data in the spatial fusion data set, and the fusion result is added to the channel data to obtain an enhanced data set.

[0117] In some embodiments, feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer, including:

[0118] Normalizing the enhanced data set to obtain a third normalized data set;

[0119] Performing a dropout operation on the third normalized data set to obtain a down-fitting data set;

[0120] Inputting the third normalized data set and the down-fit data set into a preset Transformer model for feature extraction to obtain a first feature vector and a second feature vector;

[0121] Fusing the first feature vector and the second feature vector to obtain a fused feature vector;

[0122] Inputting the fused feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer;

[0123] The Transformer model only includes a decoder structure.

[0124] It can be understood that the feature data was dropout-operated and input into the improved Transformer model respectively, which improved the model evaluation speed and prevented the model from overfitting.

[0125] In some embodiments, the error state evaluation method of the current transformer further includes:

[0126] When it is determined that an abnormal current transformer exists among the multiple current transformers based on the error state evaluation results of the multiple current transformers, a three-phase contribution index of each current transformer is determined based on the three-phase current amplitudes of the same time sequence in the secondary current data of each current transformer;

[0127] It is determined that the current transformer with the largest three-phase contribution index among multiple current transformers is the out-of-tolerance current transformer.

[0128] In some embodiments, Figure 2 As shown, the error state evaluation method of the current transformer provided by the present invention includes:

[0129] S201: Collect secondary current data of current transformer and calculate index parameters such as zero sequence unbalance, negative sequence unbalance, and fluctuation unbalance;

[0130] Under normal circumstances, the negative-sequence imbalance and zero-sequence imbalance of the secondary current of the current transformer are relatively stable. When the transformer measurement error changes, the data distribution of its negative-sequence and zero-sequence imbalance will change and no longer meet the distribution characteristics under normal operating conditions.

[0131] (1) Collect the secondary current data of the three-phase current transformer that has just been put into operation or just been overhauled, and calculate the zero-sequence unbalance and negative-sequence unbalance:

[0132]

[0133] Zero sequence unbalance:

[0134] Negative sequence imbalance:

[0135] in, is the three-phase current, That is the corresponding a-phase zero-sequence component.

[0136] (2) Under normal operating conditions, the secondary current fluctuation component of the three-phase current transformer is relatively balanced, and the fluctuation parameter m3 is defined as:

[0137]

[0138] in, They are the fluctuation components caused by the loads of phase A, phase B, and phase C just after commissioning or maintenance, respectively. Represents the difference between the current at the current moment and the current at the previous moment, t=1, 2, ..., T.

[0139] S202: introducing an additional error to a mutual sensor in the three-phase circuit, and constructing a training set, wherein the training set includes normal data, warning data, and abnormal data;

[0140] For the current transformer, different additional errors are introduced according to formula (5) to simulate the current transformer in different states (normal, alarm, abnormal):

[0141] I i ′=(1+Δf i )I i (5)

[0142] Among them, I i is the secondary current amplitude of the ith transformer, Δf i is the additional ratio error introduced by the i-th transformer, I i ′ is the current amplitude after the additional error is introduced.

[0143] Through the above steps, sample groups of different states are constructed. Based on formulas (1) to (4), the sample groups are calculated to obtain m1, m2, and m3, and the time series data set is constructed:

[0144] M1={m 11 ,m 12 ,...,m 1t} (6)

[0145] M2={m 21 ,m 22 ,...,m 2t} (7)

[0146] M3={m 31 ,m 32 ,...,m 3t} (8)

[0147] m 3t It represents the fluctuation unbalance parameter of the tth point after a group of mutual inductors introduces errors, t=1,2,...T, T is the sequence length.

[0148] S203: Using the attention mechanism to enhance the features of the training data set, first spatially fuse the indicator data separately, and then perform channel fusion on the spatially fused indicator data.

[0149] Data feature enhancement processing, including:

[0150] Spatial fusion process reference Figure 3 , specifically:

[0151] Input M1, M2, and M3 of a group of mutual inductors into the data feature space fusion model respectively, taking M1 as an example:

[0152] ① Normalize M1 and get Input the normalized data into the multi-convolution model:

[0153]

[0154] Among them, α1, α2, α3, and α4 are the weight parameters of each convolution respectively.

[0155] ②Calculate the attention of data:

[0156]

[0157] in, Represent three linear projection layers respectively.

