A high-voltage testing instrument detection data digitization identification system and method
By combining sliding window time series analysis, Kalman filtering algorithm, temporal convolutional neural network and adaptive fuzzy logic control algorithm, high-pressure test data is processed intelligently, solving the problems of low accuracy, poor real-time performance and poor adaptability of existing systems. It realizes efficient and intelligent data processing and fault early warning, and improves the safety and efficiency of test equipment.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- STATE GRID BEIJING ELECTRIC POWER CO
- Filing Date
- 2026-02-11
- Publication Date
- 2026-06-02
Smart Images
Figure CN122132909A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent data processing and optimized control technology, and in particular to a digital recognition system for detection data of high-voltage testing instruments. Background Technology
[0002] High-voltage testing instruments are widely used in industries such as power, electronics, and machinery, primarily for testing the pressure-bearing capacity, withstand voltage, and other key performance indicators of electrical equipment. With the continuous improvement of industrial automation, traditional high-voltage test data processing methods can no longer meet the needs of modern production. Traditional test data acquisition systems often rely on manual monitoring and single hardware interfaces, resulting in limited data processing methods, low efficiency, and susceptibility to human factors. Furthermore, data acquired in traditional systems typically requires manual intervention for analysis and identification, leading to inefficiency and a high risk of errors.
[0003] Existing high-voltage test data processing technologies typically employ traditional signal acquisition and processing techniques, such as digital signal processing algorithms and simple filtering and denoising techniques. While these methods can improve data clarity to some extent, they still fall short when dealing with complex test data. In particular, noise and abnormal fluctuations in the data often affect the accuracy of fault detection and parameter identification, thereby impacting the safety and reliability of the entire test process. Due to the lack of intelligent processing and analysis methods, traditional systems have weak capabilities in processing dynamic data under complex operating conditions, failing to detect potential fault modes in a timely and accurate manner, resulting in insufficient assurance of the accuracy and efficiency of test results.
[0004] Furthermore, existing high-pressure test data analysis typically relies on manually defined rules for parameter extraction and fault detection. These rules are often based on static, experience-driven models, which struggle to handle complex and variable test data, leading to delays or misjudgments in the system's identification of abnormal states. Even when some systems incorporate machine learning or data mining algorithms, many existing technologies still face the challenge of effectively integrating algorithmic models. Due to the dynamic changes in data during testing, existing system models fail to consider the specificities of real-world application scenarios and lack the ability to flexibly adjust, further reducing the system's accuracy and stability.
[0005] Although some existing technologies have begun to incorporate advanced techniques such as deep learning, fuzzy control, and genetic algorithms, their practical application results are not ideal. For example, in high-pressure testing, data fluctuations and noise are often complex, and existing algorithms have low accuracy in processing this data, leading to delays in fault detection and parameter optimization of testing equipment. Many deep learning-based models are prone to overfitting during training and lack sufficient generalization ability, resulting in insufficient robustness and adaptability in practical applications. Furthermore, existing multi-objective optimization methods, such as locally weighted regression algorithms, typically rely on manually set parameters and cannot adaptively adjust them in practical applications, leading to low response speed and control precision during testing.
[0006] Against this backdrop, existing high-pressure testing instrument data processing systems generally suffer from low accuracy, poor real-time performance, and poor adaptability. Traditional methods cannot provide a complete, multi-functional, intelligent solution and cannot effectively address the dynamic and complex monitoring needs during high-pressure testing. Therefore, how to establish a digital identification system for high-pressure testing instrument data to intelligently collect, analyze, optimize, and control test data during high-pressure testing has become a core problem that urgently needs to be solved by those skilled in the art. Summary of the Invention
[0007] The purpose of this invention is to propose a digital identification system and method for high-pressure testing instrument data. This invention intelligently analyzes and processes high-pressure testing data by combining sliding window time series analysis, Kalman filtering algorithm, temporal convolutional neural network, and adaptive fuzzy logic control algorithm. The system can acquire multi-dimensional data generated by high-pressure testing instruments in real time, and further optimize data accuracy through noise removal, anomaly detection, and key parameter extraction. By combining advanced data analysis and identification technologies, it can accurately identify abnormal patterns in the testing process and generate real-time early warning signals. Simultaneously, the system employs an improved local weighted regression algorithm for multi-objective optimization adjustment, achieving precise control and optimization decision-making of test parameters.
