Transformer control method and system based on artificial intelligence

Abstract

The invention provides a transformer control method and system based on artificial intelligence, and relates to the technical field of transformer control. The method comprises the steps that historical control data are collected, clustering analysis based on control correlation is carried out, and historical control correlation clustering data are formed; performing control prediction analysis on the historical control correlation clustering data, and establishing control prediction model data; according to the method, real-time monitoring data is acquired, control prediction analysis is carried out in combination with control prediction model data, real-time control prediction data is formed, effective and accurate data prediction is realized through reasonable data processing analysis, and a more intelligent transformer control mode is provided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of transformer control, in particular to a transformer control method and system based on artificial intelligence. BACKGROUND

[0002] Traditional transformer control is mainly reactive, such as automatic voltage regulation and on-load voltage regulation. It is based on real-time measurement of voltage, current and other parameters, compared with the set value, and then driven by the classic control law such as PID to drive the actuator. This control is lagging and difficult to deal with complex and nonlinear system dynamics.

[0003] Currently, there is also a big data-based data processing to realize automatic control of transformer control, but this way is mainly based on the characteristics of big data to realize, and does not have the predictability of new data, limiting the range and ability of the accuracy of automatic control.

[0004] Therefore, it is an urgent problem to design a transformer control method and system based on artificial intelligence, to realize effective and accurate data prediction through reasonable data processing and analysis, and to provide a more intelligent transformer control method. SUMMARY

[0005] The purpose of the present application is to provide a transformer control method based on artificial intelligence, which realizes reasonable data clustering by analyzing the correlation between monitoring data and corresponding control data under different control items based on historical control data, forms a clustering basic big data with more control correlation in the historical control data of different control items, and then realizes reasonable control prediction analysis by using the clustering basic big data, establishes a corresponding control prediction model, so as to realize control parameter prediction based on real-time monitoring data, and provides more accurate and reasonable control prediction information. On the one hand, the historical control data is reasonably processed by control correlation clustering, which can effectively ensure the pertinence of the subsequent model establishment. Through the training and learning of these targeted clustering basic big data, the control prediction of different control conditions can be more accurately and effectively realized. On the other hand, the control prediction model established by the historical control clustering data can realize effective and accurate data prediction based on monitoring data, which is more adaptive and accurate than the big data processing method.

[0006] The application also aims to provide an artificial intelligence-based transformer control system, which collects basic big data and real-time monitoring data through a clustering collection unit, further cleans and arranges the basic big data in combination with a clustering analysis unit to form more effective and reasonable clustered data for model training, and establishes an artificial intelligence control parameter prediction model by using different clustered data as basic training data by means of a training learning unit, so that the real-time prediction unit completes the analysis and prediction of control parameters on the basis of sufficient collection of real-time monitoring data, different functional units are connected to each other to form a close whole for realizing transformer control prediction analysis, and the system is an important material basis for completing artificial intelligence control of transformers.

[0007] In the first aspect, the application provides an artificial intelligence-based transformer control method, which comprises the following steps: collecting historical control data, performing clustering analysis based on control correlation to form historical control correlation clustered data, performing control prediction analysis on the historical control correlation clustered data to establish control prediction model data, acquiring real-time monitoring data, and performing control prediction analysis in combination with the control prediction model data to form real-time control prediction data.

[0008] In the application, the method realizes reasonable data clustering by performing correlation analysis on historical control data under different control sub-items based on monitoring data and corresponding control data, forms clustered basic big data with better control correlation in historical control data of different control items, and then realizes reasonable control prediction analysis by using the clustered basic big data to establish a corresponding control prediction model, so as to realize control parameter prediction based on real-time monitoring data and provide more accurate and reasonable control prediction information. On the one hand, the historical control data is reasonably processed in terms of control correlation clustering, which can effectively ensure the pertinence of subsequent model establishment. Control prediction of different control conditions can be more accurately and effectively realized by training and learning of these pertinence clustered basic big data. On the other hand, the control prediction model established by means of historical control clustered data can realize effective and accurate data prediction based on monitoring data, which is more adaptive and accurate than the big data processing-based mode.

[0009] As a possible implementation manner, the method of collecting historical control data, performing clustering analysis based on control correlation, and forming historical control correlation clustered data comprises the following steps: extracting data of the historical control data for different control items to form corresponding historical sub-item control data of different control items; classifying the different historical sub-item control data for monitoring control to form corresponding historical sub-item control classification data; and performing correlation clustering analysis on the different historical sub-item control data to form historical control correlation clustered data.

[0010] In this invention, before performing reasonable control correlation clustering analysis on historical control data, it is necessary to first extract and parameterize the historical data appropriately. Here, data extraction is performed on a per-sub-item basis; that is, the period from a specific time point after the last control operation to a specific time point after the next control operation is considered as a single control sub-item for data extraction. After obtaining the sub-item control data, each sub-item needs to be reasonably parameterized. Data parameterization is a crucial foundation for subsequent data correlation analysis and predictive model establishment; only after data parameterization can correlation clustering analysis be performed.

[0011] As one possible approach, historical control data is extracted for different control items to form historical sub-item control data corresponding to each control item. This includes: extracting different types of monitoring feature parameters from different historical sub-item control data and establishing corresponding historical sub-item monitoring parameter matrices. m represents the number of different historical sub-item control data, and n represents the number of different types of monitoring characteristic parameters; for different historical sub-item control data, different types of control characteristic parameters are extracted, and corresponding historical sub-item control parameter matrices are established. k represents the number of different types of control characteristic parameters; for different historical sub-term control data, the corresponding historical sub-term monitoring parameter matrix is ​​used. and the corresponding historical sub-term control parameter matrix These are combined to form the corresponding historical sub-item control data.

[0012] In this invention, the data extraction from historical control data mainly involves the reasonable extraction and parameterization of different data information from the secondary control data. It should be noted that the extracted data includes two main categories: one is parameter data collected for monitoring to provide basic input for control analysis; the other is different types of control parameter data formed after combining the analysis of monitoring data. These two types of data also differ in their extraction methods. For monitoring parameter data, since the monitoring data obtained from the secondary control data contains time parameters, reasonable feature information extraction is required to complete subsequent correlation analysis and establish a predictive model. The feature information referred to here mainly eliminates the influence of time parameters. For example, for information such as voltage, current, oil temperature, and ambient temperature that change over time, feature information such as mean, average rate of change, variance, and root mean square error can be used to express the characteristics of the corresponding parameter type. Of course, multiple feature information can be selected simultaneously to express the same parameter; the specific feature information type selected can be determined according to the actual situation. Feature extraction of control parameter data mainly involves the change in the parameter relative to the current value during control processing, such as the increase in temperature, the change in current, and the change in voltage. This change can be a relative value or a relative rate of change. Multiple feature expression parameters can be selected simultaneously for the same control parameter, determined based on the specific situation. It's also important to note that for matrix data of the same type extracted from different sub-items, the elements in the matrix must correspond to each other to avoid mismatches affecting the rationality and accuracy of subsequent analysis. After extraction, these two types of data are combined to form the control parameter matrix, thus forming all the necessary control data for the sub-item control parameter.

