Intelligent judgment method and system for comprehensive performance of electric energy meter

Abstract

The invention discloses an intelligent judgment method and system for comprehensive performance of an electric energy meter. The method comprises the following steps: preprocessing test data of comprehensive performance detection of the electric energy meter based on a rough set theory, and extracting key features from the preprocessed test data through a feature extraction technology of scale invariant feature transformation to obtain key features of the test data; and performing comprehensive analysis on the preprocessed test data and key features to obtain a comprehensive analysis result, and performing comprehensive judgment on the comprehensive performance of the electric energy meter according to the comprehensive analysis result to realize intelligent judgment on the comprehensive performance of the electric energy meter. The accuracy and the stability of the feature points are ensured, so that a reliable data basis is provided; information of different levels is effectively integrated, and comprehensiveness and accuracy of comprehensive evaluation are improved; samples with abnormal performance are effectively identified, and the accuracy of abnormal identification is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electric energy metering, and in particular to an intelligent method and system for determining the comprehensive performance of an electric energy meter. Background Art

[0002] According to the quality control requirements of smart electric energy meters, the quality inspection of metering equipment is divided into full performance inspection before arrival, sampling inspection after arrival, and full inspection and acceptance inspection after arrival. Among them, the full performance test items are the most, up to fifty or sixty items, including accuracy requirement test, electrical requirement test, insulation, electromagnetic compatibility test, fee control safety test, climate impact test, mechanical test, etc. The traditional electric energy meter performance detection method mainly relies on manual testing and manual judgment. This method has many shortcomings, including complex data processing process, difficulty in achieving full automation, large amount of data leading to low efficiency of manual processing, and difficulty in accurately identifying and judging abnormal phenomena. These problems make it difficult for the efficiency and accuracy of comprehensive performance detection of electric energy meters to meet the development needs of modern smart grids. Traditional data processing methods usually require manual intervention to analyze and judge a large amount of test data. Due to the huge amount of data and the inclusion of a lot of redundant information, manual processing is not only time-consuming and labor-intensive, but also prone to errors, affecting the reliability of the test results. In addition, manual detection methods have limitations in identifying abnormal phenomena, often relying on the experience of detection personnel, and it is difficult to ensure consistency and accuracy.

[0003] With the development of smart grid and Internet of Things technologies, the performance requirements of electric energy meters are constantly increasing, and detection methods also need to be upgraded accordingly. Although the existing automated detection methods can partially replace manual operations, they are still insufficient in processing complex data and comprehensively evaluating performance. In particular, in terms of the fusion and analysis of multi-dimensional feature data, the scientific calculation of feature weights, and the intelligent determination of comprehensive performance, traditional methods are difficult to achieve efficient and accurate detection.

[0004] Chinese patent application No. 202310170845.4 discloses an invention entitled "A method, device and readable storage medium for checking the performance of an electric energy meter device". The content of the method includes: when receiving a mixed inspection mode calibration task, controlling the installed standard electric energy meter and the calibrated electric energy meter to be installed in at least one calibration unit in the electric energy meter automatic calibration pipeline specified by the mixed inspection mode calibration task according to a preset quantity ratio, and performing the experimental project of the mixed inspection mode calibration task. At the end of the execution of the mixed inspection mode calibration task, obtaining the first calibration data generated by the installed standard electric energy meter based on the experimental project and the second calibration data generated by the calibrated electric energy meter based on the experimental project, and analyzing the calibration performance of the calibration unit according to the first calibration data. If the calibration performance of the calibration unit is qualified, analyzing the metering performance of the calibrated electric energy meter according to the second calibration data. The above method can verify the calibration performance of the calibration unit while calibrating the installed standard electric energy meter, and at the same time does not affect the calibration of ordinary calibrated electric energy meters, thereby improving the flexibility and convenience of calibration.

[0005] However, the above technology has at least the following technical problems: the intelligent determination process of the comprehensive performance of the electricity meter is complex and difficult to achieve full automation, the manual processing efficiency is low due to the large amount of data, and abnormal phenomena are difficult to accurately identify and determine. Summary of the invention

[0006] The present invention provides an intelligent determination method for comprehensive performance of an electric energy meter, so as to solve the technical problems that the existing intelligent determination process for comprehensive performance of an electric energy meter is complex and difficult to realize full automation, the manual processing efficiency is low due to the large amount of data, and abnormal phenomena are difficult to accurately identify and determine.

