Version difference analysis method and system of metering automation system

Abstract

The invention discloses a version difference analysis method and system for a metering automation system, and the method comprises the steps: collecting a plurality of feature values of the metering automation system under a current version to construct a correlation coefficient matrix, and determining a version weight corresponding to each feature value according to the correlation coefficient matrix, obtaining a plurality of real-time feature values corresponding to the current version according to the version weights and the feature values; performing nonlinear mapping on each real-time feature value according to a preset multi-dimensional feature segmentation nonlinear mapping function to obtain a plurality of mapping feature values corresponding to the current version; performing spectral decomposition on the plurality of mapping feature values to obtain a plurality of feature vectors corresponding to the current version, and dynamically adjusting a fusion weight corresponding to each feature vector; and obtaining a plurality of reference feature vectors corresponding to the reference version, so as to determine difference information between the current version and the reference version according to the fusion weight, the reference feature vectors and the feature vectors, thereby improving the precision of version difference analysis.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and specifically to a method and system for analyzing version differences in a metering automation system. Background Technology

[0002] With the development of technology, metrological automation systems that use various sensors, actuators, controllers and software to measure and control physical quantities are gradually being widely used in industrial, laboratory and other environments to improve work efficiency and accuracy.

[0003] In the widespread use of metrology automation systems, the base version of the metrology automation system will be updated as the developers improve and control it. Therefore, managing the current version of the metrology automation system and analyzing the differences between it and the standard version are key to ensuring stable operation.

[0004] However, existing technologies primarily rely on manual inspection and simple version number comparison for version difference analysis. Manual inspection is not only time-consuming and labor-intensive, but also prone to inaccurate results due to human factors. It struggles to handle complex system environments and large-scale version updates, thus failing to comprehensively evaluate the performance and functionality of the metrology automation system, impacting the overall system performance and reliability. While version number-based difference analysis techniques typically focus only on version numbers or differences in certain functional modules, they cannot comprehensively assess the metrology automation system's performance and compatibility with the current base version, nor can they provide detailed difference analysis reports. Consequently, they fail to meet the demands of modern metrology automation systems for efficient and intelligent management. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention discloses a method and system for version difference analysis of a metrology automation system, which improves the accuracy of version difference analysis.

[0006] To achieve the above objectives, this invention discloses a method for version difference analysis of a metrology automation system, comprising:

[0007] The system collects multiple feature values ​​of the automated metering system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension;

[0008] Construct a correlation coefficient matrix corresponding to multiple eigenvalues, and select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix;

[0009] Based on the eigenvalues ​​of the principal component eigenvalues, the version weight corresponding to each eigenvalue is determined, so as to obtain multiple real-time eigenvalues ​​corresponding to the current version based on the version weights and the eigenvalues.

[0010] Each real-time feature value is nonlinearly mapped according to a preset multidimensional feature segmentation nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version.

[0011] Spectral decomposition is performed on the multiple mapping feature values ​​to obtain multiple feature vectors corresponding to the current version, and the fusion weights corresponding to each feature vector are dynamically adjusted.

[0012] Multiple reference feature vectors corresponding to the reference version are obtained, and the difference information between the current version and the reference version is determined based on the fusion weight, the reference feature vectors, and the feature vectors.

[0013] This invention discloses a method for version difference analysis of a metrology automation system. By extracting multi-dimensional feature values ​​of the metrology automation system across different versions, and comparing these feature values, a comprehensive version difference analysis can be performed, improving the accuracy of the analysis. Specifically, during the version difference analysis process, multiple feature values ​​of the metrology automation system under the current version and different dimensions are first extracted. Simultaneously extracting multiple feature values ​​under both functional and performance dimensions comprehensively reveals the characteristics of the metrology automation system in the current version, improving the accuracy of the version analysis. Next, considering that version updates involve optimizations in different dimensions, to better reflect the changes in features across these dimensions, a correlation coefficient matrix is ​​constructed between multiple feature values. Principal component feature values ​​are extracted based on the feature values ​​of this correlation coefficient matrix. The weight of each feature value in the current version is dynamically adjusted using these principal component feature values. Furthermore, multiple real-time feature values ​​corresponding to the current version are generated using the dynamically adjusted weights and the original feature values, thereby improving the accuracy of feature extraction. Furthermore, during version difference analysis, feature values ​​of different magnitudes have varying degrees of importance in similarity calculations. To improve the accuracy of the difference analysis, a multidimensional feature piecewise nonlinear mapping function is used to nonlinearly map each real-time feature value, adjusting its importance in the similarity calculation process. Finally, spectral decomposition is performed on multiple mapped feature values ​​to obtain the fusion weights for the current version. These weights are then combined with multiple reference feature vectors corresponding to the obtained reference version to determine the difference information between the current version and the reference version, thereby improving the accuracy of the difference analysis.

[0014] As a preferred example, the automated data acquisition and measurement system in its current version includes several characteristic values, including:

[0015] Based on several functional evaluation indicators corresponding to the functional dimensions, the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators are obtained; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators and data storage indicators.

[0016] Based on several performance evaluation indicators corresponding to the performance dimension, the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators are obtained; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators, and system throughput indicators.

[0017] The above scheme extracts comprehensive features of the metrology automation system in the current version by collecting feature values ​​from functional and performance dimensions, thereby improving the accuracy of version difference analysis. During the feature value extraction process, different evaluation indicators are set for different dimensions to ensure that the extracted feature values ​​fully reflect the characteristics of the current dimension. This provides a comprehensive and accurate basis for subsequent difference analysis, thus improving the accuracy of the difference analysis.

[0018] As a preferred example, the step of constructing a correlation coefficient matrix corresponding to multiple eigenvalues ​​and selecting multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix includes:

[0019] Each of the functional feature values ​​and each of the performance feature values ​​are normalized to obtain multiple normalized functional feature values ​​and multiple normalized performance feature values.

[0020] Based on the normalized functional eigenvalues ​​and the normalized performance eigenvalues, obtain the Pearson correlation coefficient between any two eigenvalues, and construct a correlation coefficient matrix corresponding to multiple eigenvalues ​​based on the Pearson correlation coefficients;

[0021] The eigenvalues ​​corresponding to each eigenvalue are obtained according to the correlation coefficient matrix, and multiple principal component eigenvalues ​​are selected from the multiple eigenvalues ​​based on the eigenvalues ​​and the cumulative contribution rate of the multiple eigenvalues.

[0022] The above scheme first eliminates dimensional differences between different eigenvalues ​​through normalization, enabling accurate difference analysis based on eigenvalues ​​with unified dimensions. Then, a matrix is ​​constructed using the Pearson correlation coefficient to discover hidden correlations between eigenvalues. These hidden correlations are then used to determine the most critical features, improving the efficiency and accuracy of feature extraction. Finally, principal components are selected using eigenvalues ​​and cumulative contribution rates, automatically focusing on the most critical features and avoiding the subjectivity and bias of manual selection. This provides an accurate basis for adjusting the weights of each eigenvalue, thereby improving the precision of weight adjustment.

[0023] As a preferred example, the step of determining the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component features, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values, includes:

[0024] Obtain the sum of the eigenvalues ​​of the principal components corresponding to the eigenvalues ​​of the principal components;

[0025] For any one of the principal component eigenvalues:

[0026] The ratio of the eigenvalue corresponding to the principal component eigenvalue to the sum of the eigenvalues ​​is obtained, and the ratio is used as the variance contribution rate of the principal component eigenvalue.

[0027] For any one of the aforementioned eigenvalues:

[0028] Obtain the squared loading of the eigenvalue in each of the principal component eigenvalues, and multiply the squared loading by the variance contribution rate corresponding to each of the principal component eigenvalues ​​to obtain the product of multiple load sharing rates corresponding to the eigenvalue.

[0029] The version weight corresponding to the feature value is obtained by summing the products of multiple load sharing rates.

[0030] Each of the aforementioned feature values ​​is multiplied by its corresponding version weight to obtain multiple real-time feature values ​​corresponding to the current version.

[0031] In the above scheme, based on the sum of the eigenvalues ​​of the extracted principal component eigenvalues ​​and the variance contribution rate of each principal component eigenvalue, the product of the squared loading of the eigenvalue in each principal component eigenvalue and the variance contribution rate corresponding to the principal component eigenvalue is obtained. Then, the sum of the multiple products corresponding to a eigenvalue is used as the weight corresponding to the eigenvalue. In this way, the importance of each eigenvalue is determined by the degree of correlation between each eigenvalue and the principal component eigenvalue, thereby improving the accuracy of weight allocation and thus improving the accuracy of version difference analysis.

[0032] As a preferred example, the step of performing nonlinear mapping on each of the real-time feature values ​​according to a preset multidimensional feature segmentation nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version includes:

[0033] Based on the multiple preset feature value intervals in the multidimensional feature segmentation nonlinear mapping function, the multiple real-time feature values ​​are divided into a first interval feature value, a second interval feature value, and a third interval feature value.