[0013] As one possible implementation, different historical sub-item control data are categorized for monitoring and control purposes to form corresponding historical sub-item control classification data. This includes: determining the maximum total number of different monitoring feature parameters to be extracted and the maximum total number of different control feature parameters to be extracted, and setting the maximum total as the unified total for classification; for different historical sub-item control data, based on the unified total for classification, the corresponding historical sub-item monitoring parameter matrix is... The parameters are categorized and standardized to form a unified matrix of historical sub-item monitoring parameters. For different control characteristic parameters, the corresponding historical sub-item control parameter matrix is ​​then processed according to the unified total number of categories. The parameters are classified and processed uniformly to form a unified matrix of corresponding historical sub-item control parameters.

[0014] In this invention, the data classification of historical sub-item control data mainly considers two aspects. First, the subsequent prediction model aims to predict accurate and reasonable control characteristic information based on the acquired monitoring characteristic data. Therefore, it is necessary to establish a mapping relationship between the monitoring parameter matrix extracted from the historical sub-item data and the corresponding control parameter matrix. Second, when establishing the mapping relationship between the monitoring parameter matrix and the control parameter matrix, the rules of matrix mapping must be considered, namely, the element positions and total number of elements in the two matrices. Therefore, when establishing the mapping relationship, i.e., during the process of unifying the total number of elements in both the monitoring parameter matrix and the control parameter matrix, it is also necessary to consider the supplementary positions of the matrix with fewer elements during element supplementation, so that the two matrices formed have uniformity in element position and element structure arrangement. Of course, for the element supplementation of the matrix with fewer elements, the parameter value of the supplemented element can be a unit parameter value or other constant, as long as it does not affect the subsequent mapping relationship analysis.

[0015] As one possible approach, correlation clustering analysis is performed on different historical sub-item control data to form historical control correlation clustering data. This includes: for different historical sub-item control data, based on the corresponding historical sub-item monitoring parameter matrix... and the corresponding historical sub-term control parameter matrix Perform matrix merging to form the historical sub-item merged parameter matrix corresponding to the historical sub-item control data. Where i represents the index of a different parameter in the historical sub-item merged parameter matrix, and the total number of parameters in the historical sub-item merged parameter matrix is ​​equal to the number of parameters in the historical sub-item monitoring parameter matrix. Total number of parameters and historical sub-term control parameter matrix The sum of the total number of parameters; merging parameter matrices for different historical terms. Perform non-repetitive analysis on the relevant clusters to generate historical control correlation cluster data.

[0016] In this invention, the essence of correlation clustering analysis is to cluster historical control data of different sub-items under the same operating conditions, thereby providing a more reasonable and effective foundation of big data for establishing prediction models for corresponding operating conditions. Considering that although the control sub-items differ under different operating conditions, the synergistic changes between monitoring parameters, between control parameters, and between monitoring and control parameters are closely correlated, merging the monitoring parameter matrix and the control parameter matrix before conducting correlation analysis can fully ensure the grasp of the correlation magnitude of these three synergistic relationships. Simply performing correlation analysis between monitoring parameter matrices or between control parameter matrices alone cannot fully reflect the correlation of these three synergistic relationships. Of course, for data of different sub-items, the elements in different merged parameter matrices correspond to each other in terms of parameter type to ensure the rationality and effectiveness of the analysis. For correlation analysis, utilizing the correlation characteristics of the matrices and extracting representative matrices for clustering around representative matrices is more reasonable and efficient.

[0017] As one possible implementation, the parameter matrix is ​​merged for different historical sub-terms. Perform non-repetitive analysis on the relevant clusters to generate historical control correlation clustering data, including: merging parameter matrices for different historical sub-terms. Extract any one and label it as the starting matrix for analysis; perform cluster non-repetition analysis based on the starting matrix in the following manner: determine the merging parameter matrix of the starting matrix and other historical terms. The cosine similarity value is obtained by merging the historical terms corresponding to the smallest cosine similarity value into the parameter matrix. Extract and determine the representative matrix; for the initial analysis matrix and the representative matrix, determine the parameter matrix to be merged with the remaining historical terms. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the closely related threshold. This forms the corresponding initial clustering matrix set. For different initial clustering matrix sets, the following judgment is made: if different initial clustering matrix sets contain the same historical term merging parameter matrix... Then, the initial analysis matrix is ​​determined as the calibration representative matrix, and the initial clustering matrix set corresponding to the initial analysis matrix is ​​determined as the historical control correlation clustering set corresponding to the calibration representative matrix; if there are no identical historical second-order merging parameter matrices in different initial clustering matrix sets... Then the selected representative matrix extracted this time will be merged with the remaining historical sub-items into a parameter matrix. The cosine similarity value is the smallest and is different from the extracted historical term merging parameter matrix. Historical term merged parameter matrix The newly determined representative matrix is ​​then used to determine the parameter matrix for merging with the remaining historical terms. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the closely related threshold. This forms the corresponding initial clustering matrix set; the evaluation continues for different initial clustering matrix sets until any two initial clustering matrix sets contain the same historical term merging parameter matrix. The initial analysis matrix and all selected representative matrices are then determined as calibration representative matrices, and the initial clustering matrix corresponding to the calibration representative matrix is ​​calibrated as the historical control correlation cluster set. Different calibration representative matrices and their corresponding historical control correlation cluster sets are combined to form historical control correlation cluster data.