[0008] In a first aspect, the present invention provides a digital recognition system for detection data of high-pressure testing instruments, comprising: The data acquisition and conversion module is used to acquire analog signals generated by the high-voltage testing instrument in real time and convert the analog signals into digital signals through an analog-to-digital converter. The data preprocessing module is used to remove noise and detect anomalies in digital signals; The key parameter extraction module is used to extract key parameters related to high-voltage testing from the processed digital signal, and the accuracy is optimized by using sliding window time series analysis and Kalman filtering algorithm. The data analysis and identification module is used to combine temporal convolutional neural networks and adaptive fuzzy logic control algorithms to analyze and extract key parameters, identify abnormal patterns and issue warnings. The feedback adjustment module is used to perform multi-objective optimization adjustment based on the analysis results using an improved local weighted regression algorithm, adjust the experimental parameters in real time, and automatically generate optimization strategies. The display control module is used to display test data, analysis results, anomaly identification information and optimization suggestions in real time, and provides trend charts, heat maps, alarm information and automatically generates alarm prompts.
[0009] In a second aspect, the present invention provides a method for digitally identifying detection data from a high-pressure testing instrument, comprising: S1. Real-time acquisition of analog signals generated by high-voltage testing instruments, and conversion of analog signals into digital signals through an analog-to-digital converter; S2. Perform noise removal on the digital signal, detect anomalies, and filter valid data; S3. Extract key parameters related to the high-pressure test from the processed digital signal, perform local data aggregation through sliding window time series analysis, and use the Kalman filter algorithm to dynamically correct and optimize the accuracy of the data. S4. Combining temporal convolutional neural networks and adaptive fuzzy logic control algorithms, we perform in-depth analysis of the extracted key parameters, identify potential abnormal patterns, and generate early warning signals. S5. Based on the analysis results, the improved local weighted regression algorithm is applied to perform multi-objective optimization. The experimental parameters are adjusted in real time through the feedback mechanism and the optimization strategy is automatically generated. S6. Real-time display of test data, analysis results, anomaly identification information and optimization suggestions. Data changes are presented intuitively through trend charts and heat maps, and alarm prompts are automatically generated.
[0010] Optionally, step S3 includes the following specific steps: S31. Perform sliding window time series analysis on the processed digital signal, setting the time window as... The data is then aggregated within each window. First, the original signal within each window is defined, and then a weighted average method is used for data aggregation, as shown in the following formula:
[0011] in, The original signal, For the first The sampled value at time 10:00. For aggregate values, These are weighting coefficients; S32. Calculate the local trend characteristics of the data within the window, using the following formula:
[0012] in, The trend function of the time series is defined as the time series forecasting model. It reflects the local trend of change in the time series; S33. Input the sliding window aggregate value and trend characteristics into the Kalman filter algorithm, and set the system state equation as follows:
[0013] in, This is the system state vector at the current moment, containing several key parameters. Here is the state transition matrix. To control the input matrix, To control the input, For process noise, For process noise covariance; S34. Define the measurement equation as follows:
[0014] in, For the observed values, For the observation matrix, To observe noise, in accordance with ,in To measure the noise covariance; S35. Update the system state using the Kalman gain, and calculate the Kalman gain as follows:
[0015] in, The estimation error covariance matrix for the previous time step. To measure the noise covariance; S36. Update the state estimate as follows:
[0016] S37, Updated Status Corresponding to the current key parameter values for optimization After completing accuracy optimization, the key parameters are output after Kalman filter correction. .
[0017] Optionally, step S4 includes the following specific steps: S41. Perform multi-layer convolution operations, using network parameters... Update the convolution kernel to obtain the temporal features after convolution. The output is then processed using an activation function.