[0034] The first nonlinear transformation function corresponding to the first interval feature value, the second nonlinear transformation function corresponding to the second interval feature value, and the third nonlinear transformation function corresponding to the third interval feature value are retrieved from the multidimensional feature segmentation nonlinear mapping function, so as to perform nonlinear mapping on the first interval feature value, the second interval feature value, and the third interval feature value respectively, to obtain multiple mapped feature values.

[0035] The above scheme preprocesses the feature values ​​of each dimension using a pre-defined multi-dimensional feature segmentation nonlinear mapping function. Different mapping rules are applied based on the interval in which the feature value lies, highlighting the feature differences in key intervals. Specifically, for feature values ​​in low-value intervals, since their impact on system performance is relatively small, the mapping function compresses them, reducing their weight in similarity calculation and preventing small fluctuations from having an excessive impact. For feature values ​​in mid-value intervals, their relative changes are maintained to accurately reflect system performance differences. For feature values ​​in high-value intervals, since even small changes can have a significant impact on system performance, the mapping function amplifies their value, emphasizing their importance in similarity calculation. This approach allows feature values ​​from different value spaces to play a reasonable role in similarity calculation, improving the accuracy of version difference analysis.

[0036] As a preferred example, the step of performing spectral decomposition on the multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjusting the fusion weights corresponding to each feature vector, includes:

[0037] Obtain the correlation coefficients between each pair of the multiple mapping feature values ​​to construct a mapping correlation coefficient matrix corresponding to the multiple mapping feature values;

[0038] The mapping correlation coefficient matrix is ​​subjected to spectral decomposition to obtain multiple first spectral decomposition eigenvalues ​​and a first eigenvector corresponding to each first spectral decomposition eigenvalue;

[0039] Based on the magnitude of the first spectral decomposition eigenvalue, several second spectral decomposition eigenvalues ​​and a second eigenvector corresponding to each second spectral decomposition eigenvalue are selected from the multiple first spectral decomposition eigenvalues ​​according to the selection rule from large to small.

[0040] Multiple preset initial fusion weights are obtained, and the preset multiple initial fusion weights are dynamically adjusted according to a preset particle swarm optimization algorithm to obtain multiple optimized fusion weights.

[0041] In the above scheme, after obtaining the mapping features with adjusted importance, the correlation coefficient between the mapping features is used to cluster and mine the hidden associations between the features. Then, based on the hidden associations, the feature vector corresponding to each mapping feature and the fusion weight corresponding to each feature vector are determined. Then, the fusion weight vector is dynamically generated based on the fusion weight to improve the accuracy of the difference analysis between versions.

[0042] As a preferred example, obtaining multiple reference feature vectors corresponding to the reference version, and determining the difference information between the current version and the reference version based on the fusion weights, the reference feature vectors, and the feature vectors, includes:

[0043] Based on the optimized fusion weights, the multiple second feature vectors and the multiple reference feature vectors are weighted and fused to obtain the fusion feature value corresponding to the current version and the reference fusion feature value corresponding to the reference version.

[0044] Obtain the fusion similarity between the fusion feature value and the reference fusion feature value;

[0045] When the fusion similarity is greater than or equal to a preset similarity threshold, it is determined that there is no difference between the current version and the reference version.

[0046] The above scheme achieves dynamic adjustment and optimization of feature weights through adaptive feature fusion similarity calculation. The fusion weight vector is automatically adjusted based on the correlation between features and historical data, making the similarity calculation more accurately reflect the actual similarity between versions, thereby improving the accuracy of similarity calculation and thus improving the accuracy of version difference analysis.

[0047] As a preferred example, obtaining multiple reference feature vectors corresponding to the reference version, and determining the difference information between the current version and the reference version based on the fusion weights, the reference feature vectors, and the feature vectors, includes:

[0048] When the fusion similarity is less than the similarity threshold, it is determined that there is a difference between the current version and the reference version, and the difference value between the corresponding second feature vector and the reference feature vector is obtained, so as to extract the difference feature vector from multiple second feature vectors according to the difference value;

[0049] The difference feature vector is mapped back to the original feature space based on the second spectral decomposition feature value, the original feature value corresponding to the difference feature vector is determined, and then the difference dimension is determined based on the original feature value.

[0050] In the above scheme, after matching analysis, due to the fusion and dimensionality reduction of feature vectors, directly comparing the differences of each feature vector cannot accurately reflect the differences of the original features. Therefore, based on the same fusion weight vector and dimensionality reduction matrix used in the process of generating feature vectors, the fused feature vectors are mapped back to the original feature space to determine the specific difference dimensions, thereby improving the accuracy of difference analysis.

[0051] On the other hand, the present invention discloses a version difference analysis system for a metrology automation system, including a multidimensional data acquisition module, a principal component analysis module, a weight adjustment module, a nonlinear mapping module, a feature fusion module, and a difference analysis module;

[0052] The multidimensional data acquisition module is used to acquire multiple feature values ​​of the metering automation system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension;

[0053] The principal component analysis module is used to construct a correlation coefficient matrix corresponding to multiple eigenvalues, and to select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix.

[0054] The weight adjustment module is used to determine the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component feature values, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values;

[0055] The nonlinear mapping module is used to perform nonlinear mapping on each of the real-time feature values ​​according to a preset multidimensional feature piecewise nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version.

[0056] The feature fusion module is used to perform spectral decomposition on multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjust the fusion weights corresponding to each feature vector.

[0057] The difference analysis module is used to obtain multiple reference feature vectors corresponding to the reference version, so as to determine the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors and the feature vectors.

[0058] This invention discloses a version difference analysis system for a metrology automation system. By extracting multi-dimensional feature values ​​of the metrology automation system across different versions, it comprehensively analyzes version differences through comparison of these feature values, thereby improving the accuracy of version difference analysis. Specifically, during the version difference analysis process, multiple feature values ​​of the metrology automation system under the current version and different dimensions are first extracted. Simultaneously extracting multiple feature values ​​under both functional and performance dimensions comprehensively showcases the characteristics of metrology automation in the current version, improving the accuracy of version analysis. Next, considering that version updates involve optimizations in different dimensions, to better reflect the changes in features across these dimensions, a correlation coefficient matrix is ​​constructed between multiple feature values. Principal component feature values ​​are extracted based on the feature values ​​of this correlation coefficient matrix. The weight of each feature value in the current version is dynamically adjusted using these principal component feature values. Furthermore, multiple real-time feature values ​​corresponding to the current version are generated using the dynamically adjusted weights and the original feature values, thereby improving the accuracy of feature extraction. Furthermore, during version difference analysis, feature values ​​of different magnitudes have varying degrees of importance in similarity calculations. To improve the accuracy of the difference analysis, a multidimensional feature piecewise nonlinear mapping function is used to nonlinearly map each real-time feature value, adjusting its importance in the similarity calculation process. Finally, spectral decomposition is performed on multiple mapped feature values ​​to obtain the fusion weights for the current version. These weights are then combined with multiple reference feature vectors corresponding to the obtained reference version to determine the difference information between the current version and the reference version, thereby improving the accuracy of the difference analysis.

[0059] As a preferred example, the multidimensional data acquisition module includes a functional indicator unit and a performance indicator unit;

[0060] The functional indicator unit is used to obtain the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators based on several functional evaluation indicators corresponding to the functional dimension; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators and data storage indicators;

[0061] The performance indicator unit is used to obtain the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators based on several performance evaluation indicators corresponding to the performance dimension; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators and system throughput indicators.

[0062] The above scheme extracts comprehensive features of the metrology automation system in the current version by collecting feature values ​​from functional and performance dimensions, thereby improving the accuracy of version difference analysis. During the feature value extraction process, different evaluation indicators are set for different dimensions to ensure that the extracted feature values ​​fully reflect the characteristics of the current dimension. This provides a comprehensive and accurate basis for subsequent difference analysis, thus improving the accuracy of the difference analysis. Attached Figure Description

[0063] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0064] Figure 1 This is a flowchart illustrating a method for analyzing version differences in a metrology automation system provided in an embodiment of the present invention.

[0065] Figure 2 This is a schematic diagram of the structure of a version difference analysis system for a metrology automation system provided in an embodiment of the present invention;

[0066] Figure 3 This is a flowchart illustrating a method for analyzing version differences in a metrology automation system, provided in another embodiment of the present invention. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0068] Example 1

[0069] Reference Figure 1 This embodiment provides a method for version difference analysis of a metering automation system to improve the accuracy of version difference analysis. Specifically, it includes the following steps:

[0070] Step 101: Collect multiple feature values ​​of the metering automation system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension.

[0071] In this embodiment, the steps mainly include: obtaining the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators based on several functional evaluation indicators corresponding to the functional dimension; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators, and data storage indicators; obtaining the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators based on several performance evaluation indicators corresponding to the performance dimension; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators, and system throughput indicators.

