Intelligent control method and system for calibration of full-automatic sectional direct-current voltage divider

Abstract

The invention discloses an intelligent control method and system for calibration of a full-automatic sectional direct-current voltage divider, and relates to the technical field of intelligent control and decision making, and the method comprises the steps: automatically recognizing basic parameters of a calibrated voltage divider through a scanning test, detecting the sectional characteristics of the calibrated voltage divider, and building a characteristic model of the calibrated voltage divider; determining a calibration scheme according to the characteristic model, generating test point distribution, time sequence control parameters and a security policy, performing segmented calibration based on the calibration scheme, calculating error information of each test point, obtaining an independent compensation coefficient of each segment according to segmentation characteristics by combining temperature compensation, and performing compensation correction; and obtaining the corrected test point calibration result distribution to evaluate the health state of the calibrated voltage divider, and outputting a calibration conclusion including health state evaluation. Through the intelligent calibration process, the calibration efficiency and accuracy of the direct current voltage divider are remarkably improved, meanwhile, the operation risk is reduced, and reliable technical guarantee is provided for high-voltage direct current measurement.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control and decision-making technology, and more specifically, to an intelligent control method and system for fully automatic segmented DC voltage divider calibration. Background Art

[0002] As a key device in the high-voltage DC measurement system, the accuracy of the DC voltage divider directly affects the reliability of the entire measurement system. The traditional DC voltage divider calibration method has the following main technical defects: (1) Strong dependence on manual operation: The existing calibration process requires the participation of technicians throughout the process, including manually adjusting the output voltage of the standard source, observing and recording the measurement data, and manually switching the range. This "staring at the screen" operation mode is not only inefficient, but also prone to human errors. (2) Imperfect segment characteristic calibration: For segmented DC voltage dividers, the calibration accuracy of the traditional method in the segment transition area is insufficient. Operators often find it difficult to accurately capture the characteristic changes near the segment switching point, resulting in incomplete calibration data in the segment connection area. (3) Poor measurement synchronization: During the manual calibration process, there is a time difference between the acquisition of the standard value and the calibrated value. Especially when the voltage changes rapidly or there are fluctuations, this asynchrony will introduce significant measurement errors. (4) Weak data processing capabilities: The traditional method relies on manual calculation and judgment, and it is difficult to process complex data.

[0003] Some existing calibration systems implement program-controlled voltage output, but manual intervention is still required for range switching and data recording, making truly fully automated calibration impossible. Most calibration equipment is designed only for specific ranges and lacks the ability to automatically switch across the full range, making it unable to meet the complete calibration requirements of segmented voltage dividers. Existing highly automated systems often use fixed algorithms to process data, lacking intelligent adaptive processing capabilities and struggling to cope with complex operating conditions and nonlinear characteristics. Therefore, developing an intelligent calibration control method with full-range automatic switching and intelligent adaptive processing capabilities to automatically identify voltage divider characteristics and optimize calibration strategies is a pressing issue. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention proposes an intelligent control method and system for fully automatic segmented DC voltage divider calibration, which effectively solves the problems of low efficiency, poor accuracy, and insufficient safety of traditional calibration methods, and provides an innovative solution for the precise calibration of high-voltage DC measuring equipment.

[0005] A first aspect of the present invention provides an intelligent control method for fully automatic segmented DC voltage divider calibration, comprising the following steps:

[0006] The basic parameters of the calibrated voltage divider are automatically identified through the scan test, the segmented characteristics of the calibrated voltage divider are detected according to the basic parameters, and the characteristic model of the calibrated voltage divider is established;

[0007] A calibration scheme is determined according to the characteristic model of the calibrated voltage divider, test point distribution, timing control parameters and safety strategies are generated, segmented calibration is performed based on the calibration scheme, and after the segmented calibration is completed, full-range continuous scan test is performed;

[0008] The error information of each test point is calculated, the independent compensation coefficients of each segment are obtained in combination with temperature compensation for the segmented characteristics, and the calibration results of the test points are compensated and corrected;

[0009] The health status of the calibrated voltage divider is evaluated by using the corrected test point calibration results, a calibration conclusion containing the health status evaluation is output, and a calibration report is generated for visual display.

[0010] In the scheme, the basic parameters of the calibrated voltage divider are automatically identified through the scan test, specifically:

[0011] The self-checking program of the calibrated voltage divider is executed, the connection state and normal working state of the key components of the calibrated voltage divider are verified, the calibration parameters and compensation coefficients in the pre-preparation stage are loaded, the accurate test reference is obtained, and in addition, the environment monitoring is initialized, the environment data is collected, and the device self-checking and initialization are completed;

[0012] The voltage division ratio of the calibrated voltage divider is obtained through low-voltage scan test, the standard deviation of the voltage division ratio of each voltage point is calculated, the linearity is preliminarily judged using the standard deviation of the voltage division ratio, the linearity flag is generated according to the judgment result, and the nonlinear segment is marked;

[0013] The voltage division ratio change rate is identified according to the low-voltage scan test result, the voltage division ratio mutation interval is located, dense voltage points are set in the voltage division ratio mutation interval, the local slope of adjacent points is calculated, and the slope mutation point is identified as a segmented turning candidate point according to the comparison between the local slope and a preset threshold; the segmented turning candidate point is verified by repeated scanning, and the segmented turning voltage point is obtained;

[0014] The input impedance and output impedance of the calibrated voltage divider are evaluated, and the rated voltage range, voltage division ratio, segmented turning voltage point, input impedance and output impedance of the calibrated voltage divider are output as basic parameters.