[0018] In this invention, the formation of control correlation clustering data is based on the magnitude of correlation for classification and clustering. It is understood that the correlation between merged parameter matrices under different operating conditions is relatively weak. This correlation is represented in this application by the magnitude of the cosine similarity between matrices. The smaller the cosine similarity between two matrices, the weaker their correlation. Therefore, during analysis, an arbitrary matrix is ​​first extracted as the starting matrix for analysis, and the cosine similarity between the starting analysis matrix and other matrices is determined. While it's not possible to directly determine whether the data from two matrices belong to the same operating condition based solely on the cosine similarity, identifying the matrix with the smallest cosine similarity can help determine the control data least likely to belong to the same operating condition. Thus, by repeatedly using the extracted matrix data least likely to belong to the same operating condition as a reference, the cosine similarity between this matrix and the remaining matrices can be continuously obtained to determine other parameter matrices least likely to belong to the same operating condition. For clustering based on the representative matrix, considering data validity and to avoid clustering potentially different working conditions together, the chosen correlation threshold can be strictly limited, such as by a large cosine similarity value. Of course, this type of clustering based on the correlation threshold requires evaluation of the newly extracted representative matrix each time it is extracted. It should be noted that with progressively extracted representative matrices, as the number of remaining matrices gradually decreases, it is highly likely that the extracted matrix data will correspond to data from previous representative matrices under the same working conditions. Therefore, it is necessary to limit the number of extractions of the representative matrix, based on whether there are intersecting matrix data within the clustered data of the representative matrix. This effectively avoids generating different clustered data corresponding to different representative matrices due to more detailed feature divisions for the same working condition.

[0019] As one possible approach, control prediction analysis is performed on historical control correlation clustering data to establish control prediction model data. This includes: for different historical control correlation clusters in the historical control correlation clustering data, a unified matrix is ​​established based on the different historical sub-term monitoring parameters within the historical control correlation clusters. And the corresponding historical secondary control parameter unified matrix A mapping matrix group is formed, where u represents the number of different historical control correlation clusters, and x represents the unified matrix of historical sub-item monitoring parameters. The unified parameter number for different monitoring characteristics, where y represents the unified matrix of historical sub-item control parameters. The unified parameter numbering for different control characteristics, and the unified matrix of historical sub-term control parameters. The total number and arrangement of elements in the matrix are consistent with the historical sub-item monitoring parameters. Same; establish clustering mapping relationship C u *F u =D u Based on the different mapping matrices in the historical control correlation clusters, the clustering mapping relationships are analyzed to determine the prediction relationship matrix F. u , where C u D represents the unified matrix of monitoring parameters corresponding to the historical control correlation cluster set numbered u. u C represents the unified control parameter matrix corresponding to the historical control correlation cluster set numbered u. u *F u C represents u The product of Fu; the historical secondary monitoring parameter matrix and cluster mapping relationship under the calibration representative matrix corresponding to different historical control correlation cluster sets are used to form control prediction model data.

[0020] In this invention, the predictive analysis mainly involves establishing the correlation between the unified monitoring parameter matrix and the unified control parameter matrix under the corresponding cluster set. The goal is to obtain the constant matrix that maps from the unified monitoring parameter matrix to the unified control layer matrix. Here, one approach is to select a number of mapping matrix groups that correspond to the number of unknowns in the constant matrix, or to extract multiple mapping matrix groups with a number of unknowns corresponding to the number of unknowns in the constant matrix, solve them, and then average them to form the final constant matrix. Any approach that establishes a reasonable mapping relationship is acceptable.

[0021] One possible approach is to acquire real-time monitoring data and combine it with control prediction model data to perform control prediction analysis, thereby forming real-time control prediction data. This includes: extracting the corresponding real-time monitoring parameter matrix based on the real-time monitoring data, and performing the following analysis on the real-time monitoring parameter matrix and the corresponding historical sub-term monitoring parameter matrices under different calibration representative matrices: determining the cosine similarity value between the real-time monitoring parameter matrix and the corresponding historical sub-term monitoring parameter matrices under different calibration representative matrices, and labeling the clustering mapping relationship corresponding to the largest cosine similarity value as the real-time analysis and prediction model; and performing control prediction analysis based on the real-time analysis and prediction model and the real-time monitoring parameter matrix to form real-time control prediction data.

[0022] In this invention, the control prediction analysis based on real-time monitoring data using control prediction model data first requires confirming the operating condition to which the real-time monitoring data belongs. The confirmation method is to extract the real-time monitoring data to form a real-time monitoring parameter matrix similar to the historical sub-term monitoring parameter matrix. Then, based on the different calibration representation matrices corresponding to the historical sub-term monitoring parameter matrices, the magnitude of the cosine similarity value is defined. It can be understood that the larger the cosine similarity value, the closer the real-time monitoring parameter matrix is ​​to the corresponding operating condition. Therefore, the operating condition with the largest cosine similarity value is determined as the operating condition of the real-time monitoring situation, and the corresponding mapping relationship is used as the prediction model for real-time analysis to complete the control prediction analysis.

[0023] As one possible implementation, control prediction analysis is performed based on the real-time analysis and prediction model and the real-time monitoring parameter matrix to form real-time control prediction data. This includes: classifying and unifying the real-time monitoring parameter matrix to form a corresponding unified real-time monitoring parameter matrix; determining the unified real-time control parameter matrix based on the unified real-time monitoring parameter matrix and the real-time analysis and prediction model; and determining all control prediction parameters based on the unified real-time control parameter matrix.

[0024] In this invention, the predictive analysis first requires obtaining a unified control monitoring parameter matrix based on the predictive analysis model. Before the analysis, the real-time monitoring parameter matrix needs to be classified and unified to provide a unified data type for the predictive model, thereby obtaining a unified control parameter matrix. The extraction of different types of control parameters in the unified control parameter matrix can be achieved in the reverse direction of the classification and unification of the historical sub-item control parameter matrix.

[0025] Secondly, the present invention provides an artificial intelligence-based transformer control system, comprising: a data acquisition unit for acquiring historical control data and real-time monitoring data; a clustering analysis unit for performing clustering analysis based on control correlation on the historical control data acquired by the data acquisition unit to form historical control correlation clustering data; a training and learning unit for performing control prediction analysis on the historical control correlation clustering data formed by the clustering analysis unit to establish control prediction model data; and a real-time prediction unit for performing control prediction analysis based on the real-time monitoring data acquired by the data acquisition unit and the control prediction model data established by the training and learning unit to form real-time control prediction data.

[0026] In this invention, the system collects basic big data and real-time monitoring data through a clustering acquisition unit, and further cleans and organizes the basic big data through a clustering analysis unit to form more effective and reasonable clustering data for model training. With the help of a training and learning unit, different clustering data are used as the basis for training to establish an artificial intelligence control parameter prediction model. This model is then handed over to a real-time prediction unit to complete the analysis and prediction of control parameters based on the full collection of real-time monitoring data. The different functional units are interconnected to form a tight whole for realizing transformer control prediction and analysis, which is an important material basis for realizing artificial intelligence control of transformers.