[0072] In this embodiment, the above steps extract comprehensive features of the metering automation system in the current version by collecting feature values ​​of functional and performance dimensions, thereby improving the accuracy of version difference analysis. During the specific feature value extraction process, different evaluation indicators are set for different dimensions to ensure that the extracted feature values ​​fully reflect the characteristics of the current dimension, thus providing a comprehensive and accurate analytical basis for subsequent difference analysis and improving the accuracy of the difference analysis.

[0073] Step 102: Construct a correlation coefficient matrix corresponding to multiple eigenvalues, and select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix.

[0074] In this embodiment, the step mainly includes: normalizing each of the functional feature values ​​and each of the performance feature values ​​to obtain multiple normalized functional feature values ​​and multiple normalized performance feature values; obtaining the Pearson correlation coefficient between any two feature values ​​based on the normalized functional feature values ​​and the normalized performance feature values, and constructing a correlation coefficient matrix corresponding to multiple feature values ​​based on the Pearson correlation coefficient; obtaining the eigenvalues ​​corresponding to each feature value based on the correlation coefficient matrix, and selecting multiple principal component feature values ​​from the multiple feature values ​​based on the eigenvalues ​​and the cumulative contribution rate of the multiple eigenvalues.

[0075] In this embodiment, the above steps eliminate dimensional differences between different eigenvalues ​​through normalization, enabling accurate difference analysis based on eigenvalues ​​with unified dimensions. Then, a matrix is ​​constructed using the Pearson correlation coefficient to discover hidden correlations between eigenvalues. These hidden correlations are then used to determine the most critical features, improving the efficiency and accuracy of feature extraction. Finally, principal components are screened using eigenvalues ​​and cumulative contribution rates, automatically focusing on the most critical features and avoiding the subjectivity and bias of manual selection. This provides an accurate basis for adjusting the weights of each eigenvalue, thereby improving the precision of weight adjustment.

[0076] Step 103: Determine the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component features, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values.

[0077] In this embodiment, the step mainly includes: obtaining the sum of the eigenvalues ​​of the eigenvalues ​​corresponding to the principal component eigenvalues; for any principal component eigenvalue, obtaining the ratio of the eigenvalue corresponding to the principal component eigenvalue to the sum of the eigenvalues, and using the ratio as the variance contribution rate of the principal component eigenvalue; for any eigenvalue, obtaining the squared load of the eigenvalue in each principal component eigenvalue, and multiplying the squared load by the variance contribution rate corresponding to each principal component eigenvalue to obtain the product of the multiple load sharing rates corresponding to the eigenvalue; summing the products of the multiple load sharing rates to obtain the version weight corresponding to the eigenvalue; and multiplying each eigenvalue by its corresponding version weight to obtain the multiple real-time eigenvalues ​​corresponding to the current version.

[0078] In this embodiment, the above steps are based on the sum of the eigenvalues ​​of the extracted principal component eigenvalues ​​and the variance contribution rate of each principal component eigenvalue. The product of the squared load of the eigenvalue in each principal component eigenvalue and the variance contribution rate corresponding to the principal component eigenvalue is obtained. Then, the sum of the multiple products corresponding to a eigenvalue is used as the weight corresponding to the eigenvalue. In this way, the importance of each eigenvalue is determined by the degree of correlation between each eigenvalue and the principal component eigenvalue, thereby improving the accuracy of weight allocation and thus improving the accuracy of version difference analysis.

[0079] Step 104: Perform nonlinear mapping on each of the real-time feature values ​​according to the preset multidimensional feature segmentation nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version.

[0080] In this embodiment, the step mainly includes: dividing the multiple real-time feature values ​​into a first interval feature value, a second interval feature value, and a third interval feature value according to the multiple preset feature value intervals in the multidimensional feature segmentation nonlinear mapping function; retrieving the first nonlinear transformation function corresponding to the first interval feature value, the second nonlinear transformation function corresponding to the second interval feature value, and the third nonlinear transformation function corresponding to the third interval feature value from the multidimensional feature segmentation nonlinear mapping function, so as to perform nonlinear mapping on the first interval feature value, the second interval feature value, and the third interval feature value respectively, to obtain multiple mapped feature values.

[0081] In this embodiment, the above steps preprocess the feature values ​​of each dimension using a preset multi-dimensional feature piecewise nonlinear mapping function. Different mapping rules are applied according to the intervals in which the feature values ​​lie, highlighting the feature differences in key intervals. Specifically, for feature values ​​in low-value intervals, since their impact on system performance is relatively small, they are compressed using the mapping function to reduce their weight in similarity calculation, avoiding excessive influence from small fluctuations. For feature values ​​in mid-value intervals, their relative changes are maintained to accurately reflect system performance differences. For feature values ​​in high-value intervals, since even small changes can have a significant impact on system performance, they are amplified using the mapping function to emphasize their importance in similarity calculation. This approach allows feature values ​​from different value spaces to play a reasonable role in similarity calculation, improving the accuracy of version difference analysis.

[0082] Step 105: Perform spectral decomposition on the multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjust the fusion weights corresponding to each feature vector.

[0083] In this embodiment, the step mainly includes: obtaining the correlation coefficients between pairs of the multiple mapping feature values ​​to construct a mapping correlation coefficient matrix corresponding to the multiple mapping feature values; performing spectral decomposition on the mapping correlation coefficient matrix to obtain multiple first spectral decomposition feature values ​​and a first feature vector corresponding to each first spectral decomposition feature value; selecting several second spectral decomposition feature values ​​and a second feature vector corresponding to each second spectral decomposition feature value from the multiple first spectral decomposition feature values ​​according to the size of the first spectral decomposition feature values ​​and according to the selection rule from large to small; obtaining multiple preset initial fusion weights to dynamically adjust the multiple preset initial fusion weights according to a preset particle swarm optimization algorithm to obtain multiple optimized fusion weights.

[0084] In this embodiment, after obtaining the mapping features with adjusted importance, the above steps use the correlation coefficient between the mapping features to cluster and mine the hidden associations between the features. Then, based on the hidden associations, the feature vector corresponding to each mapping feature and the fusion weight corresponding to each feature vector are determined. Finally, a fusion weight vector is dynamically generated based on the fusion weight to improve the accuracy of the difference analysis between versions.

[0085] Step 106: Obtain multiple reference feature vectors corresponding to the reference version, so as to determine the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors and the feature vectors.

[0086] In this embodiment, the step mainly includes: performing weighted fusion on multiple second feature vectors and multiple reference feature vectors according to the optimized multiple fusion weights to obtain the fusion feature value corresponding to the current version and the reference fusion feature value corresponding to the reference version; obtaining the fusion similarity between the fusion feature value and the reference fusion feature value; when the fusion similarity is greater than or equal to a preset similarity threshold, determining that there is no difference between the current version and the reference version.

[0087] When the fusion similarity is less than the similarity threshold, it is determined that there is a difference between the current version and the reference version, and the difference value between the corresponding second feature vector and the reference feature vector is obtained. Based on the difference value, the difference feature vector is extracted from multiple second feature vectors. The difference feature vector is mapped back to the original feature space based on the second spectral decomposition feature value, the original feature value corresponding to the difference feature vector is determined, and then the difference dimension is determined based on the original feature value.

[0088] In this embodiment, the above steps achieve dynamic adjustment and optimization of feature weights through adaptive feature fusion similarity calculation. The fusion weight vector is automatically adjusted based on the correlation between features and historical data, making the similarity calculation more accurately reflect the actual similarity between versions, thereby improving the accuracy of the similarity calculation and thus improving the accuracy of version difference analysis. After matching analysis, due to the fusion and dimensionality reduction of feature vectors, directly comparing the differences item by item of the feature vectors cannot accurately reflect the differences of the original features. Therefore, based on the same fusion weight vector and dimensionality reduction matrix used in the feature vector generation process, the fused feature vectors are mapped back to the original feature space to determine the specific difference dimensions, thereby improving the accuracy of difference analysis.

[0089] On the other hand, refer to Figure 2 This embodiment also discloses a version difference analysis system for a metrology automation system, including a multidimensional data acquisition module 201, a principal component analysis module 202, a weight adjustment module 203, a nonlinear mapping module 204, a feature fusion module 205, and a difference analysis module 206.

[0090] The multidimensional data acquisition module 201 is used to acquire multiple feature values ​​of the metering automation system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension.

[0091] The principal component analysis module 202 is used to construct a correlation coefficient matrix corresponding to multiple eigenvalues, and to select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix.

[0092] The weight adjustment module 203 is used to determine the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component features, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values.

[0093] The nonlinear mapping module 204 is used to perform nonlinear mapping on each of the real-time feature values ​​according to a preset multidimensional feature segmentation nonlinear mapping function, so as to obtain multiple mapped feature values ​​corresponding to the current version.

[0094] The feature fusion module 205 is used to perform spectral decomposition on multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjust the fusion weights corresponding to each feature vector.

[0095] The difference analysis module 206 is used to obtain multiple reference feature vectors corresponding to the reference version, so as to determine the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors and the feature vectors.

[0096] In this embodiment, the multidimensional data acquisition module 201 includes a functional indicator unit and a performance indicator unit.

