Transformer substation missing data preprocessing method based on regression analysis

Abstract

The invention discloses a transformer substation missing data preprocessing method based on regression analysis, and relates to the technical field of transformer substation missing data preprocessing, and the method comprises the following steps: carrying out the data analysis of filling data through an error evaluation index after the filling of an interpolation method is completed, and obtaining the matching degree of the filling data and real data; the matching degree of the filling data and the real data is compared with a preset deviation threshold value, and the level of the error propagation hidden danger is judged; setting a corresponding confidence interval according to the level of the error propagation hidden danger, and dynamically adjusting the confidence interval of the filling data based on a Bayesian algorithm; according to the dynamic adjustment result, the confidence degree of the data filling result is adjusted to be different levels, early warning processing is conducted on the filling data according to the levels of the confidence degree, and the method and device are used for solving the error propagation problem caused by data filling.

Description

Technical Field

[0001] The present invention relates to the technical field of regression analysis, and more specifically, to a method for preprocessing missing data of a substation based on regression analysis. Background Art

[0002] Substation missing data refers to real-time monitoring data that cannot be collected in the power system due to equipment failure, communication interruption or human error. These missing data may include key information such as voltage, current, temperature, equipment status, etc., which directly affects the safe operation and performance evaluation of the substation. The existence of missing data will not only lead to inaccurate analysis results, but also may affect decision-making and fault diagnosis. Therefore, it is crucial to adopt effective preprocessing methods, such as regression analysis, to fill these missing data to ensure the reliability and stability of the substation. Through reasonable data filling, data quality can be improved, thereby enhancing the overall operation efficiency and safety of the power system.

[0003] For continuous numerical data in substations, there are three commonly used interpolation methods (Newton interpolation, piecewise linear interpolation, and spline interpolation) and their applicable fault data types. These interpolation methods are used to fill in the gaps in key parameters such as temperature, humidity, and load caused by sensor anomalies or missing data, thereby ensuring the normal operation of the equipment. At the same time, other statistical methods such as frequency statistics are introduced for processing missing data, especially for data between unrelated devices.

[0004] In terms of intelligent threshold-based early warning, a dynamic fault threshold judgment standard is established by fitting the transformer winding temperature surface to reduce the false alarm rate of overheating faults. In addition, for gas pressure monitoring of GIS equipment, the importance of real-time monitoring is emphasized, and an intelligent early warning mechanism with floating standards is proposed in combination with historical data to improve the accuracy of fault early warning. These studies provide strong support for data management and safe operation of substations.

[0005] The above disclosed technical solution has at least the following technical problems: in the case of serious data missing, the extensive use of interpolation filling will introduce errors, and these errors may be amplified in the subsequent analysis and early warning process. This is particularly significant in the monitoring of key parameters such as temperature and humidity, which may affect fault warnings and thus affect the safety of the entire substation. How to balance the accuracy of data filling and the effectiveness of early warning is a key issue. In response to the above problems, the present invention proposes a solution. Summary of the invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a substation missing data preprocessing method based on regression analysis, which solves the error propagation problem caused by data filling by balancing the accuracy of data filling and the effectiveness of early warning.

[0007] To achieve the above object, the present invention provides the following technical solutions: The method for preprocessing missing data of a substation based on regression analysis comprises the following steps: after the interpolation method is completed, the filled data is analyzed using an error evaluation index to obtain the matching degree between the filled data and the real data; the matching degree between the filled data and the real data is compared with a preset deviation threshold to judge the level of the error propagation hidden danger; the corresponding confidence interval is set according to the level of the error propagation hidden danger, and the confidence interval of the filled data is dynamically adjusted based on the Bayesian algorithm; according to the result of the dynamic adjustment, the confidence of the data filling result is adjusted to different levels, and the filled data is early-warning processed according to the confidence level.