[0007] To this end, the present invention adopts the following technical solution.

[0008] In a first aspect, the present invention provides an intelligent method for determining the comprehensive performance of an electric energy meter, comprising:

[0009] Step S1, preprocessing the test data of the comprehensive performance detection of the electric energy meter based on the rough set theory, extracting key features from the preprocessed test data through the feature extraction technology of scale-invariant feature transformation, and obtaining the key features of the test data;

[0010] Step S2, comprehensively analyzing the pre-processed test data and key features to obtain comprehensive analysis results, and comprehensively judging the comprehensive performance of the electric energy meter based on the comprehensive analysis results to achieve intelligent judgment of the comprehensive performance of the electric energy meter.

[0011] Furthermore, in step S1, the test data of the comprehensive performance detection of the electric energy meter is preprocessed based on the rough set theory, specifically including:

[0012] The test data of the comprehensive performance detection of the electric energy meter is obtained by collecting the test data in real time through the integrated test device. The test data of the comprehensive performance detection of the electric energy meter is preprocessed using the rough set feature preprocessing algorithm based on the rough set theory to remove redundant information and extract important features.

[0013] Furthermore, in step S1, attribute reduction is performed by calculating the importance of attributes in the rough set feature preprocessing algorithm.

[0014] Furthermore, in step S1, the test data for the comprehensive performance detection of the electric energy meter includes AC withstand voltage test data, pulse voltage test data, constant magnetic field test data, power frequency magnetic field test data, surge immunity test data and electrical fast transient pulse group immunity test data.

[0015] Furthermore, in step S1, the key features are extracted from the preprocessed test data by a feature extraction technique of scale-invariant feature transformation, and the specific process is as follows:

[0016] Feature point detection is performed on the preprocessed test data using feature extraction technology using scale-invariant feature transformation; to improve the accuracy and stability of feature points, each potential feature point needs to be accurately located; to ensure the rotational invariance of feature points, a main direction needs to be assigned to each feature point; after the feature points are assigned directions, feature descriptors are generated based on the feature points that have been assigned directions.

[0017] Furthermore, in step S2, the preprocessed test data and key features are comprehensively analyzed using a multi-level feature fusion and adaptive evaluation algorithm to obtain a comprehensive analysis result.

[0018] Furthermore, in step S2, in the multi-level feature fusion and adaptive evaluation algorithm, the preprocessed test data and key features are integrated into a multi-dimensional feature matrix; nonlinear transformation functions and logarithmic transformations are used to perform feature transformations to enhance the separability between features; the multi-dimensional feature matrix is ​​decomposed into multiple levels for fusion; the feature weights are calculated using information entropy and mutual information; based on the fused feature matrix and feature weights, a multivariate scoring function is introduced to calculate the comprehensive performance score of each sample; in order to identify outliers, anomaly detection based on the three standard deviation method is used.

[0019] Furthermore, the multidimensional feature matrix is ​​decomposed into three levels for fusion. For the features of each level, the fused feature matrix is ​​calculated respectively. The calculation formula of the feature matrix is ​​as follows:

[0020]

[0021] Among them, F is the fused feature matrix, which represents the final feature set after 3-level fusion; Tk represents the feature matrix of the kth level; α k is the layer fusion coefficient, which indicates the weight of each layer in the fusion process; W k is the weight matrix of the kth level, indicating the importance weight of each feature at the kth level;

[0022] The formula for calculating feature weights using information entropy and mutual information is as follows:

[0023]

[0024] Among them, f ij is the jth eigenvalue of the i-th sample of the fused feature matrix F; n is the number of samples; is the number of features; w j is the weight of the jth feature; represents the normalization of the jth eigenvalue of the i-th sample; Represents the information entropy of the jth feature.

[0025] Furthermore, the calculation formula for the comprehensive performance score of each sample is as follows:

[0026]

[0027] Among them, S i is the comprehensive performance score of the i-th sample; σ ij is the standard deviation of the jth feature of the ith sample; log(f ij +1) is the eigenvalue f ij Perform logarithmic transformation; is the eigenvalue f ij After a combination of logarithmic and square root transformations; The eigenvalue f ij After various nonlinear transformations, it is combined with the standard deviation;

[0028] The specific content of anomaly detection based on the three-sigma method is: set the threshold τ = [τ1, τ2] as the mean μ of the comprehensive performance score S Plus or minus three times the standard deviation 3σ S :

[0029] τ1=μ S +3σ S

[0030] τ2=τ S -3σ S

[0031] Among them, μ S and σ S are the mean and standard deviation of the comprehensive performance score, respectively;

[0032] Finally, the comprehensive performance score of each sample is compared with the set threshold τ to evaluate the performance indicators of the electric energy meter; if S i If the threshold τ1 or τ2 is exceeded, the sample performance is judged to be abnormal, otherwise the sample performance is judged to be normal.