[0097] The functional indicator unit is used to obtain the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators based on several functional evaluation indicators corresponding to the functional dimensions; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators and data storage indicators.

[0098] The performance indicator unit is used to obtain the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators based on several performance evaluation indicators corresponding to the performance dimension; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators and system throughput indicators.

[0099] This embodiment provides a method and system for version difference analysis of a metrology automation system. By extracting multi-dimensional feature values ​​of the metrology automation system under different versions, a comprehensive version difference analysis is performed through comparison of these multi-dimensional feature values, improving the accuracy of the version difference analysis. Specifically, during the version difference analysis process, multiple feature values ​​of the metrology automation system under the current version and different dimensions are first extracted. Simultaneously extracting multiple feature values ​​under both functional and performance dimensions comprehensively showcases the characteristics of metrology automation under the current version, improving the accuracy of the version analysis. Next, considering that version updates involve optimizations in different dimensions, to better reflect the changes in features across different dimensions, a correlation coefficient matrix is ​​constructed between multiple feature values. Principal component feature values ​​are extracted based on the feature values ​​of the correlation coefficient matrix. The weight of each feature value in the current version is dynamically adjusted using these principal component feature values. Then, multiple real-time feature values ​​corresponding to the current version are generated using the dynamically adjusted weights and the original feature values, thereby improving the accuracy of feature extraction. Furthermore, during version difference analysis, feature values ​​of different magnitudes have varying degrees of importance in similarity calculations. To improve the accuracy of the difference analysis, a multidimensional feature piecewise nonlinear mapping function is used to nonlinearly map each real-time feature value, adjusting its importance in the similarity calculation process. Finally, spectral decomposition is performed on multiple mapped feature values ​​to obtain the fusion weights for the current version. These weights are then combined with multiple reference feature vectors corresponding to the obtained reference version to determine the difference information between the current version and the reference version, thereby improving the accuracy of the difference analysis.

[0100] Example 2

[0101] In the base version management of metrology automation systems, existing technologies relying on manual inspection are not only time-consuming and labor-intensive, but also prone to inaccurate results due to human factors, making them ill-suited for complex system environments and large-scale version updates. Secondly, existing version comparison technologies typically focus only on version numbers or differences in certain functional modules, failing to comprehensively assess the performance and compatibility of base versions or provide detailed difference analysis reports to help users quickly locate problems. Furthermore, existing technologies lack intelligent version management tools and cannot fully utilize advanced technologies such as big data analytics and cloud computing, making it difficult to meet the demands of modern metrology automation systems for efficient and intelligent management. Simultaneously, when conducting version difference analysis, the inability to comprehensively assess the performance and functional differences across different versions can lead to performance bottlenecks in actual operation, impacting the overall system performance and reliability.

[0102] To solve the above technical problems, refer to Figure 3This embodiment provides a method for version difference analysis of a metrology automation system, which improves the accuracy, efficiency, and detail of version difference analysis. Specifically, the method includes the following steps:

[0103] Step 301: Collect multiple functional feature values ​​in the functional dimension and multiple performance feature values ​​in the performance dimension of the current version of the metering automation system.

[0104] In this embodiment, multiple functional feature values ​​of the current version of the metering automation system are obtained based on multiple functional evaluation indicators corresponding to the functional dimension. Similarly, multiple performance feature values ​​of the current version of the metering automation system are obtained based on multiple performance evaluation indicators corresponding to the performance dimension. The functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators, and data storage indicators; while the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators, and system throughput indicators.

[0105] Specifically, in this embodiment, when considering multiple functional characteristic values ​​corresponding to each evaluation indicator under different functional evaluation indicators corresponding to the data acquisition function dimension, taking the data acquisition indicator as an example, the data acquisition indicator includes the accuracy, completeness, and real-time performance of data acquisition. Accuracy refers to the closeness of the acquired data to the true value; completeness refers to the comprehensiveness of the acquired data, whether it covers all the required information; and real-time performance refers to the timeliness of data acquisition, whether the data can be acquired within the specified time. Specifically, the accuracy of the data acquisition is determined by comparing the acquired data with the true value and calculating the error range (e.g., meter reading error not exceeding ±0.5%). The completeness of the data acquisition is determined by checking whether the acquired data items meet the requirements of the metering automation system and checking for data continuity and missing data. The accuracy of the data acquisition is determined by recording the data acquisition timestamp and comparing whether the time interval from data acquisition to transmission completion is within the specified range (e.g., not exceeding 1 minute). After collecting the three characteristic data corresponding to the accuracy, completeness, and real-time performance of data acquisition, these three characteristic data are standardized so that they are all within the range of [0,1] to eliminate the influence of dimensions and orders of magnitude. For example, assuming the original accuracy is 95% (if the true value is 100, the collected value is 95), the corresponding standardized accuracy is 0.95; the completeness is 80% (if 100 data points should be collected, but 80 are actually collected), the standardized value is 0.8; and the real-time performance is 90% (if the data collection is scheduled to be completed within 1 minute, it is actually completed within 1.1 minutes), the standardized value is 0.9. Each feature is assigned a weight based on its relative importance in the data collection function. For example, the accuracy weight is 0.4, the completeness weight is 0.3, and the real-time performance weight is 0.3. The feature value of the data collection dimension is calculated by weighted average, using the formula: Data collection dimension feature value = Accuracy weight × Standardized accuracy + Completeness weight × Standardized completeness + Real-time performance weight × Standardized real-time performance.

[0106] Preferably, based on the characteristic requirements corresponding to the data acquisition indicators, the characteristic requirements corresponding to the data transmission indicators cover the stability and timeliness of data transmission. Stability refers to whether data is lost, erroneous, or interrupted during transmission; timeliness refers to whether data can be transmitted from the sending end to the receiving end within a specified time. In this first embodiment, the stability of data transmission is determined by monitoring the packet loss rate and error rate during transmission (e.g., a loss rate below 1%). The timeliness of data transmission is determined by calculating the transmission time from the sending end to the receiving end to ensure it meets the requirements of the metering automation system (e.g., a delay of no more than 5 seconds). Data on data transmission stability, such as the reciprocal of the packet loss rate (a lower packet loss rate indicates higher stability; assuming a packet loss rate of 5%, the stability value is 1 - 0.05 = 0.95), and timeliness data, such as the reciprocal of the transmission delay (assuming a specified delay of no more than 5 seconds, with an actual delay of 3 seconds, the timeliness value is 1 - (3 / 5) = 0.4), are collected and standardized. The weights for stability and timeliness are set to 0.6 and 0.4 respectively. The weighted average method is used to calculate the eigenvalue of the data transmission dimension: Eigenvalue of data transmission dimension = Stability weight × Standardized stability + Timeliness weight × Standardized timeliness. For example, if the standardized stability is 0.9 (corresponding to the standardized result of the original stability data being 0.9) and the timeliness is 0.6, then the eigenvalue of the data transmission dimension = 0.6 × 0.9 + 0.4 × 0.6 = 0.78.

[0107] In this first embodiment, the characteristic requirements corresponding to the data processing indicators include the accuracy and efficiency of data processing. Accuracy refers to whether the processed data meets the expected results; efficiency refers to the speed of data processing and resource consumption. Specifically, the accuracy of data processing is determined by comparing the degree of agreement between the processed data and the expected results, such as an error rate below 0.1%. Efficiency data is determined by recording the response time of data processing, such as processing 1000 data points in no more than 10 seconds, and the resource consumption rate, such as CPU utilization below 60%. Accuracy data (e.g., the reciprocal of the error rate, assuming an error rate of 0.1%, the accuracy value is 1 - 0.001 = 0.999) and efficiency data (e.g., the reciprocal of CPU utilization, assuming a CPU utilization rate not exceeding 60% and an actual utilization rate of 50%, the efficiency value is 1 - (50 / 60) = 0.167) are collected and standardized. Based on the actual situation, the weights for accuracy and efficiency are determined to be 0.7 and 0.3 respectively. The eigenvalues ​​of the data processing dimension are calculated using a weighted average method: Eigenvalue of data processing dimension = Accuracy weight × Standardized accuracy + Efficiency weight × Standardized efficiency. For example, if the standardized accuracy is 0.98 and the efficiency is 0.5, then the eigenvalue of the data processing dimension = 0.7 × 0.98 + 0.3 × 0.5 = 0.784.