[0008] In a preferred embodiment, the error evaluation index performs data analysis on the filled data to obtain the matching degree between the filled data and the real data, specifically: constructing the filled data and the real data into time series respectively, calculating the distance between each point in the filled data and the real data, and using the Euclidean distance as the distance metric; constructing a cumulative distance matrix and initializing the points of the first cumulative distance matrix according to the first missing data, and calculating the cumulative distance values ​​of the remaining points through a recursive relationship; from the last point in the lower right corner of the cumulative distance matrix, backtracking in the x direction and the y direction in turn to the starting point of the cumulative distance matrix to obtain the backtracked cumulative distance matrix; and calculating the matching degree between the filled data and the real data according to the end point value of the backtracked cumulative distance matrix.

[0009] In a preferred embodiment, the cumulative distance matrix is ​​traced back in the x direction and the y direction to the starting point of the cumulative distance matrix in reverse order from the last point in the lower right corner of the cumulative distance matrix to obtain the cumulative distance matrix after backtracking, specifically: From , then backtrack once in the x direction; if From , then backtrack once in the y direction; if From , then backtrack once in the x and y directions.

[0010] In a preferred embodiment, the degree of match between the padded data and the real data is compared with a preset deviation threshold to determine the level of error propagation hidden dangers, specifically: by analyzing the error levels between the historical padded data and the real data, a deviation range is set; the degree of match between the padded data and the real data is compared with a preset deviation threshold; if the degree of match is higher than the threshold, the level of error propagation hidden danger is set to an acceptable level; if it is lower than the threshold and the difference is less than the standard deviation of the match, the level of error propagation hidden danger is set to a warning level; if it is lower than the threshold and the difference is greater than the standard deviation of the match, the level of error propagation hidden danger is set to a large error level.

[0011] In a preferred embodiment, the corresponding confidence interval is set according to the level of error propagation hidden danger, specifically: obtaining a standard confidence level, and adjusting the corresponding confidence level according to the level of error propagation hidden danger; and obtaining the confidence interval based on the confidence interval calculation under the normal distribution assumption.

[0012] In a preferred embodiment, the confidence interval of the filled-in data is dynamically adjusted based on the Bayesian algorithm, specifically: based on historical experience, a Bayesian prior distribution is set to describe the initial mean; after the interpolation method is completed, new observation data are gradually added to calculate the mean and variance of the filled-in data based on the prior distribution; after each new observation data is added, the calculation of the mean and variance is adjusted according to the posterior distribution to obtain a new mean and variance calculation formula, and then the confidence interval is dynamically adjusted.

[0013] In a preferred embodiment, the confidence of the data filling result is adjusted to different levels, and the filled data is pre-warned according to the confidence level, specifically: the dynamically adjusted confidence interval is compared with the preset error requirement threshold, and the confidence of the data filling result is divided into different levels; for high-confidence filled data, the system automatically confirms the accuracy of the data and directly uses these data in data analysis and prediction; for filled data with confidence levels, the system should take early warning measures. Additional monitoring of the filled data is carried out, and other relevant data sources are compared and verified; when the adjusted confidence level is still low, the system re-traces the previous filling results based on the new observation data, and automatically adjusts the set Bayesian prior distribution, and switches the linear interpolation method to the spatiotemporal interpolation method.

[0014] Technical effects and advantages of the substation missing data preprocessing method based on regression analysis of the present invention: 1. The present invention can effectively determine the potential risk level of error propagation by comparing the matching degree between the filled data and the real data with the preset deviation threshold. A reasonable deviation range is set by analyzing the error level of historical data to ensure the accuracy of the assessment. When the matching degree is higher than the threshold, the potential risk of error propagation is set to an acceptable level; if it is lower than the threshold but the difference is within the standard deviation, it is set to a warning level; if the difference exceeds the standard deviation, it is determined to be a large error level. This process improves the accuracy of error assessment of data filling results, which is helpful for early warning and risk control.