[0027] The beneficial effects of the transformer control method and system based on artificial intelligence provided by this invention are as follows:

[0028] This method achieves reasonable data clustering by performing correlation analysis on historical control data under different control sub-items based on monitoring data and corresponding control data. This forms a more control-correlated basic big data set of historical control data for different control sub-items. Then, it uses the basic big data set of clusters to achieve reasonable control prediction analysis and establish a corresponding control prediction model. This enables control parameter prediction based on real-time monitoring data, providing more accurate and reasonable control prediction information. On the one hand, reasonable control correlation clustering of historical control data can effectively ensure the relevance of subsequent model building. Training on these targeted basic big data sets of clusters can more accurately and effectively achieve control prediction for different control conditions. On the other hand, the control prediction model established with the help of historical control cluster data can achieve effective and accurate data prediction based on monitoring data, which is more adaptable and accurate than the big data processing method.

[0029] The system collects basic big data and real-time monitoring data through a clustering acquisition unit, and further cleans and organizes the basic big data through a clustering analysis unit to form more effective and reasonable clustering data for model training. With the help of the training and learning unit, different clustering data are used as the basis for training to build an artificial intelligence control parameter prediction model. This model is then handed over to the real-time prediction unit to complete the analysis and prediction of control parameters based on the full collection of real-time monitoring data. The different functional units are interconnected to form a tight whole for realizing transformer control prediction and analysis, which is an important material basis for realizing artificial intelligence control of transformers. Attached Figure Description

[0030] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments of the present invention will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0031] Fig. 1 A flowchart illustrating the steps of an artificial intelligence-based transformer control method provided in an embodiment of the present invention;

[0032] Fig. 2 This is a schematic diagram of a transformer control system based on artificial intelligence, provided as an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0034] Traditional transformer control is primarily reactive, such as automatic voltage regulation and on-load tap changer. It relies on real-time measurements of parameters like voltage and current, comparing them with setpoints, and then using classic control laws like PID controllers to drive the actuators. This type of control is lagging and struggles to handle complex, nonlinear system dynamics.

[0035] Currently, there are also methods that use big data to process data for transformer control to achieve automated control. However, this method mainly relies on the characteristic information of big data and lacks predictive capabilities, especially for new data, which limits the scope and capability of the accuracy of automatic control.

[0036] refer to Figs. 1-2This invention provides an artificial intelligence-based transformer control method. This method achieves reasonable data clustering by performing correlation analysis on historical control data under different control sub-items based on monitoring data and corresponding control data. This forms a larger cluster of historical control data with higher control relevance, and then utilizes this clustered data to achieve reasonable control prediction analysis, establishing a corresponding control prediction model. This enables control parameter prediction based on real-time monitoring data, providing more accurate and reasonable control prediction information. On the one hand, reasonable control relevance clustering of historical control data effectively ensures the relevance of subsequent model building. Training on this targeted clustered data allows for more accurate and effective control prediction for different control conditions. On the other hand, the control prediction model established using historical control clustering data can achieve effective and accurate data prediction based on monitoring data, which is more adaptable and accurate compared to methods based on big data processing.

[0037] An artificial intelligence-based transformer control method specifically includes the following steps:

[0038] S1: Collect historical control data, perform cluster analysis based on control correlation, and form historical control correlation cluster data.

[0039] Historical control data is collected and cluster analysis based on control correlation is performed to form historical control correlation cluster data. This includes: extracting data from historical control data for different control items to form historical sub-item control data corresponding to different control items; classifying the different historical sub-item control data for monitoring and control to form corresponding historical sub-item control classification data; and performing correlation cluster analysis on the different historical sub-item control data to form historical control correlation cluster data.

[0040] Before conducting reasonable control correlation clustering analysis on historical control data, it is necessary to first extract and parameterize the historical data appropriately. Here, data extraction is performed on a per-sub-item basis; that is, the period from a specific time point after the last control was implemented to a specific time point after the next control is implemented is considered as a control sub-item for data extraction. After obtaining the sub-item control data, each sub-item data needs to be reasonably parameterized. Data parameterization is a crucial foundation for subsequent data correlation analysis and predictive model building; only after data parameterization can correlation clustering analysis be performed.

[0041] Historical control data is extracted for different control items to form historical sub-item control data corresponding to each control item. This includes: extracting different types of monitoring feature parameters from different historical sub-item control data and establishing corresponding historical sub-item monitoring parameter matrices. m represents the number of different historical sub-item control data, and n represents the number of different types of monitoring characteristic parameters; for different historical sub-item control data, different types of control characteristic parameters are extracted, and corresponding historical sub-item control parameter matrices are established. k represents the number of different types of control characteristic parameters; for different historical sub-term control data, the corresponding historical sub-term monitoring parameter matrix is ​​used. and the corresponding historical sub-term control parameter matrix These are combined to form the corresponding historical sub-item control data.

[0042] Data extraction from historical control data primarily involves the reasonable extraction and parameterization of different data information from the secondary control data. It's important to note that the extracted data falls into two main categories: one is parameter data collected for monitoring to provide basic input for control analysis; the other is different types of control parameter data formed after combining monitoring data with analysis. These two types of data also differ in their extraction methods. For monitoring parameter data, since the monitoring data obtained from the secondary control data contains time parameters, reasonable feature information extraction is necessary to complete subsequent correlation analysis and predictive model building. The feature information referred to here mainly eliminates the influence of time parameters. For example, for information such as voltage, current, oil temperature, and ambient temperature that change over time, features such as mean, average rate of change, variance, and root mean square error can be used to represent the corresponding parameter type. Of course, multiple feature information types can be selected simultaneously for the same parameter; the specific feature information type selected can be determined based on the actual situation. Feature extraction of control parameter data mainly involves the change in the parameter relative to the current value during control processing, such as the increase in temperature, the change in current, and the change in voltage. This change can be a relative value or a relative rate of change. Multiple feature expression parameters can be selected simultaneously for the same control parameter, determined based on the specific situation. It's also important to note that for matrix data of the same type extracted from different sub-items, the elements in the matrix must correspond to each other to avoid mismatches affecting the rationality and accuracy of subsequent analysis. After extraction, these two types of data are combined to form the control parameter matrix, thus forming all the necessary control data for the sub-item control parameter.