[0108] In this first embodiment, the characteristic requirements corresponding to the data analysis indicators involve the accuracy, depth, and visualization effect of the data analysis. Accuracy refers to the reliability of the analysis results; depth refers to the breadth and depth of the analysis, and whether it can uncover valuable information; visualization effect refers to whether the presentation of the analysis results is intuitive and easy to understand. Specifically, the accuracy of the data analysis is determined by verifying the consistency between the analysis results and the actual operating conditions, such as a prediction accuracy rate higher than 90%. The depth of the data analysis is determined by evaluating whether the analysis results can reveal potential patterns in the data, such as discovering abnormal electricity consumption patterns through cluster analysis, or by user survey questionnaire scores, such as an intuitiveness score ≥ 4.5 / 5. After collecting the accuracy data (e.g., prediction accuracy rate, assuming a prediction accuracy rate of 90%), depth data (e.g., expert scoring, with a maximum score of 5 and a depth score of 4), and visualization effect data (e.g., user survey questionnaire scores, with a maximum score of 5 and a score of 4.5), standardization processing is performed. The weights for accuracy, depth, and visualization effect are set at 0.5, 0.3, and 0.2 respectively. The weighted average method is used to calculate the eigenvalues ​​of the data analysis dimensions: Eigenvalue = Accuracy weight × Standardized accuracy + Depth weight × Standardized depth + Visualization weight × Standardized visualization effect. For example, if the standardized accuracy is 0.9 (corresponding to a standardized result of the original accuracy data being 0.9), the depth is 0.8 (corresponding to a standardized result of the original depth data being 4), and the visualization effect is 0.9 (corresponding to a standardized result of the original visualization effect data being 4.5), then the eigenvalue of the data analysis dimension = 0.5 × 0.9 + 0.3 × 0.8 + 0.2 × 0.9 = 0.87.

[0109] In this first embodiment, the data storage metrics correspond to the characteristics of data storage capacity and query efficiency. Capacity refers to the amount of data the storage system can hold; query efficiency refers to the speed at which data is retrieved from the storage system. Specifically, the storage capacity is determined by testing the storage system's maximum storage capacity before reaching a performance bottleneck, such as maintaining stable operation after storing 1TB of data. The query efficiency is determined by recording the average response time of query operations, such as querying 10,000 records in no more than 2 seconds. After collecting data on storage capacity (e.g., performance after storing 1TB of data; assuming stable operation after storing 1TB, the capacity characteristic value is set to 1; if performance issues occur after storing 0.8TB, the capacity characteristic value is set to 0.8) and query efficiency data (e.g., an average response time of 1.5 seconds, specified to not exceed 2 seconds, the query efficiency characteristic value is set to 1 - (1.5 / 2) = 0.25), these two characteristic data are standardized. The weights for capacity and query efficiency are determined to be 0.5 and 0.5 respectively. The weighted average method is used to calculate the data storage dimension feature value: Data storage dimension feature value = capacity weight × standardized capacity + query efficiency weight × standardized query efficiency. For example, if the standardized capacity is 0.9 (corresponding to the standardized result of the original capacity data being 0.9) and the query efficiency is 0.6, then the data storage dimension feature value = 0.5 × 0.9 + 0.5 × 0.6 = 0.75.

[0110] In this first embodiment, consistent with the process of extracting feature data for functional dimensions, the performance feature values ​​corresponding to each performance evaluation indicator are obtained according to the feature requirements of the performance evaluation indicators.

[0111] Preferably, the characteristic requirement of the system operation accuracy index is the accuracy of the metering automation system's operating results, including the accuracy of data calculation, measurement, etc. Specifically, the current version of the metering automation system is verified using a standard test dataset, and the error range between the calculated result and the true value is determined. For example, for an electricity meter reading, if the true value is 100 and the calculated result is 99.9, the error is 0.1%. A large amount of such error data is collected and standardized. Standardized value = (original value - mean) / standard deviation. Assuming the mean error is 0.15% and the standard deviation is 0.05%, then the accuracy characteristic value = (0.1 - 0.15) / 0.05 = -1. This reflects the accuracy of the current version of the metering automation system in data calculation, measurement, etc. Preferably, principal component analysis (PCA) can be used to calculate the weights. After collecting, cleaning, and standardizing the data, a correlation coefficient matrix is ​​constructed, spectral decomposition is performed, eigenvalues ​​and principal components are calculated, and weights are determined. The weight of the accuracy feature is derived based on its importance in the system and its correlation with other features; for example, a weight of 0.25.

[0112] In this first embodiment, the system stability index is characterized by the ability of the metrology automation system to maintain a stable state during long-term operation, including the ability to handle and recover from abnormal situations. Specifically, a long-term operation of the metrology automation system (e.g., 72 hours) can be simulated, recording the number of anomalies and recovery times. For example, if the system experiences 3 anomalies within 72 hours, with an average recovery time of 1 minute, the number of anomalies and recovery times are standardized. Assume the standardized value for the number of anomalies is 0.2 (mean 5, standard deviation 2), and the standardized value for the recovery time is 0.8 (mean 2 minutes, standard deviation 0.5 minutes). When constructing the stability feature value, a weighted average method can be used to combine these two standardized values, with weights of 0.6 and 0.4 respectively, to calculate the stability feature value: 0.2 × 0.6 + 0.8 × 0.4 = 0.44. Similarly, principal component analysis is used to collect stability-related data, and through data preprocessing, constructing a correlation coefficient matrix, and spectral decomposition, the weights of the stability features are determined, such as 0.2.

[0113] In this first embodiment, the system reliability index is characterized by the probability that the metrology automation system can operate continuously and stably under various conditions, including the reliability of both hardware and software. Specifically, fault injection tests are performed, and the recovery success rate of the metrology automation system in the event of hardware failure or software error is statistically analyzed. For example, if 100 fault injection tests are performed and the system successfully recovers 99 times, the recovery success rate is 99%. The recovery success rate data is collected and standardized. Assuming the mean recovery success rate is 95% and the standard deviation is 2%, then the reliability characteristic value = (99-95) / 2 = 2. Principal component analysis is used to collect reliability data, and after a series of calculation steps, the weight of the reliability characteristic is determined, such as 0.15.

[0114] In this first embodiment, the characteristic requirement of the system response rate refers to the time interval from receiving a request to returning a result by the metering automation system, reflecting the response speed of the metering automation system. Specifically, the average response time of multiple requests is recorded using performance monitoring tools. For example, the specified response time is ≤200 milliseconds, and the actual average response time is 180 milliseconds. Response time data is collected and standardized. Assuming the mean response time is 220 milliseconds and the standard deviation is 30 milliseconds, then the characteristic value of this response time = (180-220) / 30 = -1.33. Preferably, principal component analysis can be used to collect response time data and determine its weight, such as 0.2.

[0115] In this first embodiment, the system throughput indicator refers to the amount of data or tasks that the metering automation system can process per unit time, reflecting the processing capacity of the metering automation system. Specifically, the number of transactions processed per second by the system is recorded under stress testing. For example, if TPS is specified as ≥1000, the actual TPS is 1200. Throughput data is collected and standardized. Assuming the mean throughput is 900 and the standard deviation is 150, then the characteristic value of throughput = (1200-900) / 150 = 2. Preferably, principal component analysis is used to collect throughput data and determine its weight, such as 0.2.

[0116] In this embodiment, by extracting the above-mentioned multi-dimensional features, the characteristics of the metering automation system under the current base version can be comprehensively and accurately characterized, providing a solid foundation for subsequent version difference analysis.

[0117] Step 302: Obtain the correlation coefficients between different feature values ​​to construct a correlation coefficient matrix corresponding to multiple feature values, and then determine the first weight corresponding to each of the functional feature values ​​and the second weight corresponding to each of the performance feature values ​​based on the correlation coefficient matrix and principal component analysis.

[0118] In this embodiment, weight analysis is performed on the features of each dimension to determine the importance and influence of different features in the metrology automation system. Specifically, after collecting multidimensional feature data containing functional and performance feature values, the multidimensional feature data can be cleaned. Specifically, a Z-score dynamic threshold detection method can be used to handle outliers (e.g., setting |Z-score|>3 as an outlier) to avoid the blind spots of fixed thresholds in consistency methods. For missing values, interpolation combined with expert feedback is used for imputation (e.g., predicting missing values ​​based on functional dimension correlations), rather than simple mean imputation.

[0119] Next, each feature value is standardized according to a preset standardization formula. The standardization formula is: Standardized value = (Original value - Mean) / Standard deviation. The mean and standard deviation are calculated using a rolling window, such as data from the past three months, to achieve dynamic updates and avoid weight bias caused by static standardization. Example: When standardizing the throughput feature value, a real-time window with a mean of 900 and a standard deviation of 150 is used for adaptive adjustment.

[0120] Next, the Pearson correlation coefficient between any two feature values ​​is obtained using the Pearson correlation coefficient formula, and a correlation coefficient matrix corresponding to the multidimensional feature data is constructed based on the Pearson correlation coefficient; wherein, the Pearson correlation coefficient formula is:

[0121]

[0122] Where m is the number of samples, xik Let i be the value of the i-th feature in the k-th sample. Let x be the mean of the i-th feature. j,k Let i be the value of the i-th feature in the k-th sample. Let r be the mean of the j-th feature. i,j The Pearson correlation coefficient represents the relationship between the i-th feature and the j-th feature.

[0123] Next, the Locality Sensitive Hash (LSH) algorithm is integrated to uncover nonlinear correlations between features (such as the hidden dependency between throughput and data transmission stability) and construct a more comprehensive correlation coefficient matrix. The elements in this matrix range from -1 to 1, but redundant dimensions are dynamically compressed by grouping similar features through hierarchical clustering (e.g., grouping response time and stability into the same group), rather than using a consistent matrix structure.