[0015] 2. The present invention can achieve more accurate risk management by dynamically adjusting the confidence of the data filling results and performing early warning processing accordingly. The filled data is divided into different levels according to the confidence level, so that high-confidence data can be quickly confirmed and put into use, thereby improving data utilization efficiency; for medium and low-confidence data, appropriate early warning measures are taken to ensure the reliability of data quality through additional monitoring and data verification. When the confidence level is too low, the system will automatically backtrack and adjust the data filling strategy to dynamically optimize the data processing process. This mechanism based on the Bayesian algorithm and switching between different interpolation methods ensures the flexibility and accuracy of the data and greatly reduces the risk of error propagation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a structural schematic diagram of the substation missing data preprocessing method based on regression analysis of the present invention. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0018] Embodiment 1, Figure 1 The present invention provides a substation missing data preprocessing method based on regression analysis, which includes the following steps: S1, after the interpolation filling is completed, the filled data is analyzed using the error evaluation index to obtain the matching degree between the filled data and the real data; In this embodiment, the error evaluation index performs data analysis on the filled data to obtain the matching degree between the filled data and the real data, specifically: The filled data and the real data are constructed into time series respectively, and the distance between each point in the filled data and the real data is calculated, and the Euclidean distance is used as the distance metric; Construct a cumulative distance matrix and initialize the points of the first cumulative distance matrix according to the first missing data, and calculate the cumulative distance values ​​of the remaining points through recursive relations; From the last point in the lower right corner of the cumulative distance matrix, trace back in the x direction and the y direction to the starting point of the cumulative distance matrix to obtain the cumulative distance matrix after backtracking; The matching degree between the filled data and the real data is calculated based on the endpoint value of the cumulative distance matrix after backtracking.

[0019] It should be noted that the optimal path is to trace back from the last point in the lower right corner of the cumulative distance matrix in the x and y directions to the starting point of the cumulative distance matrix. The optimal path can find the optimal time alignment between the filled data and the real data and minimize the error propagation between the two.

[0020] It should be noted that the calculation of each element in the cumulative distance matrix includes the minimum cumulative distance of all points on the path of the filled data and the real data.

[0021] It should be noted that the method of matching the fill-in data with the real data in this application has the following advantages: Time transformation flexibility: DTW can handle the offset and deformation problems of time series on the time axis, and is suitable for processing data with seasonal and periodic changes, especially meteorological data such as temperature and humidity.

[0022] Robustness: Due to the optimization of cumulative distance, DTW is not overly affected by single point outliers, making it more robust.

[0023] The specific calculation method of the distance matrix is:

[0024] The cumulative distance matrix is ​​constructed as follows:

[0025] In the formula, is the cumulative distance matrix, which represents the minimum cumulative distance from the starting point (1,1) to (i,j); is the distance matrix.

[0026] The cumulative distance matrix is ​​traced back from the last point in the lower right corner of the cumulative distance matrix in the x direction and the y direction to the starting point of the cumulative distance matrix, and the cumulative distance matrix after backtracking is obtained, which is specifically: like From , then backtrack once in the x direction; like From , then backtrack once in the y direction; like From , then backtrack once in the x and y directions.

[0027] The method for evaluating the matching degree between filled data and real data in this application provides a flexible and accurate method for evaluating the matching degree between interpolated filled data and real data by using dynamic time warping. This method has the following significant advantages in processing time series data, especially in the case of missing data: Temporal transformation flexibility: It can effectively handle the nonlinear deformation and offset of time series on the time axis, and is particularly suitable for data with seasonal or periodic changes, such as temperature, humidity and other meteorological data. This means that even if there is a certain time misalignment between the filled data and the real data, the optimal path can be found through flexible alignment, so as to more accurately evaluate the similarity between the two. This time alignment capability can reduce the error propagation problem caused by time asynchrony during the data filling process.

[0028] Robustness: In the calculation process, the influence of single outliers can be effectively handled by constructing a cumulative distance matrix. Even if some data points deviate significantly from the real data due to anomalies (such as noise, sudden failures, etc.), the optimal matching path of the two sequences can still be found by accumulating the minimum distance, thereby reducing the impact of single-point anomalies on the overall result. Compared with other matching evaluation methods, this robustness is more suitable for complex and changeable practical application scenarios.

