Early warning method and system for stability of water-turbine generator set and storage medium
Through the early warning method of water turbine generator sets based on big data, using Kolmogorov-Smirnov inspection and data preprocessing, the problem of low early warning accuracy in the existing technology is solved, and early detection of equipment abnormalities is achieved , improves the safety and reliability of the equipment.
Patent Information
- Application Number
- CN202411016637.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-29
- Publication Date
- 2024-12-06
AI Technical Summary
The existing stability warning methods for water turbine generator sets are not accurate, and equipment abnormalities cannot be detected early, resulting in the defects developing to a certain extent before they are issued.
Using a big data-based early warning method, data preprocessing and stability data screening are obtained by obtaining historical data, Kolmogorov-Smirnov test is used to determine whether the data meets the Burr distribution, and the median is calculated to set the warning threshold. , and extract features in real-time monitoring for early warning operations.
It realizes automated historical data cleaning and management, timely discovers potential problems of hydrowheel generator sets, improves the safety and reliability of the equipment, and can detect abnormalities in advance in the early stages of equipment defects.
Smart Images

Figure CN119089014A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of hydroelectric generator set operation data analysis, and in particular to an early warning method, system and storage medium for the stability of a hydroelectric generator set. Background Art
[0002] The stability parameters of hydro-turbine generator sets are key indicators that affect the safe operation of the units, and stability warnings can guide the safe operation of the units.
[0003] At present, the stability warning of hydro-turbine generator sets mostly adopts the general recommended value formed by the industry's many years of equipment operation and maintenance experience. The recommended value is generally a common value for multiple models in the industry. It is not very practical and targeted for specific units, and the accuracy of the warning is not high. Warnings are often issued only when equipment defects develop to a certain extent, and abnormalities cannot be detected in advance at the early stage of equipment defects. Summary of the invention
[0004] The main purpose of the present invention is to provide a method, system and storage medium for early warning of the stability of a hydro-turbine generator set, so as to solve the technical problems in the prior art that the early warning accuracy is not high and abnormalities cannot be detected in advance at the early stage of equipment defects.
[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a hydropower generator set stability warning method based on big data, comprising the following steps: S1: obtaining historical data of stability measurement points of all units of the same type in the power station under operating conditions within a preset first time period; S2: After data preprocessing and stability data screening of the historical data, distribution curves of each measurement point are drawn in turn according to the data of each stability measurement point, and the Kolmogorov-Smirnov test is used to determine whether each measurement point conforms to the Burr distribution; S3: Eliminate the measurement points that do not conform to the Burr distribution in the data obtained after the data preprocessing and stability data screening in S2, and save the eliminated data set as the first data set; S4: Calculate the median of each stability measurement point data in the first data set, and preset a first area warning threshold and a second area warning threshold corresponding to the stability measurement point according to the median; S5: within a preset second time period, obtaining stability measurement point data under the running state of the unit, and extracting features according to all features of the data in the first data set, and then performing the data preprocessing and stability data screening, and saving the data as a second data set, the features of the second data set are completely the same as those of the first data set; S6: Calculate the maximum value of the confidence level of the data of each stability measurement point in the second data set, and compare it with the first area warning threshold and the second area warning threshold of the corresponding measurement point to trigger a corresponding warning operation.
[0006] In a preferred embodiment, the data preprocessing comprises the following steps: A1: Null value processing: the method is to use spline interpolation to fill in the null value so that the data of each measuring point are aligned; A2: Dead value elimination. The method is: preset the number of dead value judgment points k. If the values of the measured point data do not change for more than k points in a row, the corresponding data are judged as dead values and eliminated.
[0007] In a preferred embodiment, the stability data screening comprises the following steps: B1: Thermal steady-state process screening: according to the unit temperature change curve, eliminate the data where the temperature does not reach a stable state during the unit startup and shutdown process; B2: Data fluctuation screening, the method is: Preset data fluctuation tracing time , select the measuring point , read the measuring point For each value in the data set, calculate the measurement point in turn At each time point in the data set Fluctuation value , the formula is: ; In the formula, For measuring point exist The data value at the moment, For measuring point In the time interval ( , ), For measuring point In the time interval ( , ) Setting volatility thresholds ,when When The data fluctuates too much. The data is concentrated at a certain time point. All data are removed.
[0008] Iterate through all the time points in the data set in sequence, and iterate through all the measurement points in the data set in sequence.