[0018] in, For activation function, The input features are those after Kalman filtering correction. For convolution kernel weights, For bias terms, The number of convolution kernels; S42. Features output by the convolutional network are controlled using a fuzzy logic algorithm. Perform fuzzy reasoning, construct the input and output of the fuzzy reasoning system, and define the fuzzy rule set. Each rule Expressed as: If is then output ; in, For input fuzzy sets, The output of the fuzzy control system is the inference output. Obtained through aggregation rules; S43, Output of the fuzzy inference system Deblurring is performed using the center-average method to obtain the early warning signal. The calculation formula is:
[0019] in, For fuzzy sets membership degree This is the output value; S44, if the warning signal Exceeding the set threshold This will trigger an abnormal warning and generate an abnormal alarm signal:
[0020] in, For the early warning results, This is the warning threshold; S45, The generated warning signal The output is sent to the control system to trigger the corresponding control mechanism, ensuring that any abnormalities during the test can be responded to and adjusted in a timely manner.
[0021] Optionally, step S5 includes the following specific steps: S51, Based on the warning signal in claim 4 and early warning results An improved local weighted regression algorithm was used to extract key parameters. Perform modeling and define the objective function. The formula is as follows:
[0022] in, For local weights, For the first One key parameter, For regression predictions, For parameter changes, The gradient of the key parameter, and The regularization coefficient is . The number of target parameters; S52. When performing locally weighted regression, use an adaptive weighting mechanism to adjust each key parameter. weight value And combine the feedback results from the previous moment. The weight update formula is as follows:
[0023] in, For the first Local variance of each parameter, It is a smoothing constant. For adjustment coefficients; S53. Combining the locally weighted regression results after the previous weight update, the experimental parameters are optimized using an algorithm. Adjustments are made, and optimization strategies are generated in real time, resulting in optimized experimental parameters. The calculation formula is:
[0024] in, These are the initial experimental parameters. For learning rate, The gradient of the objective function; S54, When the warning result Exceeding the set threshold At that time, an optimization adjustment formula based on conditional constraints is used to dynamically generate decision feedback signals. :
[0025] in, This is the final decision feedback signal, representing the adjusted experimental parameters; S55, through a feedback mechanism As the final adjusted test parameters This is further applied to the control of the experimental process to ensure the accuracy and efficiency of the experiment. The final decision model is as follows:
[0026] in, To adjust the coefficient, These are the experimental parameters for the final application.
[0027] The beneficial effects of this invention are: First, the high-pressure testing instrument detection data digital recognition system of this invention overcomes the limitations of traditional methods in dynamic data processing by combining sliding window time series analysis and Kalman filtering algorithm. Sliding window time series analysis can capture the time-varying characteristics of time series data, thus effectively processing multi-dimensional detection data, while the Kalman filtering algorithm performs precise correction during the dynamic updating process of the data, significantly improving the accuracy and stability of the data. This method effectively solves the problems of noise interference and outliers that may occur during the processing of high-pressure testing data, providing reliable data support for subsequent anomaly detection and fault diagnosis.
[0028] Secondly, this invention improves the accuracy and real-time performance of fault identification by combining a temporal convolutional neural network with an adaptive fuzzy logic control algorithm. The temporal convolutional neural network can efficiently process complex patterns in time-series data and automatically extract key features, while the adaptive fuzzy logic control algorithm dynamically adjusts model parameters based on real-time data, ensuring accurate fault warnings under various operating conditions. Through this combination, the system can identify potential abnormal modes in high-voltage testing instruments and issue warnings in real time, thus avoiding the problems of reliance on manual intervention and response lag in traditional methods, and improving the overall intelligence level of the monitoring system.
[0029] Finally, this invention employs an improved local weighted regression algorithm for multi-objective optimization, automatically generating optimization strategies while adjusting test parameters in real time. This optimization algorithm can dynamically adjust based on real-time monitoring data and historical operating data, ensuring not only maximum performance of the high-pressure testing instrument but also effectively reducing risks during the testing process. Combined with the display and control module, the system can display test data, analysis results, and anomaly identification information in real time, providing operators with trend charts, heat maps, and alarm information, further improving the visualization and automation level of the testing process, reducing the need for manual intervention, and enhancing the system's efficiency and safety. Attached Figure Description
[0030] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a simplified structural diagram of a high-pressure testing instrument detection data digital recognition system proposed in this invention.
[0031] Figure 2 This is a flowchart of a method for digitally identifying detection data from a high-pressure testing instrument, as proposed in this invention. Detailed Implementation
[0032] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.