[0015] In the scheme, the segmented characteristics of the calibrated voltage divider are detected according to the basic parameters, and the characteristic model of the calibrated voltage divider is established, specifically:

[0016] The calibration data, historical data and environmental data of the calibrated voltage divider are acquired, each linear segment and nonlinear segment is independently modeled according to the segmented turning voltage points, and the impedance network is modeled, and in the linear segment modeling, the Tukey weighting algorithm is used to eliminate the influence of abnormal points;

[0017] The physical mechanism of the calibrated voltage divider is acquired, the physical mechanism is used for big data retrieval analysis to acquire a supplementary model, the data model of the linear segment and the nonlinear segment and the impedance data model are integrated with the supplementary model to represent the segmented characteristics of the calibrated voltage divider;

[0018] After the multi-scale characteristic model is fused and verified, real-time data is queried and read and written in real time, the mapping relationship between the physical entity of the calibrated voltage divider and the digital twin model is established, and if the data deviation between the twin data and the actual running data is less than a preset deviation threshold, the digital twin model of the calibrated voltage divider is output as the characteristic model.

[0019] In the scheme, the calibration scheme is determined according to the characteristic model of the calibrated voltage divider, test point distribution, timing control parameters and safety strategies are generated, and specifically:

[0020] The linearity of the linear segment, the nonlinearity of the nonlinear segment, the thermal time constant and the historical calibration data trend are acquired as characteristic parameters of the calibrated voltage divider through the characteristic model of the calibrated voltage divider, the t-SNE dimension reduction algorithm is used to project the characteristic parameters to a low-dimensional space, and a corresponding feature scatter plot is generated;

[0021] The multi-source characteristic parameters in the voltage divider calibration instance are extracted, the feature scatter plot corresponding to the historical calibration requirement is generated, the feature scatter plot of the calibrated voltage divider is compared with the feature scatter plot of the historical calibration requirement, and the Wasserstein distance is used to quantify the distribution difference;

[0022] The historical calibration requirement meeting the preset distance threshold is selected, the historical calibration requirement is divided according to the test point distribution, the timing control parameters and the safety strategies, and the calibration requirement of the calibrated voltage divider is represented according to the requirement subset with data labels;

[0023] According to the requirement subset with different data labels, the scheme items with interaction in the voltage divider calibration instance are extracted, the requirement-scheme item subgraph of different data labels is constructed, the graph convolution network is used to learn the requirement-scheme item subgraph, and the local requirement feature representation corresponding to different data labels is updated through message propagation and neighbor aggregation;

[0024] The local demand feature representations corresponding to different data labels are cross-label spliced ​​to generate a global demand feature representation. The CTR estimation method is used to build a calibration solution recommendation model. The global demand feature representation and solution item feature representation are scored. The solution items with the highest scores in the three dimensions of test point distribution, timing control parameters, and security policy are screened to generate a calibration solution.

[0025] In this solution, segmented calibration is performed based on the calibration solution. After the segmented calibration is completed, a full-range continuous scan test is performed. Specifically:

[0026] Perform segmented calibration and full-range continuous sweep testing according to the calibration scheme, initialize the test point distribution of different segments and set segment-specific parameters, perform voltage control and data acquisition according to the calibration scheme, and compare the digital twin prediction value of the characteristic model in real time. When the deviation is greater than the preset threshold, retesting is triggered;

[0027] Use full-scale continuous sweep test to verify the overall consistency after segmented calibration, obtain the calibration data table, feed back the error distribution characteristics of the sweep test to the calibration scheme recommendation model and characteristic model, and update the model parameters and segmented parameters.

[0028] In this solution, the error information of each test point is calculated, and the independent compensation coefficient of each segment is obtained according to the segmented characteristics in combination with temperature compensation. The calibration results of the test points are compensated and corrected. Specifically,

[0029] Calculate the proportional error, linear error, and repeatability error for each test point, generate an error-voltage curve based on the calculated errors, analyze the error distribution of each segment using the error-voltage curve, and establish an error profile for each segment using the error distribution and the temperature-error relationship matrix combined with the segment characteristics;

[0030] A stacking strategy is introduced to build a retrieval model. A KNN model is trained based on the distance feature vectors corresponding to the Euclidean distance, Manhattan distance, and Fréchet distance. A preset number of voltage divider calibration instances are output and scored based on the calibration results.

[0031] The voltage divider calibration instances and scores corresponding to different distances are spliced ​​into meta-feature vectors, and the meta-feature vectors are used to train the neural network model to learn the weights of each distance metric. The training output retrieval model is iteratively trained, and the error profile is imported into the retrieval model to obtain the voltage divider calibration instance with the highest score. The calibration value deviation is extracted to obtain the corresponding segmented compensation coefficient, and the calibration results of the test points are compensated and corrected.