[0033] In a second aspect, the present invention provides an intelligent determination system for comprehensive performance of an electric energy meter, which is used to implement the above-mentioned intelligent determination method for comprehensive performance of an electric energy meter, and comprises:

[0034] Key feature acquisition unit: pre-process the test data of comprehensive performance detection of electric energy meters based on rough set theory, extract key features from the pre-processed test data through feature extraction technology of scale-invariant feature transformation, and obtain key features of the test data;

[0035] Intelligent performance judgment unit: conducts comprehensive analysis on the pre-processed test data and key features to obtain comprehensive analysis results, and conducts a comprehensive judgment on the comprehensive performance of the electric energy meter based on the comprehensive analysis results to achieve intelligent judgment on the comprehensive performance of the electric energy meter.

[0036] The present invention has the following beneficial effects:

[0037] 1. The rough set feature preprocessing algorithm based on rough set theory is used to preprocess the test data of the comprehensive performance test of the electric energy meter. By removing redundant information and extracting important features, the efficiency and accuracy of subsequent data analysis are greatly improved. The scale-invariant feature transformation technology is used to extract key features, effectively detect and locate feature points, ensure the accuracy and stability of feature points, and thus provide a reliable data foundation.

[0038] 2. Integrate the preprocessed test data and key features into a multi-dimensional feature matrix, use nonlinear transformation and logarithmic transformation to enhance the separability between features, decompose the feature matrix into three levels: low-level features, middle-level features, and high-level features, and use the fusion formula to perform multi-level feature fusion, effectively integrating information at different levels and improving the comprehensiveness and accuracy of the comprehensive evaluation.

[0039] 3. Use information entropy method and mutual information to calculate the importance weight of features, accurately reflect the importance of each feature in the comprehensive evaluation, and provide a scientific basis for comprehensive scoring; introduce complex multivariate scoring functions, through nonlinear transformation and multivariate combination operations, comprehensively consider the importance of different features and levels, and calculate the comprehensive performance score of each sample; adopt anomaly detection based on the three-sigma method, by setting thresholds, effectively identify samples with abnormal performance, and improve the accuracy of anomaly identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1This is a flow chart of an intelligent method for determining the comprehensive performance of an electric energy meter according to the present invention;

[0041] Figure 2 The present invention is a structural diagram of an intelligent system for determining the comprehensive performance of an electric energy meter. DETAILED DESCRIPTION

[0042] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all 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.

[0043] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0044] Example 1

[0045] Refer to the attached Figure 1 , which shows a flow chart of an intelligent method for determining the comprehensive performance of an electric energy meter provided by an embodiment of the present invention, the method comprising the following steps:

[0046] S1. Based on the rough set theory, the test data of the comprehensive performance test of the electric energy meter is preprocessed, and the key features of the preprocessed test data are extracted by the feature extraction technology of scale-invariant feature transformation to obtain the key features of the test data.

[0047] The test data of the comprehensive performance test of the electric energy meter is collected in real time through the integrated test device, and the test data of the comprehensive performance test of the electric energy meter can include AC withstand voltage test data, pulse voltage test data, constant magnetic field test data, power frequency magnetic field test data, surge immunity test data, electrical fast transient pulse group immunity test data, etc. The test data of the comprehensive performance test of the electric energy meter is preprocessed using a rough set feature preprocessing algorithm based on rough set theory to remove redundant information and extract important features, thereby improving the efficiency and accuracy of subsequent data analysis.

[0048] The specific implementation process of preprocessing is as follows:

[0049] The test data of the comprehensive performance test of the electric energy meter is represented as a decision table. The test data set of the comprehensive performance test of the electric energy meter is recorded as D = {d1, d2, ..., d n}, where n represents the number of electric energy meters, and each data point d i Contains m attributes, namely d i={a1,a2,…,a m}. The decision table is in the form of:

[0050] Γ=(U,A,V,f)

[0051] Wherein: U is the domain, i.e. the sample set of the electric energy meter, which is the test data set for the comprehensive performance detection of the electric energy meter, U=D; A is the attribute set, i.e. the various indicators of the test; V is the range of the attribute value, which indicates the value range of each attribute in the test; f is the mapping from the sample set and the attribute set to the attribute value range, which defines the specific measurement value of each electric energy meter sample on each attribute.