[0033] Example 1 like Figure 1 As shown, a high-voltage testing instrument data digital recognition system includes: The data acquisition and conversion module is used to acquire analog signals generated by the high-voltage testing instrument in real time and convert the analog signals into digital signals through an analog-to-digital converter. The data preprocessing module is used to remove noise and detect anomalies in digital signals; The key parameter extraction module is used to extract key parameters related to high-voltage testing from the processed digital signal, and the accuracy is optimized by using sliding window time series analysis and Kalman filtering algorithm. The data analysis and identification module is used to combine temporal convolutional neural networks and adaptive fuzzy logic control algorithms to analyze and extract key parameters, identify abnormal patterns and issue warnings. The feedback adjustment module is used to perform multi-objective optimization adjustment based on the analysis results using an improved local weighted regression algorithm, adjust the experimental parameters in real time, and automatically generate optimization strategies. The display control module is used to display test data, analysis results, anomaly identification information and optimization suggestions in real time, and provides trend charts, heat maps, alarm information and automatically generates alarm prompts.
[0034] Example 2 like Figure 2 As shown, based on the same inventive concept as Embodiment 1, Embodiment 2 of the present invention also provides a method for digital identification of high-pressure testing instrument detection data, including the following steps: S1. Real-time acquisition of analog signals generated by high-voltage testing instruments, and conversion of analog signals into digital signals through an analog-to-digital converter; S2. Perform noise removal on the digital signal, detect anomalies, and filter valid data; S3. Extract key parameters related to the high-pressure test from the processed digital signal, perform local data aggregation through sliding window time series analysis, and use the Kalman filter algorithm to dynamically correct and optimize the accuracy of the data. S4. Combining temporal convolutional neural networks and adaptive fuzzy logic control algorithms, we perform in-depth analysis of the extracted key parameters, identify potential abnormal patterns, and generate early warning signals. S5. Based on the analysis results, the improved local weighted regression algorithm is applied to perform multi-objective optimization. The experimental parameters are adjusted in real time through the feedback mechanism and the optimization strategy is automatically generated. S6. Real-time display of test data, analysis results, anomaly identification information and optimization suggestions. Data changes are presented intuitively through trend charts and heat maps, and alarm prompts are automatically generated.
[0035] In this embodiment, step S3 includes the following specific steps: S31. Perform sliding window time series analysis on the processed digital signal, setting the time window as... The data is then aggregated within each window. First, the original signal within each window is defined, and then a weighted average method is used for data aggregation, as shown in the following formula:
[0036] in, The original signal, For the first The sampled value at time 10:00. For aggregate values, These are weighting coefficients; S32. Calculate the local trend characteristics of the data within the window, using the following formula:
[0037] in, The trend function of the time series is defined as the time series forecasting model. It reflects the local trend of change in the time series; S33. Input the sliding window aggregate value and trend characteristics into the Kalman filter algorithm, and set the system state equation as follows:
[0038] in, This is the system state vector at the current moment, containing several key parameters. Here is the state transition matrix. To control the input matrix, To control the input, For process noise, For process noise covariance; S34. Define the measurement equation as follows:
[0039] in, For the observed values, For the observation matrix, To observe noise, in accordance with ,in To measure the noise covariance; S35. Update the system state using the Kalman gain, and calculate the Kalman gain as follows:
[0040] in, The estimation error covariance matrix for the previous time step. To measure the noise covariance; S36. Update the state estimate as follows:
[0041] S37, Updated Status Corresponding to the current key parameter values for optimization After completing accuracy optimization, the key parameters are output after Kalman filter correction. .
[0042] In this embodiment, step S4 includes the following specific steps: S41. Perform multi-layer convolution operations, using network parameters... Update the convolution kernel to obtain the temporal features after convolution. The output is then processed using an activation function.