[0158] ③Spatial data fusion:

[0159] Weighted calculation:

[0160] Data Fusion:

[0161] Among them, W q =[W q1 ,W q2 ,...,W qT ], t=1,2,...,T, T represents the time series length, k i 、v i K S 、V S The i-th data value. -(|ti|-1) / T is the relative position deviation between the t-th data and the i-th data, and its size can be controlled by the learning parameter w. u is the reward parameter of the t-th data, which is a learnable parameter. ⊙ represents the dot product, σ(·) is the Sigmoid gating function, Indicates addition, W OS A linear projection layer for output projection.

[0162] Channel Fusion Reference Figure 4 , specifically:

[0163] SP is constructed based on the different channel data after spatial fusion, and SP is input into the channel fusion model for fusion.

[0164]

[0165] ① Normalize the data to obtain Input the normalized data into the multi-scale convolutional network:

[0166]

[0167] Among them, β1, β2, and β3 are the weight parameters of each convolution.

[0168] ②Calculate the attention of data:

[0169]

[0170] in, Represent three linear projection layers respectively.

[0171] ④Channel data fusion:

[0172] Data Fusion:

[0173] Among them, SC∈R 3*T , σ(·) is the Sigmoid gating function, Indicates addition, A linear projection layer for output projection.

[0174] S204: Reference Figure 5 , the feature-enhanced data is sent to the improved Transformer model for feature extraction, and the naive Bayes algorithm is used to construct an error evaluation model for the extracted features;

[0175] The data SC after spatial fusion and channel fusion is normalized to obtain E1. In order to avoid overfitting of the model, the dropout operation is introduced to E1 to obtain E2. E1 and E2 are respectively input into the improved Transformer model for feature extraction.

[0176] E2=DO(E1) (13)

[0177] Among them, DO stands for dropout operation.

[0178] ① To improve the model evaluation speed, the Transformer here adopts a decoder-only structure, which contains L layers of decoders.

[0179] The data E1 is divided into N blocks according to the length S, and each block is represented as:

[0180] fs m ={x (m-1)S+1 ,...,x mS} (14)

[0181] Where, m = 1, 2, ..., N, fs m ∈R 3*S .

[0182]

[0183] H l =Transblock(H l-1 ),l=1,2,...,L (16)

[0184]

[0185] Indicates the embedding vector obtained through the embedding layer, TE m Indicates the added position code. e , W d Respectively represent the weight coefficients of encoding and decoding. l-1 represents the output of the l-1 layer decoder, Transblock (H l-1 ) indicates H l-1 The feature data of the layer is decoded to obtain the output of the decoder of the lth layer. Transblock(·) represents the decoder. represents the feature vector obtained after the mth data block passes through L layers of transformer decoding layers, N fs m After the data block is processed, the feature vector is obtained

[0186] ② Perform feature extraction on data E2 according to step ① above, and obtain N is the number of blocks to be divided.

[0187] ③Characteristics and Fusion is performed to obtain the fused features:

[0188] (2) Input the fused feature CS into the Naive Bayesian Classifier (NBC) to complete the status evaluation of the transformer group;

[0189] Status judgment:

[0190] Y λ =NBC(CS) (25)

[0191] Among them, λ=0,1,2, Y0 is normal, Y1 is warning, and Y2 is abnormal.

[0192] Define the loss function:

[0193] in, is the real status label. When one of the three phases is abnormal, is abnormal; when one phase alarms, It is an alarm; when the three phases are normal, is normal, a=1,2,...,A, A is the number of sample groups. τ is the weight coefficient, ρ a is the maximum posterior probability of sample a obtained through the naive Bayes model; D β is the distance between samples in normal, warning or abnormal state, D β′ It is the distance between samples that are classified as normal, warning or abnormal.

[0194] S205: further calculating the transformer group evaluated to be abnormal, and locating the abnormal transformer;

[0195] Obtain the secondary current data of the abnormal three-phase current transformer and calculate the contribution index of the three phases:

[0196]

[0197] in, is any one of the three phases, that is: j represents the time series, j=1,2,...T. is the mean current amplitude of the three-phase current in the same time sequence. The current transformer with the largest contribution index Q among the three phases is the current transformer with the largest contribution index Q among the three phases in the same group.

[0198] For the 0.2-level current transformer, the actual detection shows that the error is 0.03%. By adding additional errors to simulate the transformers in different states, the additional error varies in the range of [-0.4%, 0.4%]. The state with an error of (-0.2%, 0.2%) is defined as normal; the state with errors of (-0.3%, -0.2%] and [0.2%, 0.3%) is defined as an alarm; the state with errors of (-∞, -0.3%] and [0.3%, ∞) is defined as abnormal.