[0029] Minimize error propagation: By constructing a cumulative distance matrix and calculating the minimum cumulative distance, the error propagation between the infill data and the real data is minimized on the optimal path. The selection of the optimal path follows the principle of minimum error accumulation, ensuring that the final evaluated matching result can truly reflect the global similarity between the infill data and the real data, without distorting the overall evaluation result due to local data deviation.

[0030] Dynamic adjustment and accuracy improvement: Since the matching degree between the filled data and the real data is obtained by back-tracing the cumulative distance matrix, this method can dynamically adjust the path to adapt to the errors caused by nonlinear changes in the actual data. This dynamic adjustment not only improves the accuracy of the matching degree evaluation after data filling, but also provides a reliable basis for subsequent data analysis and early warning.

[0031] By using Euclidean distance, the difference between each point in the infill data and the real data is simply and clearly measured, and an intuitive distance matrix is ​​constructed, making the error assessment of each point highly interpretable.

[0032] Construction of the cumulative distance matrix: The recursive relationship of the cumulative distance matrix fully considers the historical cumulative error on the path, so that each step of the decision is based on the principle of minimizing the error. Its construction method is not only efficient, but also ensures the controllability of the overall error and avoids excessive propagation of the error in the path.

[0033] Backtracking of the optimal path: The backtracking process from the lower right corner to the upper left corner of the cumulative distance matrix can find the optimal alignment path between the filled data and the real data, ensuring the accuracy of the matching evaluation results. Each step in the process is selected based on the minimum error accumulation to ensure that the final evaluation result can maximize the true credibility of the filled data.

[0034] S2, compare the matching degree between the filled data and the real data with the preset deviation threshold to determine the level of error propagation risk; In this embodiment, the matching degree between the padded data and the real data is compared with a preset deviation threshold to determine the level of the error propagation risk, specifically: By analyzing the error level between historical fill-in data and real data, the deviation range is set; Compare the matching degree between the filled data and the real data with the preset deviation threshold; If the matching degree is higher than the threshold, the level of error propagation potential is set to an acceptable level; If it is lower than the threshold and the difference is less than the standard deviation of the matching degree, the level of error propagation hidden danger is set to the warning level; If it is lower than the threshold and the difference is greater than the standard deviation of the matching degree, the level of error propagation hazard is set to a larger error level.

[0035] This application can effectively determine the level of potential risk of error propagation by comparing the degree of match between the filled data and the real data with the preset deviation threshold. A reasonable deviation range is set by analyzing the error level of historical data to ensure the accuracy of the assessment. When the degree of match is higher than the threshold, the potential risk of error propagation is set to an acceptable level; if it is lower than the threshold but the difference is within the standard deviation, it is set to a warning level; if the difference exceeds the standard deviation, it is determined to be a large error level. This process improves the accuracy of error assessment of data filling results, which helps with early warning and risk control.

[0036] S3, set the corresponding confidence interval according to the level of error propagation risk, and dynamically adjust the confidence interval of the filled data based on the Bayesian algorithm.

[0037] The corresponding confidence interval is set according to the level of error propagation hidden danger, specifically: Obtain the standard confidence level and adjust the corresponding confidence level according to the level of error propagation risk; The confidence interval is calculated based on the confidence interval under the normal distribution assumption.

[0038] The specific calculation formula of the confidence interval is as follows:

[0039] In the formula, is the confidence interval, To fill the mean of the data, is the critical value of the corresponding confidence level under the quasi-normal distribution, for, is the standard deviation of the matching degree, To fill the total amount of data.

[0040] It should be noted that the adjustment of the confidence level may vary with the actual situation. The default acceptable level is 0.5 times the standard confidence level, the warning level is 1.1 times the standard confidence level, and the larger error level is 2.2 times the standard confidence level.