[0043] in, For activation function, The input features are those after Kalman filtering correction. For convolution kernel weights, For bias terms, The number of convolution kernels; S42. Features output by the convolutional network are controlled using a fuzzy logic algorithm. Perform fuzzy reasoning, construct the input and output of the fuzzy reasoning system, and define the fuzzy rule set. Each rule Expressed as: If is then output ; in, For input fuzzy sets, The output of the fuzzy control system is the inference output. Obtained through aggregation rules; S43, Output of the fuzzy inference system Deblurring is performed using the center-average method to obtain the early warning signal. The calculation formula is:
[0044] in, For fuzzy sets membership degree This is the output value; S44, if the warning signal Exceeding the set threshold This will trigger an abnormal warning and generate an abnormal alarm signal:
[0045] in, For the early warning results, This is the warning threshold; S45, The generated warning signal The output is sent to the control system to trigger the corresponding control mechanism, ensuring that any abnormalities during the test can be responded to and adjusted in a timely manner.
[0046] In this embodiment, step S5 includes the following specific steps: S51, Based on the warning signal in claim 4 and early warning results An improved local weighted regression algorithm was used to extract key parameters. Perform modeling and define the objective function. The formula is as follows:
[0047] in, For local weights, For the first One key parameter, For regression predictions, For parameter changes, The gradient of the key parameter, and The regularization coefficient is . The number of target parameters; S52. When performing locally weighted regression, use an adaptive weighting mechanism to adjust each key parameter. weight value And combine the feedback results from the previous moment. The weight update formula is as follows:
[0048] in, For the first Local variance of each parameter, It is a smoothing constant. For adjustment coefficients; S53. Combining the locally weighted regression results after the previous weight update, the experimental parameters are optimized using an algorithm. Adjustments are made, and optimization strategies are generated in real time, resulting in optimized experimental parameters. The calculation formula is:
[0049] in, These are the initial experimental parameters. For learning rate, The gradient of the objective function; S54, When the warning result Exceeding the set threshold At that time, an optimization adjustment formula based on conditional constraints is used to dynamically generate decision feedback signals. :
[0050] in, This is the final decision feedback signal, representing the adjusted experimental parameters; S55, through a feedback mechanism As the final adjusted test parameters This is further applied to the control of the experimental process to ensure the accuracy and efficiency of the experiment. The final decision model is as follows:
[0051] in, To adjust the coefficient, These are the experimental parameters for the final application.
[0052] Verification example: To verify the feasibility of this invention in practice, it was applied to the operation monitoring and fault detection of high-voltage electrical equipment in a power company. This company has long used high-voltage testing instruments for equipment testing and performance verification, but in practical applications, it often faces problems such as data noise interference, inaccurate outlier detection, and delayed fault warnings. Especially when conducting comprehensive analysis of long-term equipment operation data, traditional methods suffer from insufficient processing capacity and poor real-time performance.
[0053] This embodiment is applied at a high-voltage equipment maintenance site of a power company, specifically for the company's regular testing of high-voltage switchgear in its substations. This equipment undergoes multiple high-voltage tests annually, and maintenance personnel assess the equipment's health status based on the test data. Previously, this test data was primarily analyzed manually or processed using traditional algorithms, but the accuracy and timeliness of the analysis were significantly compromised. The high-voltage test instrument detection data digital recognition system of this invention, combined with sliding window time series analysis, Kalman filtering algorithms, and time-series convolutional neural networks, solves the accuracy and real-time issues of traditional methods in processing high-voltage test data.
[0054] In this scenario, the analog signals from the high-pressure testing instrument are first acquired in real time by a data acquisition and conversion module, and then converted into digital signals by an analog-to-digital converter. The digital signals then enter a data preprocessing module for noise removal and anomaly detection. Next, a key parameter extraction module extracts key parameters related to the high-pressure test, and a sliding window time series analysis method is used to perform local data aggregation to ensure that the time series characteristics of the data are not lost. Subsequently, a Kalman filter algorithm is used to dynamically correct these key parameters, improving the accuracy and stability of the data.
[0055] To further verify the advantages of this invention, during the experiment, the system processed the data from a high-voltage test of the substation in real time and compared it with traditional methods. The following is a comparison table of experimental data, which shows the differences in performance between the traditional method and the high-voltage test data processing technology of this invention at different stages.