[0043] Different historical sub-item control data are categorized for monitoring and control purposes, forming corresponding historical sub-item control classification data. This includes: determining the maximum total number of different monitoring feature parameters to be extracted and the maximum total number of different control feature parameters to be extracted, with the largest total being the unified total for classification; and for different historical sub-item control data, based on the unified total for classification, the corresponding historical sub-item monitoring parameter matrix is ​​generated. The parameters are categorized and standardized to form a unified matrix of historical sub-item monitoring parameters. For different control characteristic parameters, the corresponding historical sub-item control parameter matrix is ​​then processed according to the unified total number of categories. The parameters are classified and processed uniformly to form a unified matrix of corresponding historical sub-item control parameters.

[0044] The data classification of historical sub-item control data mainly considers two aspects. First, the subsequent predictive model aims to predict accurate and reasonable control characteristic information based on the acquired monitoring characteristic data. Therefore, it is necessary to establish a mapping relationship between the monitoring parameter matrix extracted from the historical sub-item data and the corresponding control parameter matrix. Second, when establishing the mapping relationship between the monitoring parameter matrix and the control parameter matrix, the rules of matrix mapping must be considered, namely the element positions and the total number of elements in the two matrices. Therefore, when establishing the mapping relationship, i.e., during the process of unifying the total number of elements in both the monitoring parameter matrix and the control parameter matrix, it is also necessary to consider the supplementary positions of the matrix with fewer elements when supplementing elements, so that the two matrices have uniformity in element position and element structure arrangement. Of course, for the supplementation of elements in the matrix with fewer elements, the parameter values ​​of the supplemented elements can be unit parameter values ​​or other constants, as long as they do not affect the subsequent mapping relationship analysis.

[0045] Correlation clustering analysis was performed on different historical sub-item control data to form historical control correlation clustering data, including: for different historical sub-item control data, based on the corresponding historical sub-item monitoring parameter matrix. and the corresponding historical sub-term control parameter matrix Perform matrix merging to form the historical sub-item merged parameter matrix corresponding to the historical sub-item control data. Where i represents the index of a different parameter in the historical sub-item merged parameter matrix, and the total number of parameters in the historical sub-item merged parameter matrix is ​​equal to the number of parameters in the historical sub-item monitoring parameter matrix. Total number of parameters and historical sub-term control parameter matrix The sum of the total number of parameters; merging parameter matrices for different historical terms. Perform non-repetitive analysis on the relevant clusters to generate historical control correlation cluster data.

[0046] The essence of correlation clustering analysis is to cluster historical control data of different sub-items under the same operating conditions, thereby providing a more reasonable and effective foundation of big data for establishing prediction models for corresponding operating conditions. Considering that although the control sub-items differ under different operating conditions, the synergistic changes between monitoring parameters, between control parameters, and between monitoring and control parameters are closely correlated, merging the monitoring parameter matrix and the control parameter matrix before conducting correlation analysis can fully ensure the grasp of the magnitude of these three synergistic relationships. Simply performing correlation analysis between monitoring parameter matrices or between control parameter matrices alone cannot fully reflect the correlation of these three synergistic relationships. Of course, for data of different sub-items, the elements in different merged parameter matrices correspond to each other in terms of parameter type to ensure the rationality and effectiveness of the analysis. For correlation analysis, utilizing the correlation characteristics of matrices and extracting representative matrices for clustering around representative matrices is more reasonable and efficient.

[0047] Merge parameter matrices for different historical terms Perform non-repetitive analysis on the relevant clusters to generate historical control correlation clustering data, including: merging parameter matrices for different historical sub-terms. Extract any one and label it as the starting matrix for analysis; perform cluster non-repetition analysis based on the starting matrix in the following manner: determine the merging parameter matrix of the starting matrix and other historical terms. The cosine similarity value is obtained by merging the historical terms corresponding to the smallest cosine similarity value into the parameter matrix. Extract and determine the representative matrix; for the initial analysis matrix and the representative matrix, determine the parameter matrix to be merged with the remaining historical terms. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the closely related threshold. This forms the corresponding initial clustering matrix set. For different initial clustering matrix sets, the following judgment is made: if different initial clustering matrix sets contain the same historical term merging parameter matrix... Then, the initial analysis matrix is ​​determined as the calibration representative matrix, and the initial clustering matrix set corresponding to the initial analysis matrix is ​​determined as the historical control correlation clustering set corresponding to the calibration representative matrix; if there are no identical historical second-order merging parameter matrices in different initial clustering matrix sets... Then the selected representative matrix extracted this time will be merged with the remaining historical sub-items into a parameter matrix. The cosine similarity value is the smallest and is different from the extracted historical term merging parameter matrix. Historical term merged parameter matrix The newly determined representative matrix is ​​then used to determine the parameter matrix for merging with the remaining historical terms. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the closely related threshold. This forms the corresponding initial clustering matrix set; the evaluation continues for different initial clustering matrix sets until any two initial clustering matrix sets contain the same historical term merging parameter matrix. The initial analysis matrix and all selected representative matrices are then determined as calibration representative matrices, and the initial clustering matrix corresponding to the calibration representative matrix is ​​calibrated as the historical control correlation cluster set. Different calibration representative matrices and their corresponding historical control correlation cluster sets are combined to form historical control correlation cluster data.

[0048] The formation of control correlation clustering data is based on the magnitude of correlation for classification and clustering. It is understandable that the correlation between merged parameter matrices under different operating conditions is relatively weak. This correlation is represented in this application by the magnitude of the cosine similarity between matrices. The smaller the cosine similarity between two matrices, the weaker their correlation. Therefore, during analysis, an arbitrary matrix is ​​first extracted as the starting matrix, and the cosine similarity between the starting matrix and other matrices is determined. While the magnitude of the cosine similarity cannot directly determine whether the data from two matrices belong to the same operating condition, identifying the matrix with the smallest cosine similarity can determine the control data least likely to belong to the same operating condition. Thus, the extracted matrix least likely to belong to the same operating condition can be used as a reference to continuously obtain its cosine similarity with other remaining matrices to determine other parameter matrices least likely to belong to the same operating condition. For clustering based on representative matrices, considering data validity and to avoid clustering potentially different operating condition data together, the selected correlation threshold can be strictly limited, such as by a relatively large cosine similarity limit value. Of course, this clustering based on a high correlation threshold requires evaluation of the newly extracted representative matrix each time it is extracted. It's important to note that with progressively extracted representative matrices, as the number of remaining matrices gradually decreases, it's highly likely that the extracted matrix data will correspond to data from previous representative matrices under the same working conditions. Therefore, it's necessary to limit the number of extractions of the representative matrix, based on whether there are intersecting matrix data within the clustered data of the representative matrices. This effectively avoids generating different clustered data corresponding to different representative matrices due to more detailed feature divisions for the same working condition.