[0199] The technology of the present invention is used to perform online evaluation of the error state of the current transformer, and some evaluation results are compared with the actual simulation results as shown in Table 1 and Table 2:

[0200] Table 1: Current transformer error status online evaluation table

[0201] sample Additional error Actual error Actual status Evaluation of the invention 1 -0.4％ -0.37％ abnormal abnormal 2 -0.39％ -0.36％ abnormal abnormal 3 -0.38％ -0.35％ abnormal abnormal 4 -0.37％ -0.35％ abnormal abnormal ... ... ... ... 80 0.39 0.36％ abnormal abnormal 81 0.4 0.37％ abnormal abnormal

[0202] Table 2: Model evaluation accuracy

[0203]

[0204] Embodiment 2:

[0205] like Figure 6 As shown, the present invention further provides an error state evaluation device 600 for a current transformer, comprising:

[0206] The acquisition module 601 is used to acquire secondary current data of the current transformer, and determine the index parameters of the current transformer based on the secondary current data; the index parameters include: zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter;

[0207] A sample construction module 602 is used to determine an additional error introduced by the current transformer, and to construct sample groups of different states based on the additional error;

[0208] A sample calculation module 603 is used to calculate the sample group based on the indicator parameter to obtain a time series data set;

[0209] A feature enhancement module 604 is used to perform feature enhancement on the time series data set based on an attention mechanism to obtain an enhanced data set;

[0210] The error evaluation module is used to extract features from the enhanced data set to obtain a feature vector, and input the feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0211] The error state evaluation device for current transformer provided in the above embodiment can implement the technical solution described in the above current transformer error state evaluation method embodiment. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above current transformer error state evaluation method embodiment, which will not be repeated here.

[0212] like Figure 7 As shown, the present invention also provides an electronic device 700. The electronic device 700 includes a processor 701, a memory 702 and a display 703. Figure 7 Only some components of the electronic device 700 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0213] In some embodiments, the memory 702 may be an internal storage unit of the electronic device 700, such as a hard disk or memory of the electronic device 700. In other embodiments, the memory 702 may also be an external storage device of the electronic device 700, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 700.

[0214] Furthermore, the memory 702 may include both an internal storage unit of the electronic device 700 and an external storage device. The memory 702 is used to store application software installed in the electronic device 700 and various data.

[0215] In some embodiments, the processor 701 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 702, such as the error state evaluation method of the current transformer in the present invention.

[0216] In some embodiments, the display 703 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 703 is used to display information on the electronic device 700 and to display a visual user interface. The components 701-703 of the electronic device 700 communicate with each other via a system bus.

[0217] In some embodiments of the present invention, when the processor 701 executes the error state evaluation program of the current transformer in the memory 702, the following steps may be implemented:

[0218] Acquire secondary current data of the current transformer, and determine index parameters of the current transformer based on the secondary current data; the index parameters include: zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter;

[0219] Determine an additional error introduced by the current transformer, and construct sample groups of different states based on the additional error;

[0220] Based on the indicator parameters, the sample group is calculated to obtain a time series data set;

[0221] Performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset;

[0222] Feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0223] It should be understood that: when the processor 701 executes the error state evaluation program of the current transformer in the memory 702, in addition to the above functions, other functions may also be implemented, and details may refer to the description of the corresponding method embodiments above.

[0224] Furthermore, the embodiment of the present invention does not specifically limit the type of the electronic device 700 mentioned, and the electronic device 700 may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of portable electronic devices include but are not limited to portable electronic devices equipped with IOS, Android, Microsoft or other operating systems. The above-mentioned portable electronic device may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 700 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0225] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the error state evaluation method of the current transformer provided by the above methods is implemented, and the method includes:

[0226] Acquire secondary current data of the current transformer, and determine index parameters of the current transformer based on the secondary current data; the index parameters include: zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter;

[0227] Determine an additional error introduced by the current transformer, and construct sample groups of different states based on the additional error;

[0228] Based on the indicator parameters, the sample group is calculated to obtain a time series data set;

[0229] Performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset;

[0230] Feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

[0231] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0232] The error state evaluation method and device for the current transformer provided by the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for evaluating the error state of a current transformer, It is characterized in that include: Acquire secondary current data of the current transformer, and determine index parameters of the current transformer based on the secondary current data; the index parameters include: zero-sequence unbalance parameter, negative-sequence unbalance parameter and fluctuation unbalance parameter; Determine an additional error introduced by the current transformer, and construct sample groups of different states based on the additional error; Based on the indicator parameters, the sample group is calculated to obtain a time series data set; Performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset; Feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

2. The error state evaluation method of a current transformer according to claim 1, It is characterized in that Determining an index parameter of a current transformer based on the secondary current data includes: Determining a zero-sequence unbalance parameter and a negative-sequence unbalance parameter based on three different A-phase zero-sequence components in the secondary current data; Based on the fluctuation component caused by the three-phase load in the secondary current data, a fluctuation unbalance parameter is determined.