[0032] In this solution, the corrected test point calibration results are obtained to evaluate the health status of the calibrated voltage divider, and the calibration conclusion including the health status evaluation is output, specifically:

[0033] Obtaining historical calibration data of the calibrated voltage divider, training a health status model of the calibrated voltage divider according to an autoencoder structure based on the historical calibration data, encoding the historical calibration data using multi-layer sensing, and extracting nonlinear features from the historical calibration data;

[0034] The encoded data is imported into the LSTM autoencoder network to obtain time series features, and the time series features are used to guide the multi-layer perceptron to perform decoding and optimize the data reconstruction process. When the reconstruction error is less than the preset threshold, the trained health status model is output;

[0035] Importing the calibration result of the calibrated voltage divider into the health status model, obtaining an estimated calibration result of the calibrated voltage divider under a historical health status, comparing the estimated calibration result with the actual calibration result, and generating a health status abnormality warning when the data residual value is greater than a preset threshold;

[0036] The dynamic time warping algorithm is used to obtain the dynamic warping distance between the current calibration data and the historical calibration data of the calibrated voltage divider, and the corresponding dynamic warping distance curve is constructed;

[0037] A genetic algorithm is used to optimize degradation indicators to obtain the optimal degradation indicator combination corresponding to the dynamic regularization distance curve. Indicator features are extracted based on the optimal degradation indicator combination. The indicator features are imported into the fully connected layer to calculate the matching score of each fault mode, and the fault category is determined to obtain the health status assessment result of the calibrated voltage divider;

[0038] Generate standardized calibration reports and data archive using the health status assessment results of the calibrated voltage divider and the current calibration results.

[0039] A second aspect of the present invention provides an intelligent control system for fully automatic segmented DC voltage divider calibration, the system comprising: a data acquisition module, a segmented characteristic modeling module, a calibration scheme determination module, a calibration scheme execution module, an error analysis and compensation module, a health status assessment module, and a report output module;

[0040] The data acquisition module obtains basic parameters and real-time measurement data of the calibrated voltage divider;

[0041] The segmented characteristic modeling module detects the segmented characteristics of the calibrated voltage divider based on basic parameters, determines the turning voltage point of each linear segment, and establishes a characteristic model of the calibrated voltage divider;

[0042] The calibration scheme determination module determines the calibration scheme according to the characteristic model of the calibrated voltage divider, and generates the test point distribution, timing control parameters and safety strategy;

[0043] The calibration scheme execution module automatically executes the calibration process based on the calibration scheme and outputs the standard value, calibrated value and environmental parameters of each test point;

[0044] The error analysis and compensation module calculates the error information of each test point, obtains the independent compensation coefficient of each segment according to the segment characteristics in combination with temperature compensation, and performs compensation correction on the calibration results of the test points;

[0045] The health status assessment module assesses abnormal health status and faults of the voltage divider and obtains a health status assessment result of the calibrated voltage divider;

[0046] The report output module obtains the calibration conclusion including the health status assessment and generates a calibration report for visual display and data management.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] This invention reduces human intervention, improves the calibration efficiency of segmented DC voltage dividers, automatically identifies the segmented characteristics of the voltage divider, and introduces an intelligent algorithm to optimize the calibration strategy for adaptive calibration. The test plan is automatically adjusted based on the characteristics of the voltage divider being calibrated. Multiple protections are implemented to ensure safe high-voltage operation, including unattended operation. Finally, a complete calibration data management system is established, enabling data traceability and trend analysis. This intelligent calibration system significantly improves the efficiency and accuracy of DC voltage divider calibration while reducing operational risks, providing reliable technical support for high-voltage DC measurements. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or exemplary descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to these drawings without paying any creative work.

[0050] Figure 1 A flow chart showing an intelligent control method for fully automatic segmented DC voltage divider calibration is provided;

[0051] Figure 2 A flow chart showing a calibration scheme based on a characteristic model of a calibrated voltage divider is shown;

[0052] Figure 3 A flow chart showing the use of calibration results to evaluate the health status of the calibrated voltage divider is shown;

[0053] Figure 4 The block diagram of the intelligent control system for fully automatic segmented DC voltage divider calibration is shown. DETAILED DESCRIPTION

[0054] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0056] Figure 1 A flow chart of an intelligent control method for fully automatic segmented DC voltage divider calibration is shown.

[0057] like Figure 1 As shown, this embodiment provides an intelligent control method for fully automatic segmented DC voltage divider calibration, including:

[0058] S102, automatically identifying basic parameters of the calibrated voltage divider through a scan test, detecting segmented characteristics of the calibrated voltage divider based on the basic parameters, and establishing a characteristic model of the calibrated voltage divider;

[0059] S104, determining a calibration scheme based on a characteristic model of the voltage divider to be calibrated, generating a test point distribution, timing control parameters, and a safety policy, performing segmented calibration based on the calibration scheme, and performing a full-range continuous sweep test after completing the segmented calibration;

[0060] S106, calculating error information of each test point, obtaining independent compensation coefficients for each segment based on the segmented characteristics in combination with temperature compensation, and performing compensation correction on the calibration results of the test points;

[0061] S108, obtaining the corrected test point calibration results to evaluate the health status of the calibrated voltage divider, outputting a calibration conclusion including the health status evaluation, and generating a calibration report for visual display.