[0052] We further define upper and lower approximation sets to handle the uncertainty and redundant information of data. is a subset of attributes, is a subset of the object. Define the upper approximation set of P on X and lower approximation set as follows:

[0053]

[0054]

[0055] in, It is an upper approximate set. It contains all objects that may belong to set X on the attribute set P. Object x in set U has a large similarity with at least one object y in set X (i.e. similarity μ([x] P ,y)>0); It is a lower approximation set. It contains all objects that definitely belong to set X on the attribute set P. Object x in set U has a high similarity with at least one object y in set X (i.e. similarity μ([x] P ,y)=1);[x] P represents the equivalence class of object x on attribute set P, which is the set of all objects equivalent to x on P. These objects have similar characteristics on attribute P; μ([x] P ,y) is a similarity function, which is used to measure the similarity between object x and object y on attribute set P. Its value is between 0 and 1, and the larger the value, the higher the similarity; it is defined as:

[0056]

[0057] Among them, a is an attribute, which represents a test indicator of the electric energy meter; f(x,a) and f(y,a) represent the values ​​of objects x and y on attribute α, σ a It is the standard deviation of attribute a, which indicates the range of variation of attribute a in the entire data set.

[0058] Furthermore, attribute reduction is performed by calculating the importance of attributes. The importance γ(α) of attribute a∈A is defined as:

[0059]

[0060] Among them, γ(a) is the importance of attribute a, which is used to measure the relative importance of a certain attribute in the data set; U is the domain, that is, the sample set of electric energy meters, which represents the data set of all electric energy meters to be tested; U / E is the division of the sample set U of electric energy meters based on the decision attribute E, that is, the set after the samples are classified according to the decision attribute; the decision attribute E is the key attribute used to classify and evaluate the electric energy meter. In the comprehensive performance detection of electric energy meters, decision attributes can include whether it is qualified, performance level, and fault type; It is a subset of objects, belongs to a subset in U / E, and represents a class of samples; is the size of the lower approximation set, indicating that on the attribute set P, the subset The number of samples in ; It is the logarithm of the size of the lower approximation set, which is used to compress the data range and reduce the impact of the maximum value on the result; It is the average of the sum of squares of all attribute values, which represents the sum of squares of attribute values, and then normalized by dividing it by the number of samples n to reflect the comprehensive importance of each attribute; a i is the i-th attribute value, indicating the specific value of a test indicator of the electric energy meter; n is the number of samples, indicating the total number of electric energy meters in the data set; |U| is the size of the domain, that is, the total number of electric energy meters in the sample set U.

[0061] Based on the importance of the attribute γ(a), a threshold is set and the attribute subset whose attribute importance is greater than the threshold is used as the reduced set

[0062] Extract multi-level decision rules from the simplified attribute set. The process of extracting decision rules is as follows:

[0063] According to the different value combinations in the reduced set R, condition C is generated. According to the generated condition C and the decision attribute E, the support and confidence of each combination are calculated. The specific calculation is as follows:

[0064]

[0065]

[0066] Among them, sup l (C→E) represents the multi-level support of decision attribute E under condition C; conf l (C→E) represents the multi-level confidence of decision attribute E under condition C; is a sample of an electricity meter; l is the hierarchical level, indicating the level of rule extraction, increasing the sense of hierarchy and computational complexity; is the sample the absolute value of the specific attribute value; |C| is the size of the conditional attribute set C, that is, the number of samples satisfying condition C.

[0067] Select rules whose support and confidence exceed the preset thresholds. The rules are extracted from the reduced attribute set, that is, by constructing a multi-level decision table Γ ′ , generating condition C and decision attribute E.

[0068] Based on the extracted rules, perform data dimensionality reduction and retain the features crucial for performance evaluation. The finally preprocessed dataset D ′ is expressed as:

[0069] D ′ ={d ′ 1,d ′ 2,…,d ′ n}

[0070] where d i ′ ={a ′ 1,a ′ 2,…,a ′ k}, k < m, and a j ′ ∈R.