[0009] In the preferred embodiment, the calculation method of the maximum value after taking the confidence level of each stability measurement point in S6 is: pre-set the confidence ratio r, read the data of each stability measurement point in the data set, arrange the data of the same measurement point in order from large to small, eliminate the r×m data in the front of the sorted data sequence, and calculate as one if there are less than 1 data, and read the first value of the data sequence after elimination, which is the maximum value of the measurement point after taking the confidence level, where m is the total number of data points in the sorted data sequence.
[0010] In the preferred solution, the preset time span of the first time period is greater than 3 years, and the preset time span of the second time period is less than 1 month.
[0011] In a preferred embodiment, the warning threshold of the first area is smaller than the warning threshold of the second area.
[0012] In the preferred solution, the comparison with the first area warning threshold and the second area warning threshold of the corresponding measuring point in S6 to trigger the corresponding warning operation includes: When the maximum value of the confidence level of the stability measurement point in the second data set is less than the warning threshold of the first area, the corresponding stability measurement point is in a normal operating state; When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the warning threshold of the first area and less than the warning threshold of the second area, the corresponding stability measurement point is in a state of concern, triggering a first-level warning; When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the second area warning threshold, the corresponding stability measurement point is in a dangerous state and triggers a second-level warning.
[0013] A water turbine generator set stability early warning system, used to implement the water turbine generator set stability early warning method, comprising: A historical data module, used to obtain historical data of stability measurement points of all units of the same type in the power station under operating conditions within a preset first time period; A first data set module is used to perform data preprocessing and stability data screening on the historical data, and then draw the distribution curve of each measurement point in turn according to the data of each stability measurement point, and use the Kolmogorov-Smirnov test to determine whether each measurement point conforms to the Burr distribution, remove the measurement points that do not conform to the Burr distribution, and save the data set after the removal as the first data set; A warning threshold module, used to use a first data set, calculate the median of each stability measurement point data in the first data set, and preset a first area warning threshold and a second area warning threshold corresponding to the stability measurement point according to the median; A second data set module is used to obtain stability measurement point data under the running state of the unit within a preset second time period, and extract all features of the data in the first data set, perform data preprocessing and stability data screening, and save it as a second data set, and the features of the second data set are exactly the same as those of the first data set; The early warning module is used to calculate the maximum value of the confidence level of each stability measurement point data in the second data set, and compare it with the first area early warning threshold and the second area early warning threshold of the corresponding measurement point to trigger the corresponding early warning operation.
[0014] A computer-readable storage medium comprises: the storage medium stores a computer program, and when the computer program is executed by a processor, the stability early warning method of a hydro-generator set is executed.
[0015] The present invention provides an early warning method, system and storage medium for the stability of a hydro-turbine generator set, and relates to the technical field of hydro-turbine generator set operation data analysis. The method includes obtaining historical data, performing data preprocessing and stability data screening on the historical data, using the Kolmogorov-Smirnov test to determine whether each measuring point conforms to the Burr distribution, obtaining a first data set, and presetting a first area early warning threshold and a second area early warning threshold according to the median of each measuring point in the first data set; obtaining a second data set, calculating the maximum value of each measuring point in the second data set after taking the confidence, and comparing it with the first area early warning threshold and the second area early warning threshold, triggering a corresponding early warning operation. In this embodiment, historical data samples can be automatically cleaned and managed, and statistical data distribution laws can be observed. Combined with historical data analysis and real-time monitoring, potential problems of the hydro-turbine generator set can be discovered and solved in a timely manner, and equipment safety and reliability can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION
[0017] Example 1 like Figure 1 As shown, a method for early warning of stability of a hydro-generator set comprises the following steps: S1: within a preset first time period, obtain historical data of stability measurement points of all units of the same type in the power station under operating conditions.
[0018] S2: After preprocessing the historical data and screening the stability data, the distribution curve of each measurement point is drawn in turn according to the data of each stability measurement point, and the Kolmogorov-Smirnov test is used to determine whether each measurement point conforms to the Burr distribution.
[0019] S3: Eliminate the measurement points that do not conform to the Burr distribution in the data obtained after the data preprocessing and stability data screening in S2, and save the data set after elimination as the first data set.
[0020] S4: Calculate the median of the data of each stability measurement point in the first data set, and preset the first area warning threshold and the second area warning threshold of the corresponding stability measurement point according to the median.
[0021] S5: within a preset second time period, the stability measurement point data of the unit under the operating state is obtained, and after feature extraction based on all features of the data in the first data set, data preprocessing and stability data screening are performed, and the data is saved as a second data set, and the features of the second data set are exactly the same as those of the first data set.