[0124] In this first embodiment, the eigenvalues ​​of the correlation coefficient matrix are solved using mathematical software or programming algorithms and arranged in ascending order. The cumulative contribution rate of the eigenvalues ​​is calculated, but the threshold is dynamically optimized: historical data is trained using a genetic algorithm to determine the optimal cumulative contribution rate (e.g., adaptively adjusted within the 80%–90% range), avoiding the risk of excessive dimensionality reduction in consistency methods. The cumulative contribution rate reflects the representativeness of the principal components to the original data.

[0125] In this first embodiment, assuming data on five features of the current version of the metering automation system are collected, a 5×5 correlation coefficient matrix Σ is constructed as follows:

[0126] Σ=[10.80.60.40.2 0.810.50.30.1 0.6 0.5 10.7 0.4 0.40.30.710.6 0.20.10.40.61]

[0131] Given the data for each feature, calculate the covariance between every two features to construct the covariance matrix Σ. The eigenvalues ​​of matrix Σ need to be solved, i.e., solving the equation |Σ-λI|=0, where I is the identity matrix. For the matrix Σ, mathematical software (such as MATLAB, Python's NumPy library, etc.) or programming algorithms can be used to solve for the eigenvalues. Taking Python's NumPy library as an example, the NumPy library function linalg.eigvals() can be used to solve for the eigenvalues ​​by inputting the correlation coefficient matrix Σ. Assume the obtained eigenvalues ​​are arranged in descending order (λ1, λ2, λ3, λ4, λ5). Finally, the cumulative contribution rate of the eigenvalues ​​is calculated to determine the number of principal components. Generally, a cumulative contribution rate ≥ 85% is required to determine the number of principal components to retain. For example, if eigenvalues ​​λ1≥λ2≥…≥λ5 are obtained, the cumulative contribution rate of the first k eigenvalues ​​is calculated. When the cumulative contribution rate reaches the requirement, the number of principal components is determined to be k.

[0132] Next, the variance contribution rate of each principal component is calculated. The variance contribution rate of each principal component is the ratio of the eigenvalue of that principal component to the sum of all eigenvalues. The weight of each feature is obtained by multiplying the squared loadings of each feature in each principal component by the corresponding variance contribution rate and then summing the results over all principal components.

[0133] Step 303: Determine multiple real-time feature values ​​corresponding to the current version of the metering automation system based on the first weight, the second weight, the functional feature value, and the performance feature value.

[0134] In this embodiment, the feature set is assumed to be F = {f1, f2, ..., f...} n The corresponding weight set is W = {w1, w2, ..., w}. n}, where w i Representing feature f i The weights satisfy Based on the extracted features and weight analysis results, a feature vector for the metering automation system under the current base version is constructed. The feature vector is a multi-dimensional vector, with each dimension corresponding to a feature and its weight. Assume the feature vector of base version V is V = (v1, v2, ..., v...). n ), where v i Representing feature f i If the value of is , then the eigenvector can be represented as:

[0135] V = (w1·v1, w2·v2, ..., w n ·v n )

[0136] In this embodiment, the weight analysis output is used for adaptive feature fusion similarity calculation, reducing human error. It also supports real-time optimization; that is, in online detection mode, the weights are dynamically updated based on cloud computing and big data, overcoming the static limitations of consistency methods.

[0137] Step 304: Call the preset multidimensional feature segmentation nonlinear mapping function to perform nonlinear mapping on each of the real-time feature values ​​to obtain multiple mapped feature values ​​corresponding to the current version.

[0138] In this embodiment, traditional similarity calculation methods are all based on the original feature vector, without considering the differences in the importance of features in different value ranges. This embodiment introduces a multi-dimensional feature piecewise nonlinear mapping function to preprocess the feature values ​​of each dimension before performing similarity calculation. This function applies different mapping rules according to the interval where the feature value is located, to highlight the feature differences in key intervals. Specifically, for feature values ​​in low-value intervals, since their impact on system performance is relatively small, the mapping function compresses and reduces their weight in similarity calculation, avoiding excessive impact from small fluctuations; for feature values ​​in mid-value intervals, their relative change relationship is maintained to accurately reflect differences in system performance; and for feature values ​​in high-value intervals, since small changes may have a significant impact on system performance, the mapping function amplifies and highlights their importance in similarity calculation. This approach allows feature values ​​in different value spaces to play a reasonable role in similarity calculation, improving the accuracy of version matching results and their practical application value. For the feature values ​​corresponding to each dimension in the feature vector, the multi-dimensional feature piecewise nonlinear mapping function is defined as:

[0139]

[0140] Where a1, b1, c1, β, γ, δ, α, and θ are mapping coefficients, and t1 and t2 are interval division thresholds; the x i This represents the i-th eigenvalue. The mapping coefficients mentioned above are obtained through optimization using a genetic algorithm based on historical data and expert experience. This function performs a non-linear transformation on the eigenvalues, highlighting the characteristic differences in key intervals and making the similarity calculation more realistic.

[0141] Step 305: Calculate the fusion features of the metering automation system under the current version based on the multiple mapping feature values, and calculate the similarity between the current version and the reference version based on the reference fusion features of the metering automation system under the reference version and the fusion features.

[0142] In this embodiment, after obtaining multiple mapping feature values, a feature correlation coefficient matrix corresponding to the multiple mapping feature values ​​is constructed.

[0143] In constructing the feature correlation coefficient matrix, the correlation coefficient between each mapped feature value and other mapped feature values ​​is calculated, constructing an n×n correlation coefficient matrix R, where n is the number of feature dimensions. Each element rij in the matrix represents the correlation coefficient between the i-th feature dimension and the j-th feature dimension, calculated using the following formula:

[0144]

[0145] Where m is the number of samples, x ik It is the value of the i-th feature dimension in the k-th sample. It is the sample mean of the i-th feature dimension.

[0146] The correlation coefficient matrix R is decomposed spectrally to obtain eigenvalues ​​λ1, λ2, ..., λn and corresponding eigenvectors v1, v2, ..., vn. The k largest eigenvalues ​​and their corresponding eigenvectors are selected based on their magnitude to form a dimensionality reduction matrix Vk. The eigenvector Vk represents a linear combination of the original features (i.e., the new feature space) and is used for subsequent difference analysis. Next, the fusion weight vector w = [w1, w2, ..., wn] is initialized, where wi represents the initial weight of the i-th feature dimension, which can be set to a uniform distribution, i.e., wi = 1 / n. The fusion weight vector is optimized using the Particle Swarm Optimization (PSO) algorithm. In the PSO algorithm, each particle represents a possible weight vector solution. The optimal weight vector is found by iteratively updating the particle's position and velocity. The update formula is as follows:

[0147]

[0148] in, Let represent the velocity of particle i in the d-th dimension at the t-th iteration. w is the inertia weight, controlling the degree to which the particle retains its past velocities. c1 and c2 are cognitive and social learning factors, respectively, controlling the particle's pursuit of the individual and global optima. r1 and r2 are random numbers in the range [0,1], introducing randomness to avoid local optima. p id It is the d-th dimension of the historical optimal position of particle i, g d It is the group's historical best position. This represents the value of particle i at the position in the d-th dimension during the t-th iteration. Through multiple iterations, the PSO algorithm finally obtains the optimal fusion weight vector w. opt Using the optimized fusion weight vector w opt The feature vectors are weighted and fused to obtain the fused feature vector X. fluse and Y fluseAssuming the feature vector of the base version to be detected is X = [x1, x2, ..., xn], and the feature vector of the standard base version is Y = [y1, y2, ..., yn], then the formula for calculating the fused feature vector is:

[0149]

[0150] Where w opt,i Let x be the weight of the i-th feature dimension after optimization. i and y i Let be the values ​​of the i-th feature dimension of the version to be detected and the standard version, respectively. Finally, calculate the fusion similarity S:

[0151]

[0152] Where · represents the dot product of vectors, and ||·|| represents the magnitude of the vector.

[0153] In this embodiment, adaptive feature fusion similarity calculation enables dynamic adjustment and optimization of feature weights. The fusion weight vector is automatically adjusted based on the correlation between features and historical data, making the similarity calculation more accurately reflect the actual similarity of the base versions. This method not only improves the accuracy of version matching but also enhances the adaptability to different operating environments and scenarios, ensuring the stability and reliability of the system under various conditions.

[0154] Step 306: Obtain version difference information between the reference version and the current version based on the similarity.

[0155] In this embodiment, based on the similarity, methods such as threshold judgment or cluster analysis can be used to achieve intelligent version matching and difference analysis. Specifically, when using threshold judgment for version analysis, a similarity threshold θ is set. If the similarity is higher than θ, the current version is considered to match the reference version and there is no difference; otherwise, a difference is considered to exist, and further analysis is required.

[0156] When using clustering algorithms (such as K-Means) for version analysis, the feature vectors of multiple versions are clustered, and the feature vector of the version to be detected is compared with the cluster center to determine its category.