[0071] Based on the preprocessed experimental data, apply the feature extraction technology of Scale-Invariant Feature Transform (SIFT) to extract key features. First, use the SIFT feature extraction technology to detect feature points in the preprocessed experimental data. By constructing a Gaussian pyramid, perform Gaussian blur processing of different scales on the preprocessed experimental data through the Difference of Gaussians method to find extreme points in multiple scale spaces. The extreme points are potential features, thereby detecting potential feature points in the preprocessed experimental data.

[0072] To improve the accuracy and stability of feature points, it is necessary to accurately locate each potential feature point. Accurately calculate the actual position of each key point through interpolation methods to determine the position of key points with sub-pixel accuracy. At the same time, filter out key points with too low contrast to reduce the influence of noise, and eliminate key points with strong but unstable edge responses through principal curvature analysis, thereby ensuring the reliability of feature points and achieving accurate positioning of feature points.

[0073] In order to ensure the rotation invariance of feature points, a main direction needs to be assigned to each feature point. First, the gradient direction and gradient magnitude are calculated in the local neighborhood where the feature point is located. Then, a direction histogram containing Q directions is constructed, and the gradient magnitude in each direction is counted. Finally, the direction with the largest magnitude in the histogram is selected as the main direction of the feature point and the main direction is assigned to the feature point. This process ensures the stability of the feature point under rotation transformation.

[0074] After the feature points are assigned directions, feature descriptors are generated based on the feature points that have been assigned directions. Feature descriptors are detailed descriptions of the local neighborhood of feature points, which are used for subsequent matching and recognition. In this step, a K×K-pixel window is constructed in the local neighborhood of the feature point and the window is divided into In each sub-region, the gradient histogram in each direction is calculated to generate a multi-dimensional feature vector. Subsequently, each multi-dimensional feature vector is normalized to reduce the impact of illumination changes and contrast changes.

[0075] Through the above steps, the scale-invariant feature transform (SIFT) technology is applied to the pre-processed test data to effectively extract key features. The key features contain important information in the test data, such as abnormal phenomena and changes in performance indicators. The extracted key features will serve as an important input for subsequent intelligent algorithm analysis, providing a reliable data basis for the intelligent determination of the comprehensive performance of the electric energy meter.

[0076] S2. Comprehensively analyze the pre-processed test data and key features to obtain comprehensive analysis results, and make a comprehensive judgment on the comprehensive performance of the electric energy meter based on the comprehensive analysis results to achieve intelligent judgment of the comprehensive performance of the electric energy meter.

[0077] The preprocessed test data and key features are comprehensively analyzed using multi-level feature fusion and adaptive evaluation algorithms to obtain the analysis results. The specific implementation process is as follows:

[0078] The preprocessed test data and key features are integrated into a multidimensional feature matrix Z, which represents the preprocessed test data and the extracted key features:

[0079]

[0080] Among them, z ij represents the jth eigenvalue of the i-th sample, the number of samples is n, and the number of features is The multi-dimensional feature matrix lays the foundation for subsequent feature transformation and fusion.

[0081] Next, nonlinear transformation function and logarithmic transformation are used to transform the features to enhance the separability between features:

[0082] T=φ(log(Z+1))

[0083] Among them, φ is a nonlinear transformation function (such as the hyperbolic tangent function), and log(Z+1) is a logarithmic transformation, which is used to prevent the zero value problem in logarithmic operations. Nonlinear transformation can map data to the range of [-1,1] and improve the distribution of data.

[0084] Next, the feature matrix T is decomposed into multiple levels for fusion. Specifically, T is decomposed into three levels: low-level features, which represent basic physical quantities (such as voltage and current); middle-level features, which represent features after some simple transformations (such as voltage fluctuation rate and current harmonic components); and high-level features, which represent features after complex transformations (such as the phase difference of voltage and current and complex waveform features).

[0085] For each level of features, calculate the fused feature matrix. The feature fusion formula is as follows:

[0086]

[0087] Among them, F is the fused feature matrix, which represents the final feature set after three-level fusion and is used to comprehensively evaluate the performance of the power transformer; T k represents the feature matrix of the kth level; α k is the hierarchical fusion coefficient, which indicates the weight of each level in the fusion process and is used to adjust the contribution of each level feature to the final fusion result; W k is the weight matrix of the kth level, which indicates the importance weight of each feature at the kth level and is used to adjust the influence of the feature at this level. Through multi-level feature fusion, information at different levels can be effectively integrated to improve the comprehensiveness and accuracy of the comprehensive evaluation.