[0056] Table 1 Comparison of High Voltage Experimental Data
[0057] As shown in Table 1, the high-voltage test data processing technology using the method of this invention demonstrates significant improvements in noise removal, anomaly detection, data accuracy, fault early warning, real-time data processing, optimization and adjustment, and equipment maintenance efficiency. For example, the accuracy of noise removal has increased from 85% to 98% compared to traditional methods, and the accuracy of anomaly detection has increased from 78% to 96%. Particularly in terms of fault early warning, this invention can provide an advance warning time of 4 hours, giving power companies sufficient time for maintenance and repair, effectively reducing downtime caused by equipment failures. Furthermore, equipment maintenance efficiency has also improved by 50%, significantly reducing manual intervention and maintenance costs.
[0058] During the experiment, the method of this invention enabled power companies to process and analyze high-voltage test data more efficiently and accurately in real time, identify potential faults in advance and take corresponding measures, avoiding the risk of failing to detect hidden problems in time as in traditional methods, and enhancing the safety and reliability of power equipment. After continuous trial operation and optimization, the system of this invention has been widely applied in different types of high-voltage equipment, and has achieved significant results in terms of equipment operation stability and maintenance efficiency, thus improving the overall operation and maintenance level of the power system.
[0059] In summary, the high-voltage testing instrument data digital identification system provided by this invention demonstrates superior performance compared to traditional methods in the detection, fault early warning, and maintenance optimization of high-voltage power equipment, providing an efficient, intelligent, and precise technical means for equipment management and operation and maintenance in the power industry.
[0060] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A digital recognition system for test data from a high-pressure testing instrument, characterized in that, include: The data acquisition and conversion module is used to acquire analog signals generated by the high-voltage testing instrument in real time and convert the analog signals into digital signals through an analog-to-digital converter. The data preprocessing module is used to remove noise and detect anomalies in digital signals; The key parameter extraction module is used to extract key parameters related to high-voltage testing from the processed digital signal, and the accuracy is optimized by using sliding window time series analysis and Kalman filtering algorithm. The data analysis and identification module is used to combine temporal convolutional neural networks and adaptive fuzzy logic control algorithms to analyze and extract key parameters, identify abnormal patterns and issue warnings. The feedback adjustment module is used to perform multi-objective optimization adjustment based on the analysis results using an improved local weighted regression algorithm, adjust the experimental parameters in real time, and automatically generate optimization strategies. The display control module is used to display test data, analysis results, anomaly identification information and optimization suggestions in real time, and provides trend charts, heat maps, alarm information and automatically generates alarm prompts.
2. A method for digitally identifying test data from a high-pressure testing instrument, characterized in that, include: S1. Real-time acquisition of analog signals generated by high-voltage testing instruments, and conversion of analog signals into digital signals through an analog-to-digital converter; S2. Perform noise removal on the digital signal, detect anomalies, and filter valid data; S3. Extract key parameters related to the high-pressure test from the processed digital signal, perform local data aggregation through sliding window time series analysis, and use the Kalman filter algorithm to dynamically correct and optimize the accuracy of the data. S4. Combining temporal convolutional neural networks and adaptive fuzzy logic control algorithms, we perform in-depth analysis of the extracted key parameters, identify potential abnormal patterns, and generate early warning signals. S5. Based on the analysis results, the improved local weighted regression algorithm is applied to perform multi-objective optimization. The experimental parameters are adjusted in real time through the feedback mechanism and the optimization strategy is automatically generated. S6. Real-time display of test data, analysis results, anomaly identification information and optimization suggestions. Data changes are presented intuitively through trend charts and heat maps, and alarm prompts are automatically generated.
3. The method for digitally identifying detection data from high-voltage testing instruments according to claim 2, characterized in that, S3 includes the following specific steps: S31. Perform sliding window time series analysis on the processed digital signal, setting the time window as... The data is aggregated within each window. First, the original signal within the window is defined, and then a weighted average method is used to aggregate the data. S32. Calculate the local trend characteristics of the data within the window; S33. Input the sliding window aggregate value and trend characteristics into the Kalman filter algorithm, and set the system state equation as follows: in, This is the system state vector at the current moment, containing several key parameters. Here is the state transition matrix. To control the input matrix, To control the input, For process noise, For process noise covariance; S34. Define the measurement equation as follows: in, For the observed values, For the observation matrix, To observe noise, in accordance with ,in To measure the noise covariance; S35. Update the system state using the Kalman gain, and calculate the Kalman gain as follows: in, The estimation error covariance matrix for the previous time step. To measure the noise covariance; S36. Update the state estimate as follows: S37, Updated Status Corresponding to the current key parameter values for optimization After completing accuracy optimization, the key parameters are output after Kalman filter correction. .