[0157] For mismatched base versions, further analysis of the differences in their feature vectors provides a detailed difference report to help users quickly locate the problem. Difference analysis can be achieved by calculating the term-by-term differences in the feature vectors:

[0158] ΔV=V1-V2=(v 11 -v 21 ,v 12 -v 22 ,...,v1n -v 2n )

[0159] After matching analysis, due to feature vector fusion and dimensionality reduction, directly comparing the differences between each feature vector may not accurately reflect the differences in the original features. Therefore, it is necessary to map the fused feature vectors back to the original feature space based on the fused weight vector and the dimensionality reduction matrix to determine the specific difference dimensions.

[0160] First, based on the calculated fusion similarity, the difference dimension in the fused feature vector is determined. Assuming the fusion similarity threshold is θ, if the fusion similarity S is lower than θ, a difference is considered to exist. Then, the fused feature vector X is compared... fluse and Y fluse For each dimension, calculate the degree of difference. This can be done using either the absolute difference method or the relative difference method. The absolute difference method directly calculates the difference between two feature values; the relative difference method divides the difference by the feature value of the standard version to obtain the relative difference ratio.

[0161] Using the dimensionality reduction matrix Vk and the fusion weight vector w opt This involves mapping the difference dimensions in the fused feature vector back to the original feature space. By analyzing the eigenvectors and eigenvalues ​​in the dimensionality reduction matrix Vk, the original feature dimension corresponding to each fused dimension and its weight contribution are determined. This is then combined with the fused weight vector w. opt The contribution of each original feature dimension to the difference is calculated to determine the specific difference dimension.

[0162] In this first embodiment, when determining the difference dimensions by comparing the feature vectors of the current base version and the reference base version, the specific dimensions where there are differences in feature values ​​are identified. For example, if the value of the "data collection accuracy" feature of the version to be tested is 0.92, while that of the standard version is 0.98, then there is a difference in this dimension. For each dimension with a difference, the degree of difference is calculated. Either the absolute difference method or the relative difference method can be used. The absolute difference method directly calculates the difference between the two feature values; the relative difference method divides the difference by the feature value of the standard version to obtain the relative difference ratio. For example, if the absolute difference is 0.06, the relative difference is 6.12% (0.06 / 0.98). Next, the feature values ​​of each dimension are plotted on a radar chart to visually display the differences between the current version and the reference version in each dimension. Each dimension corresponds to an axis, and the feature values ​​are converted into coordinate points proportionally. By comparing the shapes of the radar charts of the two versions, the difference dimensions are quickly identified.

[0163] In this embodiment, a difference report including difference dimensions, difference degree, visualization charts, and preliminary diagnostic results can also be provided. The difference information is presented in intuitive tables and charts, facilitating quick understanding and processing by users. Through the detailed difference analysis process described above, users can clearly understand the differences between base versions and their causes, thereby taking targeted optimization measures to improve the overall performance and reliability of the metering automation system.

[0164] In this first embodiment, after performing version analysis on the reference version and the current version, the current version of the metering automation system can also be tested to achieve testing management of the current version.

[0165] Specifically, the version detection process for the current version of the automated metering system is divided into online detection and local detection. Local detection involves building a highly realistic automated metering system environment locally and using simulation to detect the base version. Automation, efficiency, and intelligence are achieved through automated detection, parallel detection, and intelligent detection technologies. The specific process is as follows: Operational data from the past six months, including metering data, operation logs, and fault records, is extracted from the system's historical database. Simulated metering data, equipment status data, and network traffic data are generated according to system design specifications and testing requirements to test the base version's performance under various expected operating conditions. Next, a simulation environment highly similar to the actual operating environment is built locally, including simulated metering equipment, communication networks, and servers. Historical operational data and simulated data are injected into the simulation environment to simulate the system's operating status under different conditions. Automated detection scripts and tools are used to perform functional and performance tests on the current version of the automated metering system. Functional tests include testing of data acquisition, data processing, and data analysis modules; performance tests include testing of performance indicators such as data transmission stability and data storage query efficiency. Multi-threading / multi-process technology is employed to simultaneously test multiple current versions of automated metering systems, improving testing efficiency. Finally, functional test reports and performance test reports are generated based on the testing results. The functional test report includes test results for each functional module, such as the accuracy, completeness, and real-time performance of data acquisition, and the accuracy and efficiency of data processing. The performance test report includes test results for various performance indicators, such as the stability and timeliness of data transmission, and the capacity and efficiency of data storage and querying. Based on the test reports, problems discovered during the testing process are listed, including functional defects and performance bottlenecks, and targeted optimization suggestions are proposed, such as algorithm improvements and resource optimization.

[0166] The online detection refers to real-time monitoring during the actual operation of the metering automation system. It employs distributed data acquisition technology, high-speed communication technology, cloud computing technology, and big data analytics to achieve real-time monitoring and evaluation of the base version. The specific process is as follows: First, real-time data, including metering data, equipment status data, and network traffic data, is directly collected from the actual operating metering automation system. Next, log data such as operation logs, fault logs, and performance logs generated during system operation are collected. Then, through distributed data acquisition nodes, system operation data and log data are collected in real time and transmitted to the detection center using high-speed communication technology. On the cloud computing platform, the collected data is processed and analyzed in real time. Big data analytics is used to evaluate system performance and compatibility. Machine learning algorithms are used to monitor the system's operating status in real time, promptly identifying anomalies such as data acquisition errors, data transmission interruptions, and performance bottlenecks. Based on the real-time detection results, system parameters and resource configurations are dynamically adjusted to ensure stable system operation. Finally, a real-time monitoring report is generated based on the system's real-time operating status, such as the real-time performance of data acquisition, the stability of data transmission, and system response time. A performance evaluation report is generated based on the real-time evaluation results of system performance indicators, such as throughput and reliability. When anomalies are detected based on the real-time monitoring report and the performance evaluation report, an alarm is immediately issued to prompt maintenance personnel to handle the situation promptly. Furthermore, through comprehensive analysis of historical and real-time data, system performance trends are predicted, providing a basis for system optimization.

[0167] In this first embodiment, combining the aforementioned local and online detection modes, a complete automated detection solution is disclosed, including detection objectives and requirements, selection of the detection solution, construction of the detection environment, design and optimization of the detection process, and processing and analysis of detection data. The following is a detailed description of the detection solution construction and implementation process:

[0168] Define testing objectives and requirements, such as functional integrity testing: verifying whether the base version meets all predetermined functional requirements, such as data acquisition, processing, analysis, and storage. Performance indicator testing: evaluating the performance of the base version, including but not limited to data transmission stability, data storage and query efficiency, and system response time.

[0169] Choose the testing mode and method, such as functional testing: based on the functional test report, select appropriate test cases to verify the functionality. Performance testing: based on the performance test report, use methods such as stress testing and load testing to evaluate performance.

[0170] Set up the testing environment, such as configuring the hardware and software environment according to the actual operating requirements of the base version. Prepare the data required for testing, including historical data and simulation data.

[0171] The implementation of an automated testing solution includes the design and optimization of the testing process. This involves registering and initializing the base version to be tested. Automated test scripts are run to simulate various user operations and system loads. System status is monitored in real time during testing, and testing parameters are adjusted based on feedback. After testing, log files, performance data, and other relevant information are collected and analyzed.

[0172] The testing data is processed and analyzed, such as cleaning and organizing the collected data to remove irrelevant information. The test results are compared with the expected results to identify functional defects and performance bottlenecks. By combining historical and current test data, the future performance trend of the system is predicted.

[0173] A testing report is generated, which includes a detailed overview of the testing, the execution process, the problems found, risk assessment, and improvement suggestions. The report is presented in a structured format to ensure clarity and readability.

[0174] In summary, the current version of the automated metrology system's testing process includes configuring the environment and setting parameters for the base version to be tested, ensuring it is in a testable state, executing automated test scripts and test cases according to the established testing plan, monitoring system behavior in real time and recording key data, analyzing the collected testing data to determine whether the base version meets the expected functional and performance standards, and providing feedback on any issues found during testing to the development team to facilitate continuous version optimization.

[0175] This embodiment discloses a version difference analysis method for a metrology automation system, effectively overcoming many defects and shortcomings of existing technologies and significantly improving the efficiency and accuracy of version analysis in metrology automation systems. Specifically, by extracting multi-dimensional features and performing weight analysis, a feature vector for the current version is constructed, enabling a comprehensive evaluation of the base version. Compared to traditional manual detection and simple version comparison, this method can quickly and accurately identify differences between different versions, reducing detection errors caused by human factors and improving detection efficiency and accuracy. Secondly, the similarity of the feature vectors is calculated, and then intelligent matching is achieved by combining threshold judgment or cluster analysis. This method can not only quickly determine whether different versions match, but also provide a detailed difference analysis report when they do not match, helping users quickly locate problems and thus achieving intelligent version management. Furthermore, this embodiment not only focuses on the functional completeness of different versions, but also conducts a comprehensive evaluation from a performance perspective (such as data transmission stability and data storage query efficiency). This multi-dimensional evaluation method can effectively discover potential performance bottlenecks, ensuring the overall performance and reliability of the system in actual operation. Finally, this embodiment provides a complete automated detection solution by combining local high-fidelity simulation detection and online real-time detection. The local testing mode comprehensively tests the base version through simulation, while the online testing mode utilizes distributed data acquisition and cloud computing technologies to achieve real-time monitoring and evaluation. This dual-mode testing scheme not only improves the flexibility of testing but also ensures the real-time nature and accuracy of the test results. In summary, the method provided in this embodiment effectively solves the problems of low efficiency, poor accuracy, and insufficient intelligence in existing technologies, significantly improving the overall performance and reliability of metering automation systems and meeting the requirements of modern power systems for efficient, intelligent management and high reliability.