[0088] Then, calculate the importance weight w of each feature j . Use information entropy and mutual information to calculate feature weights:

[0089]

[0090] Among them, f ij is the jth eigenvalue of the i-th sample of the fused feature matrix F; n is the number of samples; is the number of features; w j is the weight of the jth feature; represents the normalization of the jth eigenvalue of the i-th sample; Represents the information entropy of the jth feature.

[0091] Based on the fused feature matrix F and feature weights, a multivariate scoring function is introduced to calculate the comprehensive performance score of each sample:

[0092]

[0093] Among them, S i is the comprehensive performance score of the i-th sample; w j is the weight of the jth feature, reflecting the importance of the feature in the comprehensive evaluation; f ij is the jth eigenvalue of the i-th sample in the fused feature matrix F; σ ij is the standard deviation of the jth feature of the ith sample, indicating the fluctuation of the feature value; log(f ij +1) Logarithmic transformation is performed on the eigenvalues ​​to prevent zero value problems and compress the range of eigenvalues; It is a combination of eigenvalues ​​transformed by logarithm and square root, which enhances the nonlinear expression ability of features; The eigenvalues ​​are subjected to a variety of nonlinear transformations and then combined with the standard deviation to increase the complexity and discrimination ability of the eigenvalues.

[0094] In order to identify outliers, the three-sigma method is used for anomaly detection. The threshold τ = [τ1, τ2] is set as the mean μ of the comprehensive performance score. S Plus or minus three times the standard deviation 3σ S :

[0095] τ1=μ S +3σ S

[0096] τ2=μ S -3σ S

[0097] Among them, μ S and σ S are the mean and standard deviation of the comprehensive score respectively. Through these thresholds, samples with abnormal performance can be effectively identified and the accuracy of abnormal identification can be improved.

[0098] Finally, the comprehensive performance score of each sample is compared with the set threshold τ to evaluate the performance indicators of the electric energy meter. i If the threshold value τ1 or τ2 is exceeded, the sample performance is judged to be abnormal, otherwise the sample performance is judged to be normal. Through this evaluation and judgment process, the comprehensive performance of the electric energy meter can be accurately judged to ensure the efficiency and accuracy of the test results.

[0099] In summary, an intelligent method for determining the comprehensive performance of an electric energy meter has been completed.

[0100] Example 2

[0101] This embodiment provides an intelligent system for determining the comprehensive performance of an electric energy meter. Figure 2As shown, it consists of a key feature acquisition unit and a performance intelligent judgment unit.

[0102] The key feature acquisition unit preprocesses the test data of the comprehensive performance detection of the electric energy meter based on the rough set theory, extracts key features from the preprocessed test data through the feature extraction technology of scale-invariant feature transformation, and obtains the key features of the test data.

[0103] The intelligent performance determination unit performs a comprehensive analysis on the pre-processed test data and key features to obtain a comprehensive analysis result, and performs a comprehensive determination on the comprehensive performance of the electric energy meter based on the comprehensive analysis result, thereby realizing intelligent determination of the comprehensive performance of the electric energy meter.

[0104] In the key feature acquisition unit, the preprocessing of the test data of the comprehensive performance detection of the electric energy meter based on the rough set theory specifically includes:

[0105] The test data of the comprehensive performance detection of the electric energy meter is obtained by collecting the test data in real time through the integrated test device. The test data of the comprehensive performance detection of the electric energy meter is preprocessed using the rough set feature preprocessing algorithm based on the rough set theory to remove redundant information and extract important features.

[0106] The test data for the comprehensive performance detection of the electric energy meter includes AC withstand voltage test data, pulse voltage test data, constant magnetic field test data, power frequency magnetic field test data, surge immunity test data and electrical fast transient pulse group immunity test data.

[0107] In the rough set feature preprocessing algorithm, attribute reduction is performed by calculating the importance of attributes.

[0108] In the key feature acquisition unit, the key features are extracted from the preprocessed test data by using a feature extraction technique of scale-invariant feature transformation. The specific process is as follows:

[0109] Feature point detection is performed on the preprocessed test data using feature extraction technology using scale-invariant feature transformation; to improve the accuracy and stability of feature points, each potential feature point needs to be accurately located; to ensure the rotational invariance of feature points, a main direction needs to be assigned to each feature point; after the feature points are assigned directions, feature descriptors are generated based on the feature points that have been assigned directions.