[0062] It should be noted that the self-test program of the calibrated voltage divider is executed to verify the connection status and normal working condition of the key components of the calibrated voltage divider, load the calibration parameters and compensation coefficients of the pre-preparation stage, obtain an accurate test benchmark, initialize environmental monitoring, collect environmental data, and complete equipment self-test and initialization; obtain the voltage divider ratio of the calibrated voltage divider through a low-voltage scanning test, starting from a very small voltage (such as 0.1% of the rated value) to avoid shock, and increase in steps of 5% to 10% of the rated value, covering the rated voltage range of 0% to 30%. After each voltage point stabilizes (judgment condition: fluctuation <0.01% within 1 second), the standard source output voltage and the voltage divider output value are synchronously collected to calculate the voltage divider ratio. The standard deviation of the voltage divider ratio at each voltage point is calculated and used to make an initial linearity assessment. A linearity marker is generated based on the assessment result, and nonlinear segments are marked for subsequent fine-scanning. Based on the low-voltage sweep test results, the voltage divider ratio change rate is identified, and the voltage divider ratio mutation interval is located. Within this voltage divider ratio mutation interval, densely packed voltage points are set, and the local slopes of adjacent points are calculated. These local slopes are compared with a preset threshold to identify slope mutation points as candidate segmental turning points. These candidate segmental turning points are then repeatedly scanned and verified to obtain the segmental turning voltage points. The input and output impedances of the voltage divider being calibrated are evaluated to ensure that they do not affect the load effect of the calibration system. The rated voltage range, voltage divider ratio, segmental turning voltage point, input impedance, and output impedance of the voltage divider being calibrated are output as basic parameters.

[0063] The physical entity of the voltage divider to be calibrated and the calibration data, historical data and environmental data obtained from the scan test are obtained. Each linear segment and nonlinear segment are independently modeled according to the segmented turning voltage point, and the impedance network is modeled. In the linear segment modeling, the Tukey weighted algorithm is used to eliminate the influence of abnormal points. In the independent modeling of the nonlinear segment, the Sigmoid function is used for transition in the ±5% range of the segment point.

[0064] The physical mechanism of the calibrated voltage divider is acquired and used to perform big data retrieval and analysis to obtain supplementary models, such as thermoelectric coupling models and dielectric response models. The data models for the linear and nonlinear segments, as well as the impedance data models, are integrated with the supplementary models. For example, for the linear segment, the data model is 70% data model + 30% physical model, and for the nonlinear segment, the data model is 50% data model + 50% physical model. Multi-model integration is used to characterize the segmented characteristics of the calibrated voltage divider. After fusion and verification of the multi-scale characteristic models, real-time data query and reading and writing are performed, establishing a mapping relationship between the physical entity of the calibrated voltage divider and its digital twin model. After each calibration, the model parameters are updated using the RLS algorithm. If the data deviation between the twin data and the actual operating data is less than a preset deviation threshold, the digital twin model of the calibrated voltage divider is output as the characteristic model. This dual-driven modeling approach of "data + mechanism" achieves high-fidelity digital mapping of the voltage divider's characteristics, providing a precise virtual testing environment for the intelligent calibration system.

[0065] Figure 2 A flow chart for determining a calibration solution based on a characteristic model of the calibrated voltage divider is shown.

[0066] According to an embodiment of the present invention, a calibration scheme is determined based on the characteristic model of the voltage divider to be calibrated, and a test point distribution, timing control parameters, and security strategy are generated, specifically:

[0067] S202, obtaining the linearity of the linear segment, the nonlinearity of the nonlinear segment, the thermal time constant, and the trend of historical calibration data as characteristic parameters of the calibrated voltage divider through the characteristic model of the calibrated voltage divider, projecting the characteristic parameters into a low-dimensional space using a t-SNE dimensionality reduction algorithm, and generating a corresponding characteristic scatter plot;

[0068] S204, extracting multi-source characteristic parameters from the voltage divider calibration instance, generating a characteristic scatter plot corresponding to the historical calibration requirements, comparing the characteristic scatter plot of the voltage divider to be calibrated with the characteristic scatter plot of the historical calibration requirements, and quantifying the distribution difference using Wasserstein distance;

[0069] S206, selecting historical calibration requirements that meet a preset distance threshold, dividing the historical calibration requirements according to test point distribution, timing control parameters, and security policies, and characterizing the calibration requirements of the voltage divider to be calibrated based on the requirement subsets with data tags;

[0070] S208: Extracting interactive solution items from the extraction voltage divider calibration instance based on the requirement subsets with different data labels, constructing a requirement-solution item subgraph with different data labels, learning the requirement-solution item subgraph using a graph convolutional network, and updating the local requirement feature representations corresponding to the different data labels through message propagation and neighbor aggregation;

[0071] S210, cross-label splicing of local demand feature representations corresponding to different data labels generates a global demand feature representation, uses the CTR estimation method to build a calibration solution recommendation model, calculates the scores of the global demand feature representation and the solution item feature representation, selects the solution items with the highest scores in the three dimensions of test point distribution, timing control parameters, and security policy, and generates a calibration solution.