[0049] S2: Perform control prediction analysis on historical control correlation clustering data to establish control prediction model data.

[0050] Control prediction analysis is performed on historical control correlation clustering data to establish a control prediction model. This includes: unifying the monitoring parameters of different historical sub-terms within the historical control correlation clustering data based on the different historical control correlation clusters. And the corresponding historical secondary control parameter unified matrix A mapping matrix group is formed, where u represents the number of different historical control correlation clusters, and x represents the unified matrix of historical sub-item monitoring parameters. The unified parameter number for different monitoring characteristics, where y represents the unified matrix of historical sub-item control parameters. The unified parameter numbering for different control characteristics, and the unified matrix of historical sub-term control parameters. The total number and arrangement of elements in the matrix are consistent with the historical sub-item monitoring parameters. Same; establish clustering mapping relationship C u *F u =D u Based on the different mapping matrices in the historical control correlation clusters, the clustering mapping relationships are analyzed to determine the prediction relationship matrix F. u , where C u D represents the unified matrix of monitoring parameters corresponding to the historical control correlation cluster set numbered u. u C represents the unified control parameter matrix corresponding to the historical control correlation cluster set numbered u. u *F u C represents u The product of Fu; the historical secondary monitoring parameter matrix and cluster mapping relationship under the calibration representative matrix corresponding to different historical control correlation cluster sets are used to form control prediction model data.

[0051] Predictive analysis primarily involves establishing the correlation between the unified monitoring parameter matrix and the unified control parameter matrix within the corresponding cluster sets. The goal is to obtain a constant matrix that maps from the unified monitoring parameter matrix to the unified control layer matrix. Here, one approach is to select a number of mapping matrix sets that correspond to the number of unknowns in the constant matrix, or to extract multiple mapping matrix sets whose number corresponds to the number of unknowns in the constant matrix, solve them, and then average them to form the final constant matrix. Either approach is preferable as long as a reasonable mapping relationship can be established.

[0052] S3: Acquire real-time monitoring data and combine it with control prediction model data to perform control prediction analysis and form real-time control prediction data.

[0053] Acquire real-time monitoring data and combine it with control prediction model data to perform control prediction analysis, forming real-time control prediction data. This includes: extracting the corresponding real-time monitoring parameter matrix based on the real-time monitoring data, and performing the following analysis on the real-time monitoring parameter matrix and the corresponding historical sub-term monitoring parameter matrices under different calibration representative matrices: determining the cosine similarity value between the real-time monitoring parameter matrix and the corresponding historical sub-term monitoring parameter matrices under different calibration representative matrices, and labeling the clustering mapping relationship corresponding to the largest cosine similarity value as the real-time analysis and prediction model; performing control prediction analysis based on the real-time analysis and prediction model and the real-time monitoring parameter matrix to form real-time control prediction data.

[0054] To conduct control predictive analysis based on real-time monitoring data using control predictive model data, it is first necessary to identify the operating condition to which the real-time monitoring data belongs. This is done by extracting real-time monitoring parameter matrices that are similar to historical sub-term monitoring parameter matrices. Then, based on the different calibration representation matrices corresponding to the historical sub-term monitoring parameter matrices, the magnitude of the cosine similarity value is used to define the situation. It can be understood that the larger the cosine similarity value, the closer the real-time monitoring parameter matrix is ​​to the corresponding operating condition. Therefore, the operating condition with the largest cosine similarity value is determined as the operating condition of the real-time monitoring situation, and the corresponding mapping relationship is used as the prediction model for real-time analysis to complete the control predictive analysis.

[0055] Based on the real-time analysis and prediction model and the real-time monitoring parameter matrix, control prediction analysis is performed to generate real-time control prediction data, including: classifying and unifying the real-time monitoring parameter matrix to form a corresponding unified real-time monitoring parameter matrix; determining the unified real-time control parameter matrix based on the unified real-time monitoring parameter matrix and the real-time analysis and prediction model; and determining all control prediction parameters based on the unified real-time control parameter matrix.

[0056] Predictive analysis first requires obtaining a unified control monitoring parameter matrix based on the predictive analysis model. Before analysis, the real-time monitoring parameter matrix needs to be classified and unified to provide a unified data type for the predictive model, thereby obtaining a unified control parameter matrix. The extraction of different types of control parameters in the unified control parameter matrix can be achieved by reversing the classification and unification process of the historical sub-item control parameter matrix.

[0057] This invention also provides an artificial intelligence-based transformer control system, comprising: a data acquisition unit for acquiring historical control data and real-time monitoring data; a clustering analysis unit for performing control correlation-based clustering analysis on the historical control data acquired by the data acquisition unit to form historical control correlation clustering data; a training and learning unit for performing control prediction analysis on the historical control correlation clustering data formed by the clustering analysis unit to establish control prediction model data; and a real-time prediction unit for performing control prediction analysis based on the real-time monitoring data acquired by the data acquisition unit and the control prediction model data established by the training and learning unit to form real-time control prediction data.

[0058] The system collects basic big data and real-time monitoring data through a clustering acquisition unit, and further cleans and organizes the basic big data through a clustering analysis unit to form more effective and reasonable clustering data for model training. With the help of the training and learning unit, different clustering data are used as the basis for training to build an artificial intelligence control parameter prediction model. This model is then handed over to the real-time prediction unit to complete the analysis and prediction of control parameters based on the full collection of real-time monitoring data. The different functional units are interconnected to form a tight whole for realizing transformer control prediction and analysis, which is an important material basis for realizing artificial intelligence control of transformers.

[0059] In summary, the beneficial effects of the transformer control method and system based on artificial intelligence provided by the embodiments of the present invention are as follows:

[0060] This method achieves reasonable data clustering by performing correlation analysis on historical control data under different control sub-items based on monitoring data and corresponding control data. This forms a more control-correlated basic big data set of historical control data for different control sub-items. Then, it uses the basic big data set of clusters to achieve reasonable control prediction analysis and establish a corresponding control prediction model. This enables control parameter prediction based on real-time monitoring data, providing more accurate and reasonable control prediction information. On the one hand, reasonable control correlation clustering of historical control data can effectively ensure the relevance of subsequent model building. Training on these targeted basic big data sets of clusters can more accurately and effectively achieve control prediction for different control conditions. On the other hand, the control prediction model established with the help of historical control cluster data can achieve effective and accurate data prediction based on monitoring data, which is more adaptable and accurate than the big data processing method.