4. The method for digitally identifying detection data from high-voltage testing instruments according to claim 3, characterized in that, Sliding window time series analysis was performed on the processed digital signal, with the time window set as follows: The data is then aggregated within each window. First, the original signal within each window is defined, and then a weighted average method is used for data aggregation, as shown in the following formula: in, The original signal, For the first The sampled value at time 10:
00. For aggregate values, These are the weighting coefficients.
5. The method for digitally identifying test data from high-voltage testing instruments according to claim 4, characterized in that, The formula for calculating the local trend characteristics of data within a window is as follows: in, The trend function of the time series is defined as the time series forecasting model. It reflects the local variation trend of the time series.
6. The method for digitally identifying detection data from high-voltage testing instruments according to claim 2, characterized in that, S4 includes the following specific steps: S41. Perform multi-layer convolution operations, using network parameters... Update the convolution kernel to obtain the temporal features after convolution. The output is then processed using an activation function. S42. Features output by the convolutional network are controlled using a fuzzy logic algorithm. Perform fuzzy reasoning, construct the input and output of the fuzzy reasoning system, and define the fuzzy rule set. Each rule Expressed as: : If is then output ; in, For input fuzzy sets, The output of the fuzzy control system is the inference output. Obtained through aggregation rules; S43, Output of the fuzzy inference system Deblurring is performed using the center-average method to obtain the early warning signal. The calculation formula is: in, For fuzzy sets membership degree This is the output value; S44, if the warning signal Exceeding the set threshold This will trigger an abnormal warning and generate an abnormal alarm signal: in, For the early warning results, This is the warning threshold; S45, The generated warning signal The output is sent to the control system to trigger the corresponding control mechanism, ensuring that any abnormalities during the test can be responded to and adjusted in a timely manner.
7. The method for digitally identifying detection data from high-voltage testing instruments according to claim 6, characterized in that, The output is processed using an activation function: in, For activation function, The input features are those after Kalman filtering correction. For convolution kernel weights, For bias terms, This represents the number of convolution kernels.
8. The method for digitally identifying test data from high-voltage testing instruments according to claim 7, characterized in that, S5 includes the following specific steps: S51, Based on the warning signal in claim 4 and early warning results An improved local weighted regression algorithm was used to extract key parameters. Perform modeling and define the objective function. in, For local weights, For the first One key parameter, For regression predictions, For parameter changes, The gradient of the key parameter, and The regularization coefficient is . The number of target parameters; S52. When performing locally weighted regression, use an adaptive weighting mechanism to adjust each key parameter. weight value And combine the feedback results from the previous moment. ; S53. Combining the locally weighted regression results after the previous weight update, the experimental parameters are optimized using an algorithm. Adjustments are made, and optimization strategies are generated in real time, resulting in optimized experimental parameters. The calculation formula is: in, These are the initial experimental parameters. For learning rate, The gradient of the objective function; S54, When the warning result Exceeding the set threshold At that time, an optimization adjustment formula based on conditional constraints is used to dynamically generate decision feedback signals. : in, This is the final decision feedback signal, representing the adjusted experimental parameters; S55, through a feedback mechanism As the final adjusted test parameters This is further applied to the control of the experimental process to ensure the accuracy and efficiency of the experiment. The final decision model is as follows: in, To adjust the coefficient, These are the experimental parameters for the final application.
9. The method for digitally identifying detection data of high-voltage testing instruments according to claim 8, characterized in that, objective function The formula is as follows: in, For local weights, For the first One key parameter, For regression predictions, For parameter changes, The gradient of the key parameter, and The regularization coefficient is . The number of target parameters.
10. The method for digitally identifying detection data of high-voltage testing instruments according to claim 9, characterized in that, The weight update formula is as follows: in, For the first Local variance of each parameter, It is a smoothing constant. This is for adjusting the coefficient.