[0110] In the intelligent performance determination unit, the pre-processed test data and key features are comprehensively analyzed using a multi-level feature fusion and adaptive evaluation algorithm to obtain a comprehensive analysis result.

[0111] Specifically, in the multi-level feature fusion and adaptive evaluation algorithm, the preprocessed test data and key features are integrated into a multi-dimensional feature matrix; nonlinear transformation functions and logarithmic transformations are used to transform features to enhance the separability between features; the multi-dimensional feature matrix is ​​decomposed into multiple levels for fusion; the feature weights are calculated using the information entropy method and mutual information; based on the fused feature matrix and feature weights, a multivariate scoring function is introduced to calculate the comprehensive performance score of each sample; in order to identify outliers, anomaly detection based on the three sigma method is used.

[0112] The multidimensional feature matrix is ​​decomposed into three levels for fusion. For the features of each level, the fused feature matrix is ​​calculated respectively. The calculation formula of the feature matrix is ​​as follows:

[0113]

[0114] Among them, F is the fused feature matrix, which represents the final feature set after 3-level fusion; T k represents the feature matrix of the kth level; α k is the layer fusion coefficient, which indicates the weight of each layer in the fusion process; W k is the weight matrix of the kth level, indicating the importance weight of each feature at the kth level;

[0115] The formula for calculating feature weights using information entropy and mutual information is as follows:

[0116]

[0117] Among them, f ij is the jth eigenvalue of the i-th sample of the fused feature matrix F; n is the number of samples; is the number of features; w j is the weight of the jth feature; represents the normalization of the jth eigenvalue of the i-th sample; Represents the information entropy of the jth feature.

[0118] The comprehensive performance score of each sample is calculated as follows:

[0119]

[0120] Among them, S i is the comprehensive performance score of the i-th sample; σ ij is the standard deviation of the jth feature of the ith sample; log(f ij +1) is the eigenvalue f ij Perform logarithmic transformation; is the eigenvalue f ij After a combination of logarithmic and square root transformations; The eigenvalue f ij After multiple nonlinear transformations, it is combined with the standard deviation.

[0121] The specific content of anomaly detection based on the three-sigma method is: set the threshold τ = [τ1, τ2] as the mean μ of the comprehensive performance score S Plus or minus three times the standard deviation 3σ S :

[0122] τ1=μ S +3σ S

[0123] τ2=μ S -3σ S

[0124] Among them, μ S and σ S are the mean and standard deviation of the comprehensive performance score, respectively;

[0125] Finally, the comprehensive performance score of each sample is compared with the set threshold τ to evaluate the performance indicators of the electric energy meter; if S i If the threshold τ1 or τ2 is exceeded, the sample performance is judged to be abnormal, otherwise the sample performance is judged to be normal.

[0126] The order of the embodiments of the invention is for description only and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0127] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. An intelligent method for determining the comprehensive performance of an electric energy meter, characterized in that: include: Step S1, preprocessing the test data of the comprehensive performance detection of the electric energy meter based on the rough set theory, extracting key features from the preprocessed test data through the feature extraction technology of scale-invariant feature transformation, and obtaining the key features of the test data; Step S2, comprehensively analyzing the pre-processed test data and key features to obtain comprehensive analysis results, and comprehensively judging the comprehensive performance of the electric energy meter based on the comprehensive analysis results to achieve intelligent judgment of the comprehensive performance of the electric energy meter.

2. The intelligent determination method of comprehensive performance of electric energy meter according to claim 1 is characterized in that: In step S1, the test data of the comprehensive performance detection of the electric energy meter is preprocessed based on the rough set theory, specifically including: The test data of the comprehensive performance detection of the electric energy meter is obtained by collecting the test data in real time through the integrated test device. The test data of the comprehensive performance detection of the electric energy meter is preprocessed using the rough set feature preprocessing algorithm based on the rough set theory to remove redundant information and extract important features.

3. The intelligent determination method of comprehensive performance of electric energy meter according to claim 2 is characterized in that: In step S1, attribute reduction is performed by calculating the importance of attributes in the rough set feature preprocessing algorithm.

4. The intelligent determination method of comprehensive performance of electric energy meter according to claim 1 or 2, characterized in that: In step S1, the test data for the comprehensive performance detection of the electric energy meter includes AC withstand voltage test data, pulse voltage test data, constant magnetic field test data, power frequency magnetic field test data, surge immunity test data and electrical fast transient pulse group immunity test data.