[0072] It should be noted that according to each section R 2 The linearity of the linear segment is obtained by the value, and the nonlinearity of the nonlinear segment is obtained by the radius of curvature. The linearity of the linear segment, the nonlinearity of the nonlinear segment, the thermal time constant, and the trend of historical calibration data are used as the characteristic parameters of the voltage divider under calibration. The Euclidean distance between the current voltage divider characteristics (voltage divider ratio, linearity, segmentation points, etc.) and the historical calibration data is calculated, and the top-N (N ≥ 50) historical calibration data with distances less than a threshold are selected. The historical calibration requirements are divided according to the test point distribution, timing control parameters, and safety policy. The test point distribution is normalized into a voltage percentage sequence in the [0, 1] interval. The timing parameters are structured into tuples of (settling time, number of samples, interval time), and the safety policy is encoded as a rule set of (maximum boost rate, current threshold, protection action). Based on the requirement subsets with different data labels, interactive solution items are extracted from the extracted voltage divider calibration instance. A requirement-solution item subgraph with different data labels is constructed. Requirement nodes represent the calibration requirement characteristics under specific labels, while solution item nodes are the underlying solution components. A connection is established when a requirement adopts a solution item. Use the CTR estimation method to build a calibration solution recommendation model, and calculate the scores of the global demand feature representation and solution item feature representation. represents the feature representation of the solution item corresponding to the solution item, H global represents the global demand feature representation, b j After obtaining the calibration solution, the feasibility of the recommended solution is verified through digital twin rapid simulation, and combinations that violate safety rules (such as excessive temperature rise) are eliminated.

[0073] It should be noted that according to the calibration scheme, segmented calibration and full-scale continuous scanning test are performed respectively, the test point distribution of different segments is initialized (such as low-voltage segment: [10%, 30%, 50%] rated value) and segment-specific parameters are set: sampling times and stability criteria, voltage control and data acquisition are performed according to the calibration scheme, and the digital twin prediction value of the characteristic model is compared in real time. When the deviation is greater than the preset threshold, re-testing is triggered; before reaching the segment point (such as 50% rated value), the voltage is reduced to 80% of the turning point and paused for 60 seconds to balance the thermal state, the segment point ±5% range is scanned with a step size of 0.5%, the transient response (establishment time, overshoot) is recorded, and the end data verification is performed based on the acquired data.

[0074] Use a full-scale continuous sweep test to verify the overall consistency after segmented calibration. Set the sweep range, sweep rate, and sampling density. Preferably, the sweep range is set to 0% → 105% of the rated voltage; the sweep rate is set to 0.5% of the rated value / second; and the sampling density is set to collect a set of data every 100ms. The error distribution characteristics are calculated based on real-time error. A calibration data sheet is obtained, which includes data such as the standard value, the calibrated value, temperature, and relative error. The error distribution characteristics of the sweep test are fed back into the calibration solution recommendation model and characteristic model to update the model parameters and segmented parameters. Through strict closed-loop control and multiple verifications, the calibration results are ensured to meet both segmented accuracy and full-scale consistency requirements.

[0075] The proportional error, linear error, and repeatability error are calculated for each test point. An error-voltage curve is generated based on the calculated errors. The error distribution of each segment is analyzed using the error-voltage curve. The error distribution and the temperature-error relationship matrix are combined with the segmentation characteristics to create an error profile for each segment. The temperature-error relationship matrix represents the quantitative results of the temperature effect. A stacking strategy is introduced to construct a retrieval model. A KNN model is trained based on the distance feature vectors corresponding to the Euclidean distance, Manhattan distance, and Fréchet distance. A preset number of voltage divider calibration examples are output and scored based on the calibration results. The voltage divider calibration examples corresponding to different distances and the scores are spliced ​​into meta-feature vectors. The meta-feature vectors are used to train a neural network model to learn the weights of each distance metric. The training is iterative and the output retrieval model is then imported into the retrieval model to obtain the voltage divider calibration example with the highest score. The calibration value deviation is extracted to obtain the corresponding segmentation compensation coefficient, and the calibration results of the test points are compensated and corrected.

[0076] Figure 3 A flow chart is shown for evaluating the health status of the calibrated voltage divider using the calibration results.