[0041] The confidence interval of the filled data is dynamically adjusted based on the Bayesian algorithm, specifically: Based on historical experience, a Bayesian prior distribution is set to describe the initial mean; After the interpolation method is used to fill in the data, new observation data are gradually added to calculate the mean and variance of the filled data based on the prior distribution; Each time new observation data is added, the mean and variance are calculated according to the posterior distribution to obtain new mean and variance calculation formulas, and then the confidence interval is adjusted dynamically; The new mean calculation formula is as follows:

[0042] The specific calculation formula of the new variance is as follows:

[0043] In the formula, is the new mean, is the prior variance, is the preset initial hyperparameter, New variance.

[0044] The calculation formula for the dynamically adjusted confidence interval is as follows:

[0045] In the formula, To dynamically adjust the confidence interval.

[0046] This application sets the corresponding confidence interval according to the level of error propagation hidden dangers, and dynamically adjusts the confidence interval of the filled data in combination with the Bayesian algorithm. First, the confidence level is flexibly adjusted according to different levels of error hidden dangers to ensure that the confidence interval can reflect the real risk of data filling. Then, new observation data is gradually introduced through the Bayesian algorithm, and the confidence interval is dynamically adjusted in real time during the process of updating the posterior distribution. This method can not only effectively improve the accuracy of the confidence of the filled data, but also adaptively adjust when the data is continuously updated, to ensure the rationality and reliability of the prediction results, thereby significantly improving the accuracy of early warning and risk assessment.

[0047] S4, according to the result of dynamic adjustment, the confidence of the data filling result is adjusted to different levels, and the filled data is early-warning processed according to the confidence level.

[0048] The confidence level of the data filling result is adjusted to different levels, and early warning processing is performed on the filled data according to the confidence level, specifically: Compare the dynamically adjusted confidence interval with the preset error requirement threshold, and classify the confidence of the data filling result into different levels; For high-confidence imputed data, the system automatically confirms the accuracy of the data and directly uses the data in data analysis and prediction; For the filled data with a certain confidence level, the system should take early warning measures. Additional monitoring of the filled data and comparison and verification with other relevant data sources; When the adjusted confidence level is still low, the system will re-trace the previous filling results based on the new observation data, automatically adjust the set Bayesian prior distribution, and switch from linear interpolation to spatiotemporal interpolation.

[0049] This embodiment can achieve more accurate risk management by dynamically adjusting the confidence of the data filling results and performing early warning processing accordingly. The filled data is divided into different levels according to the confidence level, so that high-confidence data can be quickly confirmed and put into use, improving data utilization efficiency; for medium and low confidence data, appropriate early warning measures are taken to ensure the reliability of data quality through additional monitoring and data verification. When the confidence level is too low, the system will automatically backtrack and adjust the data filling strategy to dynamically optimize the data processing process. This mechanism based on the Bayesian algorithm and switching between different interpolation methods ensures the flexibility and accuracy of the data and greatly reduces the risk of error propagation.

[0050] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0051] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0052] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0053] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0054] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0055] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A substation missing data preprocessing method based on regression analysis, characterized in that: The steps include: After the interpolation filling is completed, the filled data is analyzed using the error evaluation index to obtain the matching degree between the filled data and the real data; Compare the matching degree between the filled data and the real data with the preset deviation threshold to determine the level of error propagation risk; The corresponding confidence interval is set according to the level of error propagation risk, and the confidence interval of the filled data is dynamically adjusted based on the Bayesian algorithm; According to the results of dynamic adjustment, the confidence of the data filling results is adjusted to different levels, and early warning processing is performed on the filled data according to the confidence level.

2. The method for preprocessing missing data of a substation based on regression analysis according to claim 1 is characterized in that: The error evaluation index performs data analysis on the filled data to obtain the matching degree between the filled data and the real data, specifically: The filled data and the real data are constructed into time series respectively, and the distance between each point in the filled data and the real data is calculated, and the Euclidean distance is used as the distance metric; Construct a cumulative distance matrix and initialize the points of the first cumulative distance matrix according to the first missing data, and calculate the cumulative distance values ​​of the remaining points through recursive relations; From the last point in the lower right corner of the cumulative distance matrix, trace back in the x direction and the y direction to the starting point of the cumulative distance matrix to obtain the cumulative distance matrix after backtracking; The matching degree between the filled data and the real data is calculated based on the endpoint value of the cumulative distance matrix after backtracking.