5. The intelligent determination method of comprehensive performance of electric energy meter according to claim 1 is characterized in that: In step S1, the key features are extracted from the preprocessed test data by using a feature extraction technique of scale-invariant feature transformation. The specific process is as follows: Feature point detection is performed on the preprocessed test data using feature extraction technology using scale-invariant feature transformation; to improve the accuracy and stability of feature points, each potential feature point needs to be accurately located; to ensure the rotational invariance of feature points, a main direction needs to be assigned to each feature point; after the feature points are assigned directions, feature descriptors are generated based on the feature points that have been assigned directions.

6. The intelligent determination method of comprehensive performance of electric energy meter according to claim 1 is characterized in that: In step S2, the preprocessed test data and key features are comprehensively analyzed using a multi-level feature fusion and adaptive evaluation algorithm to obtain a comprehensive analysis result.

7. The intelligent determination method of comprehensive performance of electric energy meter according to claim 6 is characterized in that: In step S2, in the multi-level feature fusion and adaptive evaluation algorithm, the preprocessed test data and key features are integrated into a multi-dimensional feature matrix; nonlinear transformation functions and logarithmic transformations are used to perform feature transformations to enhance the separability between features; Decompose the multi-dimensional feature matrix into multiple levels for fusion; use information entropy method and mutual information to calculate feature weights; based on The fused feature matrix and feature weights are used to introduce a multivariate scoring function to calculate the comprehensive performance score of each sample; To identify outliers, anomaly detection based on three sigma method is used.

8. The intelligent determination method of comprehensive performance of electric energy meter according to claim 7 is characterized in that: The multidimensional feature matrix is ​​decomposed into three levels for fusion. For the features of each level, the fused feature matrix is ​​calculated respectively. The calculation formula of the feature matrix is ​​as follows: Among them, F is the fused feature matrix, which represents the final feature set after 3-level fusion; T k represents the feature matrix of the kth level; α k is the hierarchical fusion coefficient, which indicates the weight of each level in the fusion process; Wk is the weight matrix of the kth level, which indicates the importance weight of each feature at the kth level; The formula for calculating feature weights using information entropy and mutual information is as follows: Among them, f ij is the jth eigenvalue of the i-th sample of the fused feature matrix F; n is the number of samples; is the number of features; w j is the weight of the jth feature; represents the normalization of the jth eigenvalue of the i-th sample; Represents the information entropy of the jth feature.

9. The intelligent determination method of comprehensive performance of electric energy meter according to claim 8 is characterized in that: The comprehensive performance score of each sample is calculated as follows: Among them, S i is the comprehensive performance score of the i-th sample; σ ij is the standard deviation of the jth feature of the ith sample; log(f ij +1) is the eigenvalue f ij Perform logarithmic transformation; is the eigenvalue f ij After a combination of logarithmic and square root transformations; The eigenvalue f ij After various nonlinear transformations, it is combined with the standard deviation; The specific content of anomaly detection based on the three-sigma method is: set the threshold τ = [τ1, τ2] as the mean μ of the comprehensive performance score S Plus or minus three times the standard deviation 3σ S : τ1=μ S +3s S τ2=μ S -3s S Among them, μ S and σ S are the mean and standard deviation of the comprehensive performance score, respectively; Finally, the comprehensive performance score of each sample is compared with the set threshold τ to evaluate the performance indicators of the electric energy meter; if S i If the threshold τ1 or τ2 is exceeded, the sample performance is judged to be abnormal, otherwise the sample performance is judged to be normal.

10. An intelligent determination system for comprehensive performance of an electric energy meter, used to implement the intelligent determination method for comprehensive performance of an electric energy meter according to any one of claims 1 to 9, characterized in that: include: Key feature acquisition unit: pre-process the test data of comprehensive performance detection of electric energy meters based on rough set theory, extract key features from the pre-processed test data through feature extraction technology of scale-invariant feature transformation, and obtain key features of the test data; Intelligent performance judgment unit: conducts comprehensive analysis on the pre-processed test data and key features to obtain comprehensive analysis results, and conducts a comprehensive judgment on the comprehensive performance of the electric energy meter based on the comprehensive analysis results to achieve intelligent judgment on the comprehensive performance of the electric energy meter.

Images