[0077] According to an embodiment of the present invention, the corrected test point calibration result is obtained to evaluate the health status of the calibrated voltage divider, and a calibration conclusion including the health status evaluation is output, specifically:

[0078] S302, obtaining historical calibration data of the calibrated voltage divider, training a health status model of the calibrated voltage divider based on the historical calibration data using an autoencoder structure, encoding the historical calibration data using multi-layer sensing, and extracting nonlinear features from the historical calibration data;

[0079] S304: Importing the encoded data into an LSTM autoencoder network to obtain time series features, using the time series features to guide the multi-layer perceptron to perform decoding and optimize the data reconstruction process. When the reconstruction error is less than a preset threshold, the trained health status model is output;

[0080] S306: Importing the calibration result of the calibrated voltage divider into the health status model to obtain an estimated calibration result of the calibrated voltage divider under a historical health status, comparing the estimated calibration result with the actual calibration result, and generating a health status abnormality warning when the data residual value is greater than a preset threshold;

[0081] S308, using a dynamic time warping algorithm to obtain a dynamic warping distance between the current calibration data and the historical calibration data of the calibrated voltage divider, and constructing a corresponding dynamic warping distance curve;

[0082] S310: Optimizing degradation indicators using a genetic algorithm to obtain an optimal degradation indicator combination corresponding to the dynamic regularized distance curve; extracting indicator features based on the optimal degradation indicator combination; importing the indicator features into a fully connected layer to calculate a matching score for each fault mode; determining the fault category; and obtaining a health status assessment result of the calibrated voltage divider.

[0083] S312, generating a standardized calibration report and data archiving using the health status assessment result of the calibrated voltage divider and the current calibration result.

[0084] It should be noted that a multi-layer perceptron and the LeakyReLU activation function are used to capture the nonlinear relationships in the calibration data and obtain the weak nonlinear features in the calibration data. An LSTM autoencoder network is used to obtain the time series features after dimensionality reduction. The time series features are used to optimize the encoding reconstruction and reduce the reconstruction error. Finally, a multi-layer perceptron is used for decoding to reconstruct the input data. The data reconstruction error guides the training of a data-driven model corresponding to the health status of the calibrated voltage divider. Since the health status model of the calibrated voltage divider is constructed based on historical calibration data, if the current calibration data deviates significantly from the estimated calibration data driven by the historical health status, it indicates that the health status of the calibrated voltage divider is abnormal.

[0085] A dynamic time warping algorithm is used to obtain the dynamic warping distance between the current calibration data and historical calibration data of the calibrated voltage divider. The historical calibration data are aligned at the same test point voltage values ​​and uniformly converted to the error value at a standard temperature of 25°C. A eigenvector is generated for the kth calibration data. The mean of the eigenvectors of the first three calibration data is selected as a reference benchmark. The Euclidean distance between the current data and the benchmark is calculated to generate the dynamic warping distance, and a corresponding dynamic warping distance curve is constructed. A genetic algorithm is used to optimize the degradation indicators, selecting the most important, most relevant, and least redundant degradation indicator combination as the optimal degradation indicator combination. The optimal combination corresponding to the dynamic warping distance curve is obtained, with the distance growth rate, acceleration, and local mutation rate being the preferred factors. A mapping table between degradation indicators and faults is constructed. For example, when the distance growth rate increases suddenly, the possible root cause of the fault is segmented resistor aging or contact oxidation. In the fully connected layer, a matching score for each fault mode is calculated based on the degradation indicators and the fault mapping table. The DS evidence theory is used to fuse multiple indicators to improve the confidence of fault mode recognition. Determine the fault category to obtain the health status assessment result of the calibrated voltage divider, preferably obtain the health status score f according to the fault category, Among them I fault j is the fault indication factor, which takes the value of 1 when confirmed and 0.5 when suspected. BTW represents the dynamic regularized distance curve, and k is the total number of failure modes.

[0086] Generates a complete calibration certificate in accordance with the JJG standard format, including error curves and automatically annotates unqualified items and out-of-tolerance test points. Additionally, generates visualizations such as voltage divider ratio-voltage curves, error distribution graphs, and historical trend charts. Output calibration results are stored in an SQL database, supporting query and export.

[0087] Figure 4 A block diagram of an intelligent control system for fully automatic segmented DC voltage divider calibration is shown.

[0088] The second embodiment of the present invention provides an intelligent control system 4 for fully automatic segmented DC voltage divider calibration, the system comprising: a data acquisition module 401, a segmented characteristic modeling module 402, a calibration scheme determination module 403, a calibration scheme execution module 404, an error analysis and compensation module 405, a health status assessment module 406, and a report output module 407;

[0089] The data acquisition module obtains basic parameters and real-time measurement data of the calibrated voltage divider;

[0090] The segmented characteristic modeling module detects the segmented characteristics of the calibrated voltage divider based on basic parameters, determines the turning voltage point of each linear segment, and establishes a characteristic model of the calibrated voltage divider;

[0091] The calibration scheme determination module determines the calibration scheme according to the characteristic model of the calibrated voltage divider, and generates the test point distribution, timing control parameters and safety strategy;

[0092] The calibration scheme execution module automatically executes the calibration process based on the calibration scheme and outputs the standard value, calibrated value and environmental parameters of each test point;

[0093] The error analysis and compensation module calculates the error information of each test point, obtains the independent compensation coefficient of each segment according to the segment characteristics in combination with temperature compensation, and performs compensation correction on the calibration results of the test points;

[0094] The health status assessment module assesses abnormal health status and faults of the voltage divider and obtains a health status assessment result of the calibrated voltage divider;

[0095] The report output module obtains the calibration conclusion including the health status assessment and generates a calibration report for visual display and data management.