[0176] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for analyzing version differences in a metrology automation system, characterized in that, include: The system collects multiple feature values ​​of the automated metering system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension; Construct a correlation coefficient matrix corresponding to multiple eigenvalues, and select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix; Based on the eigenvalues ​​of the principal component eigenvalues, the version weight corresponding to each eigenvalue is determined, so as to obtain multiple real-time eigenvalues ​​corresponding to the current version based on the version weights and the eigenvalues. Each real-time feature value is nonlinearly mapped according to a preset multidimensional feature segmentation nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version. Spectral decomposition is performed on the multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and the fusion weights corresponding to each feature vector are dynamically adjusted. Multiple reference feature vectors corresponding to the reference version are obtained, and the difference information between the current version and the reference version is determined based on the fusion weight, the reference feature vectors, and the feature vectors.

2. The method for version difference analysis of a metering automation system according to claim 1, characterized in that, The automated data acquisition and measurement system in its current version has several characteristic values, including: Based on several functional evaluation indicators corresponding to the functional dimensions, the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators are obtained; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators and data storage indicators. Based on several performance evaluation indicators corresponding to the performance dimension, the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators are obtained; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators, and system throughput indicators.

3. The method for version difference analysis of a metering automation system according to claim 2, characterized in that, The process of constructing a correlation coefficient matrix corresponding to multiple eigenvalues ​​and selecting multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix includes: Each of the functional feature values ​​and each of the performance feature values ​​are normalized to obtain multiple normalized functional feature values ​​and multiple normalized performance feature values. Based on the normalized functional eigenvalues ​​and the normalized performance eigenvalues, obtain the Pearson correlation coefficient between any two eigenvalues, and construct a correlation coefficient matrix corresponding to multiple eigenvalues ​​based on the Pearson correlation coefficients; The eigenvalues ​​corresponding to each eigenvalue are obtained according to the correlation coefficient matrix, and multiple principal component eigenvalues ​​are selected from the multiple eigenvalues ​​based on the eigenvalues ​​and the cumulative contribution rate of the multiple eigenvalues.

4. The method for version difference analysis of a metering automation system according to claim 3, characterized in that, The step of determining the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component features, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values, includes: Obtain the sum of the eigenvalues ​​of the principal components corresponding to the eigenvalues ​​of the principal components; For any one of the principal component eigenvalues: The ratio of the eigenvalue corresponding to the principal component eigenvalue to the sum of the eigenvalues ​​is obtained, and the ratio is used as the variance contribution rate of the principal component eigenvalue. For any one of the aforementioned eigenvalues: Obtain the squared loading of the eigenvalue in each of the principal component eigenvalues, and multiply the squared loading by the variance contribution rate corresponding to each of the principal component eigenvalues ​​to obtain the product of multiple load sharing rates corresponding to the eigenvalue. The version weight corresponding to the feature value is obtained by summing the products of multiple load sharing rates. Each of the aforementioned feature values ​​is multiplied by its corresponding version weight to obtain multiple real-time feature values ​​corresponding to the current version.

5. The method for version difference analysis of a metering automation system according to claim 4, characterized in that, The step of performing nonlinear mapping on each real-time feature value according to a preset multidimensional feature piecewise nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version includes: Based on the multiple preset feature value intervals in the multidimensional feature segmentation nonlinear mapping function, the multiple real-time feature values ​​are divided into a first interval feature value, a second interval feature value, and a third interval feature value. The first nonlinear transformation function corresponding to the first interval feature value, the second nonlinear transformation function corresponding to the second interval feature value, and the third nonlinear transformation function corresponding to the third interval feature value are retrieved from the multidimensional feature segmentation nonlinear mapping function, so as to perform nonlinear mapping on the first interval feature value, the second interval feature value, and the third interval feature value respectively, to obtain multiple mapped feature values.

6. The method for version difference analysis of a metering automation system according to claim 5, characterized in that, The step of performing spectral decomposition on multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjusting the fusion weights corresponding to each feature vector, includes: Obtain the correlation coefficients between each pair of the multiple mapping feature values ​​to construct a mapping correlation coefficient matrix corresponding to the multiple mapping feature values; The mapping correlation coefficient matrix is ​​subjected to spectral decomposition to obtain multiple first spectral decomposition eigenvalues ​​and a first eigenvector corresponding to each first spectral decomposition eigenvalue; Based on the magnitude of the first spectral decomposition eigenvalue, several second spectral decomposition eigenvalues ​​and a second eigenvector corresponding to each second spectral decomposition eigenvalue are selected from the multiple first spectral decomposition eigenvalues ​​according to the selection rule from large to small. Multiple preset initial fusion weights are obtained, and the preset initial fusion weights are dynamically adjusted according to a preset particle swarm optimization algorithm to obtain multiple optimized fusion weights.

7. The method for version difference analysis of a metering automation system according to claim 6, characterized in that, The step of obtaining multiple reference feature vectors corresponding to the reference version, and determining the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors, and the feature vectors, includes: Based on the optimized fusion weights, the multiple second feature vectors and the multiple reference feature vectors are weighted and fused to obtain the fusion feature value corresponding to the current version and the reference fusion feature value corresponding to the reference version. Obtain the fusion similarity between the fusion feature value and the reference fusion feature value; When the fusion similarity is greater than or equal to a preset similarity threshold, it is determined that there is no difference between the current version and the reference version.

8. A method for analyzing version differences in a metrology automation system according to any one of claims 6-7, characterized in that, The step of obtaining multiple reference feature vectors corresponding to the reference version, and determining the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors, and the feature vectors, includes: When the fusion similarity is less than the similarity threshold, it is determined that there is a difference between the current version and the reference version, and the difference value between the corresponding second feature vector and the reference feature vector is obtained, so as to extract the difference feature vector from multiple second feature vectors according to the difference value; The difference feature vector is mapped back to the original feature space based on the second spectral decomposition feature value, the original feature value corresponding to the difference feature vector is determined, and then the difference dimension is determined based on the original feature value.

9. A version difference analysis system for a metrology automation system, characterized in that, It includes a multidimensional data acquisition module, a principal component analysis module, a weight adjustment module, a nonlinear mapping module, a feature fusion module, and a difference analysis module; The multidimensional data acquisition module is used to acquire multiple feature values ​​of the metering automation system in the current version; wherein, the multiple feature values ​​include multiple functional feature values ​​corresponding to the functional dimension and multiple performance feature values ​​corresponding to the performance dimension; The principal component analysis module is used to construct a correlation coefficient matrix corresponding to multiple eigenvalues, and to select multiple principal component eigenvalues ​​from the multiple eigenvalues ​​based on the eigenvalues ​​of the correlation coefficient matrix. The weight adjustment module is used to determine the version weight corresponding to each feature value based on the eigenvalues ​​of the principal component feature values, so as to obtain multiple real-time feature values ​​corresponding to the current version based on the version weights and the feature values; The nonlinear mapping module is used to perform nonlinear mapping on each of the real-time feature values ​​according to a preset multidimensional feature piecewise nonlinear mapping function to obtain multiple mapped feature values ​​corresponding to the current version. The feature fusion module is used to perform spectral decomposition on multiple mapped feature values ​​to obtain multiple feature vectors corresponding to the current version, and dynamically adjust the fusion weights corresponding to each feature vector. The difference analysis module is used to obtain multiple reference feature vectors corresponding to the reference version, so as to determine the difference information between the current version and the reference version based on the fusion weight, the reference feature vectors and the feature vectors.

10. A version difference analysis system for a metrology automation system according to claim 9, characterized in that, The multidimensional data acquisition module includes a functional indicator unit and a performance indicator unit; The functional indicator unit is used to obtain the functional characteristic values ​​of the metering automation system under the current version and each of the functional evaluation indicators based on several functional evaluation indicators corresponding to the functional dimension; wherein, the functional evaluation indicators include data acquisition indicators, data transmission indicators, data processing indicators, data analysis indicators and data storage indicators; The performance indicator unit is used to obtain the performance characteristic values ​​of the metering automation system under the current version and each of the performance evaluation indicators based on several performance evaluation indicators corresponding to the performance dimension; wherein, the performance evaluation indicators include system operating accuracy indicators, system stability indicators, system reliability indicators, system response rate indicators and system throughput indicators.

Images