[0061] The system collects basic big data and real-time monitoring data through a clustering acquisition unit, and further cleans and organizes the basic big data through a clustering analysis unit to form more effective and reasonable clustering data for model training. With the help of the training and learning unit, different clustering data are used as the basis for training to build an artificial intelligence control parameter prediction model. This model is then handed over to the real-time prediction unit to complete the analysis and prediction of control parameters based on the full collection of real-time monitoring data. The different functional units are interconnected to form a tight whole for realizing transformer control prediction and analysis, which is an important material basis for realizing artificial intelligence control of transformers.

[0062] In the embodiments of this application, "instruction" can include direct and indirect instructions, as well as explicit and implicit instructions. The information indicated by a certain piece of information is called the information to be instructed. In the specific implementation process, there are many ways to instruct the information to be instructed, such as, but not limited to, directly instructing the information to be instructed, such as the information to be instructed itself or its index. It can also indirectly instruct the information to be instructed by instructing other information, where there is a relationship between the other information and the information to be instructed. It can also instruct only a part of the information to be instructed, while the other parts are known or pre-agreed upon. For example, the instruction of specific information can be achieved by using a pre-agreed (e.g., protocol-defined) arrangement of various pieces of information, thereby reducing instruction overhead to some extent. At the same time, common parts of various pieces of information can be identified and uniformly indicated to reduce the instruction overhead caused by individually indicating the same information.

[0063] Furthermore, the specific indication method can also be any existing indication method, such as, but not limited to, the above-mentioned indication methods and their various combinations. Specific details of various indication methods can be found in existing technologies, and will not be repeated here. As described above, for example, when multiple pieces of information of the same type need to be indicated, the indication methods for different pieces of information may differ. In the specific implementation process, the required indication method can be selected according to specific needs. This application embodiment does not limit the selected indication method; therefore, the indication methods involved in this application embodiment should be understood to cover various methods that enable the party to be indicated to obtain the information to be indicated.

[0064] It should be understood that the information to be indicated can be sent as a whole or divided into multiple sub-information messages sent separately, and the sending period and / or timing of these sub-information messages can be the same or different. The specific sending method is not limited in this application embodiment. The sending period and / or timing of these sub-information messages can be predefined, for example, according to a protocol, or configured by the sending device by sending configuration information to the receiving device.

[0065] "Predefined" or "pre-configured" can be achieved by pre-saving corresponding codes, tables, or other means that can be used to indicate relevant information in the device. This application does not limit the specific implementation method. "Saving" can refer to saving in one or more memories. These memories can be separate installations or integrated into the encoder, decoder, processor, or communication device. Alternatively, some memories can be separately installed, while others are integrated into the decoder, processor, or communication device. The type of memory can be any form of storage medium, and this application does not limit this.

[0066] The “protocol” mentioned in the embodiments of this application may refer to a protocol family in the field of communication, a standard protocol with a similar protocol family frame structure, or a related protocol applied to future communication systems. The embodiments of this application do not specifically limit this.

[0067] In the embodiments of this application, descriptions such as "when," "under the circumstances," "if," and "if" all refer to the device making corresponding processing under certain objective circumstances, and are not limited to a specific time. They do not require the device to make a judgment action during implementation, nor do they imply any other limitations.

[0068] In the description of the embodiments of this application, unless otherwise stated, " / " indicates that the objects before and after are in an "or" relationship. For example, A / B can represent A or B. "And / or" in the embodiments of this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Furthermore, in the description of the embodiments of this application, unless otherwise stated, "multiple" refers to two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple. Additionally, to facilitate a clear description of the technical solutions of the embodiments of this application, the terms "first" and "second" are used in the embodiments of this application to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first," "second," etc., do not limit the quantity or order of execution, and that "first," "second," etc., are not necessarily different. Furthermore, in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate that something is being used as an example, illustration, or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner for ease of understanding.

[0069] It should be understood that the processor in the embodiments of this application can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0070] It should also be understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0071] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0072] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0073] In this application, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0074] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0075] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0076] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0077] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0078] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0079] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0080] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0081] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An artificial intelligence-based transformer control method, characterized by, The method comprises the following steps: collecting historical control data, performing clustering analysis based on control correlation, and forming historical control correlation clustering data; performing control prediction analysis on the historical control correlation clustering data, and establishing control prediction model data; obtaining real-time monitoring data, combining the control prediction model data to perform control prediction analysis, and forming real-time control prediction data.

2. The transformer control method based on artificial intelligence according to claim 1, characterized in that, The method of collecting historical control data, performing clustering analysis based on control correlation, and forming historical control correlation clustering data comprises the following steps: extracting data for different control items from the historical control data to form historical sub-item control data corresponding to different control items; classifying different historical sub-item control data for monitoring control to form corresponding historical sub-item control classification data; performing correlation clustering analysis on different historical sub-item control data to form the historical control correlation clustering data.

3. The transformer control method based on artificial intelligence according to claim 2, characterized in that, The method of extracting data for different control items from the historical control data to form historical sub-item control data corresponding to different control items comprises the following steps: Different types of monitoring characteristic parameters are extracted for different historical sub-item control data, and a corresponding historical sub-item monitoring parameter matrix is established m represents the number of different historical sub-item control data, and n represents the number of different types of monitoring characteristic parameters Different types of control characteristic parameters are extracted for different historical sub-item control data, and a corresponding historical sub-item control parameter matrix is established k represents the number of different types of control characteristic parameters combining the different said historical subentry control data, the corresponding said historical subentry monitoring parameter matrix and the corresponding said historical subentry control parameter matrix to form the corresponding said historical subentry control data.

4. The transformer control method based on artificial intelligence according to claim 3, characterized by, The method of classifying different historical sub-item control data for monitoring control to form corresponding historical sub-item control classification data comprises the following steps: According to the total number of different monitoring characteristic parameters to be extracted and the total number of different control characteristic parameters to be extracted, and determining the maximum total number as the classification unified total number; For different historical sub-item control data, based on the unified total according to the classification, the corresponding historical sub-item monitoring parameter matrix is... The parameters are classified and standardized to form a unified matrix of corresponding historical sub-item monitoring parameters. According to the classification uniform total number, the corresponding historical sub-item control parameter matrix is formed for different control characteristic parameters Classification uniform processing is performed to form a corresponding historical sub-item control parameter uniform matrix.