3. The error state evaluation method of a current transformer according to claim 2, It is characterized in that Based on the fluctuation component caused by the three-phase load in the secondary current data, determining the fluctuation unbalance parameter includes: Determine the fluctuation component caused by the A-phase load based on the A-phase current difference at adjacent moments in the secondary current data; Determine the fluctuation component caused by the B-phase load based on the B-phase current difference at adjacent moments in the secondary current data; Determine the fluctuation component caused by the C-phase load based on the C-phase current difference at adjacent moments in the secondary current data; The fluctuation component caused by the A-phase load, the fluctuation component caused by the B-phase load and the fluctuation component caused by the C-phase load are added together to obtain the fluctuation unbalance parameter.

4. The error state evaluation method of a current transformer according to claim 1, It is characterized in that Based on the additional error, a sample group of different states is constructed, including: Different additional errors are added with 1 and then multiplied with the secondary current data to obtain sample groups in different states.

5. The error state evaluation method of a current transformer according to claim 1, It is characterized in that The time series dataset is feature enhanced based on the attention mechanism to obtain an enhanced dataset, including: Based on the attention mechanism, the time series data set is spatially fused to obtain a spatial fused data set; Channel fusion is performed on the spatial fusion data set to obtain an enhanced data set.

6. The error state evaluation method of a current transformer according to claim 5, It is characterized in that The time series dataset is spatially fused based on the attention mechanism to obtain a spatial fused dataset, including: Normalizing the time series data set to obtain a first normalized data set; Inputting the first normalized data set into a multi-scale convolutional model to obtain a first multi-scale feature; Determine the attention value of the first multi-scale feature at three linear projection layers; After weighted calculation of two attention values ​​of the first multi-scale feature in the three linear projection layers, data fusion is performed on the weighted calculation result and the other attention value, and the fusion result is added to the time series data in the time series data set to obtain a spatial fusion data set.

7. The error state evaluation method of a current transformer according to claim 5, It is characterized in that Channel fusion is performed on the spatial fusion data set to obtain an enhanced data set, including: Normalizing the data of different channels in the spatial fusion data set to obtain a second normalized data set; Inputting the second normalized data set into a multi-scale convolutional model to obtain a second multi-scale feature; Determine the attention value of the second multi-scale feature at two linear projection layers; Based on the attention values ​​of the second multi-scale features in the two linear projection layers, data fusion is performed on different channel data in the spatial fusion data set, and the fusion result is added to the channel data to obtain an enhanced data set.

8. The error state evaluation method of a current transformer according to claim 1, It is characterized in that Feature extraction is performed on the enhanced data set to obtain a feature vector, and the feature vector is input into a naive Bayes classifier to obtain an error state evaluation result of the current transformer, including: Normalizing the enhanced data set to obtain a third normalized data set; Performing a dropout operation on the third normalized data set to obtain a down-fitting data set; Inputting the third normalized data set and the down-fit data set into a preset Transformer model for feature extraction to obtain a first feature vector and a second feature vector; Fusing the first feature vector and the second feature vector to obtain a fused feature vector; Inputting the fused feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer; The Transformer model only includes a decoder structure.

9. The error state evaluation method of a current transformer according to any one of claims 1 to 8, It is characterized in that Also includes: When it is determined that an abnormal current transformer exists among the multiple current transformers based on the error state evaluation results of the multiple current transformers, a three-phase contribution index of each current transformer is determined based on the three-phase current amplitudes of the same time sequence in the secondary current data of each current transformer; It is determined that the current transformer with the largest three-phase contribution index among multiple current transformers is the out-of-tolerance current transformer.

10. A current transformer error state evaluation device, It is characterized in that include: An acquisition module, used for acquiring secondary current data of a current transformer, and determining index parameters of the current transformer based on the secondary current data; the index parameters include: a zero-sequence unbalance parameter, a negative-sequence unbalance parameter, and a fluctuation unbalance parameter; A sample construction module, used to determine an additional error introduced by the current transformer, and to construct sample groups of different states based on the additional error; A sample calculation module, used to calculate the sample group based on the indicator parameter to obtain a time series data set; A feature enhancement module, used for performing feature enhancement on the time series dataset based on an attention mechanism to obtain an enhanced dataset; The error evaluation module is used to extract features from the enhanced data set to obtain a feature vector, and input the feature vector into a naive Bayes classifier to obtain an error state evaluation result of the current transformer.

Images