3. The method for preprocessing missing data of a substation based on regression analysis according to claim 2 is characterized in that: The cumulative distance matrix is ​​traced back from the last point in the lower right corner of the cumulative distance matrix in the x direction and the y direction to the starting point of the cumulative distance matrix, and the cumulative distance matrix after backtracking is obtained, which is specifically: like From , then backtrack once in the x direction; like From , then backtrack once in the y direction; like From , then backtrack once in the x and y directions.

4. The method for preprocessing missing data of a substation based on regression analysis according to claim 3 is characterized in that: The matching degree between the filled data and the real data is compared with the preset deviation threshold to determine the level of error propagation hidden danger, specifically: By analyzing the error level between historical fill-in data and real data, the deviation range is set; Compare the matching degree between the filled data and the real data with the preset deviation threshold; If the matching degree is higher than the threshold, the level of error propagation potential is set to an acceptable level; If it is lower than the threshold and the difference is less than the standard deviation of the matching degree, the level of error propagation hidden danger is set to the warning level; If it is lower than the threshold and the difference is greater than the standard deviation of the matching degree, the level of error propagation hazard is set to a larger error level.

5. The method for preprocessing missing data of substations based on regression analysis according to claim 4 is characterized in that: The corresponding confidence interval is set according to the level of error propagation hidden danger, specifically: Obtain the standard confidence level and adjust the corresponding confidence level according to the level of error propagation risk; The confidence interval is calculated based on the confidence interval under the normal distribution assumption.

6. The method for preprocessing missing data of a substation based on regression analysis according to claim 5 is characterized in that: The confidence interval of the filled data is dynamically adjusted based on the Bayesian algorithm, specifically: Based on historical experience, a Bayesian prior distribution is set to describe the initial mean; After the interpolation method is used to fill in the data, new observation data are gradually added to calculate the mean and variance of the filled data based on the prior distribution; Each time new observation data is added, the calculation of the mean and variance is adjusted according to the posterior distribution to obtain new mean and variance calculation formulas, and then the confidence interval is dynamically adjusted.

7. The method for preprocessing missing data of a substation based on regression analysis according to claim 6 is characterized in that: The confidence level of the data filling result is adjusted to different levels, and early warning processing is performed on the filled data according to the confidence level, specifically: Compare the dynamically adjusted confidence interval with the preset error requirement threshold, and classify the confidence of the data filling result into different levels; For high-confidence imputed data, the system automatically confirms the accuracy of the data and directly uses the data in data analysis and prediction; For the filled data with a certain confidence level, the system should take early warning measures. Additional monitoring of the filled data and comparison and verification with other relevant data sources; When the adjusted confidence level is still low, the system will re-trace the previous filling results based on the new observation data, automatically adjust the set Bayesian prior distribution, and switch from linear interpolation to spatiotemporal interpolation.

8. The method for preprocessing missing data of a substation based on regression analysis according to claim 7 is characterized in that: The specific calculation method of the distance matrix is: The cumulative distance matrix is ​​constructed as follows: In the formula, is the cumulative distance matrix, which represents the minimum cumulative distance from the starting point (1,1) to (i,j); is the distance matrix.

9. The method for preprocessing missing data of a substation based on regression analysis according to claim 8, characterized in that: The specific calculation formula of the confidence interval is as follows: In the formula, is the confidence interval, To fill the mean of the data, is the critical value of the corresponding confidence level under the quasi-normal distribution, for, is the standard deviation of the matching degree, To fill the total amount of data.

10. The method for preprocessing missing data of a substation based on regression analysis according to claim 9, characterized in that: The new mean calculation formula is as follows: The specific calculation formula of the new variance is as follows: In the formula, is the new mean, is the prior variance, is the preset initial hyperparameter, New variance. The calculation formula for the dynamically adjusted confidence interval is as follows: In the formula, To dynamically adjust the confidence interval.

Images