5. The transformer control method based on artificial intelligence according to claim 4, characterized in that, The method of performing correlation clustering analysis on different historical sub-item control data to form the historical control correlation clustering data comprises the following steps: According to the corresponding historical sub-item monitoring parameter matrix and the corresponding historical sub-item control parameter matrix Matrix merging is performed to form a historical sub-item merged parameter matrix corresponding to the historical sub-item control data Wherein, i represents the number of different parameters in the historical sub-item merged parameter matrix, and the total number of parameters in the historical sub-item merged parameter matrix is the sum of the total number of parameters in the historical sub-item monitoring parameter matrix and the total number of parameters in the historical sub-item control parameter matrix . combining parameters matrices for different said historical sub-items performing non-redundant analysis for related clusters, forming said historical control relatedness cluster data.

6. The transformer control method based on artificial intelligence according to claim 5, characterized in that, said different said historical subterm merge parameter matrices performing a non-redundant analysis for related clusters, forming said historical control relatedness cluster data, comprising: combining parameters matrices for different said historical sub-items arbitrarily extract one and label as the analysis starting matrix; According to the analysis starting matrix, the following non-repetitive clustering analysis is performed: determining cosine similarity values of the analysis starting matrix with other historical subentry merge parameter matrices , and taking the historical subentry merge parameter matrix corresponding to the minimum cosine similarity value as the selection representative matrix ​ For the analysis starting matrix and the selected representative matrix, determine the parameter matrix to be merged with the remaining historical sub-items. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the correlation threshold. And form the corresponding initial clustering matrix set; Different initial clustering matrix sets are judged as follows: if same said history subterm merge parameter matrix exists in different said initial clustering matrix set then determine said analysis start matrix as a calibration representative matrix, and determine said initial clustering matrix set corresponding to said analysis start matrix as said history control correlation clustering set corresponding to said calibration representative matrix; if there is no same said history subterm merge parameter matrix in different said initial clustering matrix set then the cosine similarity value between said selected representative matrix extracted this time and the remaining other said history subterm merge parameter matrix is the smallest and different from said history subterm merge parameter matrix which has been extracted said history subterm merge parameter matrix is determined as the selected representative matrix; For the newly determined selection representative matrix, determine the parameter matrix to be merged with the remaining historical sub-items. The combined parameter matrix of all historical terms whose cosine similarity value is not less than the correlation threshold. And form the corresponding initial clustering matrix set; The judgment on different initial clustering matrix sets is continued until any two initial clustering matrix sets in different initial clustering matrix sets have the same historical subterm merging parameter matrix If the analysis starting matrix and all the selected representative matrices are determined as calibration representative matrices, the initial clustering matrix corresponding to the calibration representative matrix is calibrated as the historical control correlation clustering set. The historical control correlation clustering data is formed by collecting different calibration representative matrices and corresponding historical control correlation clustering sets.

7. The artificial intelligence-based transformer control method of claim 6, wherein, The method of performing control prediction analysis on the historical control correlation clustering data to establish control prediction model data comprises the following steps: According to different historical control correlation cluster sets in the historical control correlation cluster data, a historical secondary item monitoring parameter uniform matrix is monitored according to different historical secondary items in the historical control correlation cluster set And a corresponding historical secondary item control parameter uniform matrix A mapping matrix group is formed, wherein u represents the number of different historical control correlation cluster sets, x represents the number of different monitoring feature uniform parameters in the historical secondary item monitoring parameter uniform matrix Y represents the number of different control feature uniform parameters in the historical secondary item control parameter uniform matrix And the total number and arrangement of elements in the historical secondary item control parameter uniform matrix Are the same as those in the historical secondary item monitoring parameter uniform matrix ​ Establishing a cluster mapping relationship C u *F u =D u , according to different mapping matrices in the historical control correlation cluster set, analyzing the cluster mapping relationship to determine a prediction relationship matrix F u , wherein C u represents a corresponding monitoring parameter uniform matrix under the historical control correlation cluster set numbered u, D u represents a corresponding control parameter uniform matrix under the historical control correlation cluster set numbered u, C u *F u represents the product of C u and Fu. The control prediction model data is formed by collecting the historical sub-item monitoring parameter matrix under the calibration representative matrix corresponding to different historical control correlation clustering sets and the clustering mapping relationship.

8. The transformer control method based on artificial intelligence according to claim 7, characterized in that, The method of obtaining real-time monitoring data, combining the control prediction model data to perform control prediction analysis, and forming real-time control prediction data comprises the following steps: According to the real-time monitoring data, the corresponding real-time monitoring parameter matrix is extracted, and the real-time monitoring parameter matrix and the historical sub-item monitoring parameter matrix corresponding to different calibration representative matrices are analyzed as follows: The cosine similarity value of the real-time monitoring parameter matrix and the historical sub-item monitoring parameter matrix corresponding to different calibration representative matrices is determined, and the clustering mapping relationship corresponding to the maximum cosine similarity value is calibrated as a real-time analysis prediction model; According to the real-time analysis prediction model and the real-time monitoring parameter matrix, control prediction analysis is performed to form the real-time control prediction data.

9. The transformer control method based on artificial intelligence according to claim 8, characterized in that, The method of performing control prediction analysis according to the real-time analysis prediction model and the real-time monitoring parameter matrix to form the real-time control prediction data comprises the following steps: Classify and unify the real-time monitoring parameter matrix to form a corresponding real-time monitoring parameter unified matrix; Determine a real-time control parameter unified matrix according to the real-time monitoring parameter unified matrix and the real-time analysis prediction model; Determine all control prediction parameters according to the real-time control parameter unified matrix.

10. An artificial intelligence-based transformer control system employing the artificial intelligence-based transformer control method according to any one of claims 1 to 9, characterized by, Comprise: A data acquisition unit for acquiring historical control data and real-time monitoring data; A cluster analysis unit for performing cluster analysis on the historical control data acquired by the data acquisition unit based on control correlation to form historical control correlation cluster data; A training and learning unit for performing control prediction analysis on the historical control correlation cluster data formed by the cluster analysis unit to establish control prediction model data; A real-time prediction unit for performing control prediction analysis on the real-time monitoring data acquired by the data acquisition unit and the control prediction model data established by the training and learning unit to form real-time control prediction data.

Images