[0096] The third aspect of the present invention provides a computer-readable storage medium, which includes a program for an intelligent control method for fully automatic segmented DC voltage divider calibration. When the program for the intelligent control method for fully automatic segmented DC voltage divider calibration is executed by a processor, the steps of the intelligent control method for fully automatic segmented DC voltage divider calibration are implemented.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described above are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple modules or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms. In addition, the functional modules in the various embodiments of the present invention can all be integrated into one processing module, or each module can be a separate module, or two or more modules can be integrated into one module; the above-mentioned integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0098] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0099] Alternatively, if the above-mentioned integrated module of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0100] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed by the present invention, which should be covered by the scope of protection of the present invention.

Claims

1. An intelligent control method for fully automatic segmented DC voltage divider calibration, characterized in that: The following steps are involved: Automatically identifying basic parameters of the calibrated voltage divider through a scanning test, detecting segmented characteristics of the calibrated voltage divider based on the basic parameters, and establishing a characteristic model of the calibrated voltage divider; Determine a calibration scheme based on the characteristic model of the voltage divider to be calibrated, generate test point distribution, timing control parameters and safety strategy, perform segmented calibration based on the calibration scheme, and perform full-range continuous sweep test after completing the segmented calibration; Calculate the error information of each test point, obtain the independent compensation coefficient of each segment according to the segment characteristics in combination with temperature compensation, and perform compensation correction on the calibration result of the test point; The corrected test point calibration results are obtained to evaluate the health status of the calibrated voltage divider, a calibration conclusion including the health status evaluation is output, and a calibration report is generated for visual display.

2. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: The basic parameters of the voltage divider to be calibrated are automatically identified through the scan test, specifically: Execute the self-test program of the calibrated voltage divider to verify the connection status and normal operation of the key components of the calibrated voltage divider, load the calibration parameters and compensation coefficients of the pre-preparation phase, obtain an accurate test benchmark, initialize environmental monitoring, collect environmental data, and complete the equipment self-test and initialization; Obtain the voltage divider ratio of the calibrated voltage divider through a low-voltage sweep test, calculate the standard deviation of the voltage divider ratio at each voltage point, use the standard deviation of the voltage divider ratio to make a preliminary linearity judgment, generate a linearity mark based on the judgment result, and mark the nonlinear segment; Identify the voltage division ratio change rate based on the low voltage scan test results, locate the voltage division ratio mutation interval, set dense voltage points in the voltage division ratio mutation interval, calculate the local slopes of adjacent points, compare the local slopes with a preset threshold, identify the slope mutation points as segmented turning point candidates, repeatedly scan and verify the segmented turning point candidates, and obtain segmented turning voltage points; The input impedance and output impedance of the calibrated voltage divider are evaluated, and the rated voltage range, voltage division ratio, segmented turning voltage point, input impedance and output impedance of the calibrated voltage divider are output as basic parameters.

3. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: The segmented characteristics of the calibrated voltage divider are detected according to the basic parameters, and a characteristic model of the calibrated voltage divider is established, specifically: Obtain the physical entity of the voltage divider to be calibrated and the calibration data, historical data, and environmental data obtained from the scan test. Independently model each linear segment and nonlinear segment based on the segmented break voltage points, and model the impedance network. In the linear segment modeling, the Tukey weighting algorithm is used to eliminate the influence of outliers. Obtaining the physical mechanism of the calibrated voltage divider, using the physical mechanism to perform big data retrieval and analysis to obtain a supplementary model, integrating the linear segment and nonlinear segment data models and the impedance data model with the supplementary model to characterize the segmented characteristics of the calibrated voltage divider; After the fusion verification of the multi-scale characteristic model, real-time query and reading and writing of real-time data are performed to establish a mapping relationship between the physical entity of the calibrated voltage divider and the digital twin model. If the data deviation between the twin data and the actual operation data is less than the preset deviation threshold, the digital twin model of the calibrated voltage divider is output as the characteristic model.

4. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: The calibration scheme is determined based on the characteristic model of the voltage divider to be calibrated, and the test point distribution, timing control parameters and safety strategy are generated. Specifically: The linearity of the linear segment, the nonlinearity of the nonlinear segment, the thermal time constant, and the trend of historical calibration data are obtained as characteristic parameters of the calibrated voltage divider through the characteristic model of the calibrated voltage divider, and the characteristic parameters are projected into a low-dimensional space using the t-SNE dimensionality reduction algorithm to generate a corresponding characteristic scatter plot; Extract multi-source characteristic parameters from the voltage divider calibration instance, generate a characteristic scatter plot corresponding to the historical calibration requirements, compare the characteristic scatter plot of the voltage divider to be calibrated with the characteristic scatter plot of the historical calibration requirements, and quantify the distribution difference using the Wasserstein distance; Selecting historical calibration requirements that meet a preset distance threshold, dividing the historical calibration requirements according to test point distribution, timing control parameters, and security policies, and characterizing the calibration requirements of the voltage divider under calibration based on the requirement subsets with data labels; Extracting interacting solution items from the extraction divider calibration instance based on the requirement subsets with different data labels, constructing a requirement-solution item subgraph with different data labels, learning the requirement-solution item subgraph using a graph convolutional network, and updating the local requirement feature representation corresponding to different data labels through message propagation and neighbor aggregation; The local demand feature representations corresponding to different data labels are cross-label spliced ​​to generate a global demand feature representation. The CTR estimation method is used to build a calibration solution recommendation model. The global demand feature representation and solution item feature representation are scored. The solution items with the highest scores in the three dimensions of test point distribution, timing control parameters, and security policy are screened to generate a calibration solution.

5. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: Perform segmented calibration based on the calibration scheme, and perform full-range continuous sweep test after completing segmented calibration, specifically: Perform segmented calibration and full-range continuous sweep testing according to the calibration scheme, initialize the test point distribution of different segments and set segment-specific parameters, perform voltage control and data acquisition according to the calibration scheme, and compare the digital twin prediction value of the characteristic model in real time. When the deviation is greater than the preset threshold, retesting is triggered; Use full-scale continuous sweep test to verify the overall consistency after segmented calibration, obtain the calibration data table, feed back the error distribution characteristics of the sweep test to the calibration scheme recommendation model and characteristic model, and update the model parameters and segmented parameters.

6. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: Calculate the error information of each test point, combine temperature compensation to obtain the independent compensation coefficient of each segment according to the segment characteristics, and perform compensation correction on the calibration results of the test points, specifically: Calculate the proportional error, linear error, and repeatability error for each test point, generate an error-voltage curve based on the calculated errors, analyze the error distribution of each segment using the error-voltage curve, and establish an error profile for each segment using the error distribution and the temperature-error relationship matrix combined with the segment characteristics; A stacking strategy is introduced to build a retrieval model. A KNN model is trained based on the distance feature vectors corresponding to the Euclidean distance, Manhattan distance, and Fréchet distance. A preset number of voltage divider calibration instances are output and scored based on the calibration results. The voltage divider calibration instances and scores corresponding to different distances are spliced ​​into meta-feature vectors, and the meta-feature vectors are used to train the neural network model to learn the weights of each distance metric. The training output retrieval model is iteratively trained, and the error profile is imported into the retrieval model to obtain the voltage divider calibration instance with the highest score. The calibration value deviation is extracted to obtain the corresponding segmented compensation coefficient, and the calibration results of the test points are compensated and corrected.

7. The intelligent control method for fully automatic segmented DC voltage divider calibration according to claim 1, characterized in that: Obtain the corrected test point calibration results to evaluate the health status of the calibrated voltage divider, and output the calibration conclusion including the health status evaluation, specifically: Obtaining historical calibration data of the calibrated voltage divider, training a health status model of the calibrated voltage divider according to an autoencoder structure based on the historical calibration data, encoding the historical calibration data using multi-layer sensing, and extracting nonlinear features from the historical calibration data; The encoded data is imported into the LSTM autoencoder network to obtain time series features, and the time series features are used to guide the multi-layer perceptron to perform decoding and optimize the data reconstruction process. When the reconstruction error is less than the preset threshold, the trained health status model is output; Importing the calibration result of the calibrated voltage divider into the health status model, obtaining an estimated calibration result of the calibrated voltage divider under a historical health status, comparing the estimated calibration result with the actual calibration result, and generating a health status abnormality warning when the data residual value is greater than a preset threshold; The dynamic time warping algorithm is used to obtain the dynamic warping distance between the current calibration data and the historical calibration data of the calibrated voltage divider, and the corresponding dynamic warping distance curve is constructed; A genetic algorithm is used to optimize degradation indicators to obtain the optimal degradation indicator combination corresponding to the dynamic regularization distance curve. Indicator features are extracted based on the optimal degradation indicator combination. The indicator features are imported into the fully connected layer to calculate the matching score of each fault mode, and the fault category is determined to obtain the health status assessment result of the calibrated voltage divider; Generate standardized calibration reports and data archive using the health status assessment results of the calibrated voltage divider and the current calibration results.

8. An intelligent control system for fully automatic segmented DC voltage divider calibration, characterized in that: An intelligent control method for fully automatic segmented DC voltage divider calibration according to any one of claims 1 to 7 is implemented, the system comprising: a data acquisition module, a segmented characteristic modeling module, a calibration scheme determination module, a calibration scheme execution module, an error analysis and compensation module, a health status assessment module, and a report output module; The data acquisition module obtains basic parameters and real-time measurement data of the calibrated voltage divider; The segmented characteristic modeling module detects the segmented characteristics of the calibrated voltage divider based on basic parameters, determines the turning voltage point of each linear segment, and establishes a characteristic model of the calibrated voltage divider; The calibration scheme determination module determines the calibration scheme according to the characteristic model of the calibrated voltage divider, and generates the test point distribution, timing control parameters and safety strategy; The calibration scheme execution module automatically executes the calibration process based on the calibration scheme and outputs the standard value, calibrated value and environmental parameters of each test point; The error analysis and compensation module calculates the error information of each test point, obtains the independent compensation coefficient of each segment according to the segment characteristics in combination with temperature compensation, and performs compensation correction on the calibration results of the test points; The health status assessment module assesses abnormal health status and faults of the voltage divider and obtains a health status assessment result of the calibrated voltage divider; The report output module obtains the calibration conclusion including the health status assessment and generates a calibration report for visual display and data management.

Images