[0022] S6: Calculate the maximum value of the confidence level of the data of each stability measurement point in the second data set, and compare it with the first area warning threshold and the second area warning threshold of the corresponding measurement point to trigger a corresponding warning operation.
[0023] This embodiment obtains historical data, performs data preprocessing and stability data screening on the historical data, and then performs Kolmogorov-Smirnov test on the processed data of each measuring point, removes the measuring points that do not conform to the Burr distribution, and saves the removed data as the first data set. The first area warning threshold and the second area warning threshold of the corresponding measuring point data are preset according to the median of the data of each measuring point in the first data set; then the maximum value of the data of each measuring point in the second data set after taking the confidence is calculated, and compared with the first area warning threshold and the second area warning threshold of the corresponding measuring point, triggering the corresponding warning operation. In this embodiment, the historical data samples can be automatically cleaned and managed, and the statistical data distribution law can be observed. Combined with historical data analysis and real-time monitoring, potential problems of the hydro-turbine generator set can be discovered and solved in a timely manner, and the safety and reliability of the equipment can be improved.
[0024] In this embodiment, the Kolmogorov-Smirnov test is a non-parametric test method used to test whether a sample comes from a specific probability distribution.
[0025] Burr distribution is a continuous probability distribution, and its probability density function is usually expressed in the following form: ; in, >0, >0, and is the shape parameter, >0 is the scale parameter.
[0026] In this embodiment, a distribution curve is drawn and a Kolmogorov-Smirnov test is performed to determine whether the data conforms to the Burr distribution, so as to perform a corresponding stability warning.
[0027] In the preferred solution, the preset time span of the first time period is greater than 3 years, and the preset time span of the second time period is less than 1 month.
[0028] The above settings ensure the integrity and periodicity of the data, making subsequent data predictions more accurate.
[0029] Data preprocessing and stability data screening in step S2 help remove noise and outliers and ensure the accuracy of subsequent analysis.
[0030] In the preferred embodiment, data preprocessing includes the following steps: A1: Null value processing: use spline interpolation to fill in the null values so that the data of each measuring point are aligned.
[0031] A2: Dead value elimination. The method is: preset the number of dead value judgment points k. If the values of the measured point data do not change for more than k points in a row, the corresponding data are judged as dead values and eliminated.
[0032] Null value processing uses spline interpolation to infer the possible values of missing values based on the trends of known data points, thereby filling in the missing data. Through null value processing, the filtered data has higher integrity and higher data reliability.
[0033] In dead value elimination, it is necessary to reasonably set the threshold of the dead value judgment point number k in order to remove the impact that may be caused by sensor failure or data collection errors and ensure the accuracy of subsequent data analysis.
[0034] In a preferred embodiment, the stability data screening comprises the following steps: B1: Thermal steady-state process screening: According to the temperature change curve of the unit, eliminate the data where the temperature does not reach a stable state during the unit startup and shutdown process.
[0035] B2: Data fluctuation screening, the method is: Preset data fluctuation tracing time , select the measuring point , read the measuring point For each value in the data set, calculate the measurement point in turn At each time point in the data set Fluctuation value , the formula is: ; In the formula, For measuring point exist The data value at the moment, For measuring point In the time interval ( , ), For measuring point In the time interval ( , ) is the standard deviation in .
[0036] Setting volatility thresholds ,when When The data fluctuates too much. The data is concentrated at a certain time point. All data are removed.
[0037] Iterate through all the time points in the data set in sequence and traverse all the measurement points in the data set in sequence.
[0038] Thermal steady-state process screening is to ensure that only data from the unit in a stable state is included in the analysis. By observing the temperature change curve of the unit, data from the unit that did not reach a stable temperature during the start-up and shutdown process are eliminated, eliminating the interference of non-steady-state data on the analysis results.
[0039] Data fluctuation screening is to remove noise and outliers in the data, retain stable data for analysis, and improve the accuracy and reliability of data analysis. In this embodiment, t=60s, =5, use the above method to traverse all the measurement points in the data set in turn, and eliminate the data at all time points with excessive fluctuations.
[0040] In the preferred solution, the calculation method of the maximum value of the confidence of each stability measurement point in step S6 is: preset confidence ratio , read the data of each stability measurement point in the data set, arrange the data of the same measurement point in order from large to small, and remove the data at the front in the sorted data sequence. If there are less than 1 data, it will be counted as 1. The first value of the data sequence after elimination is read, which is the maximum value of the confidence level of the measuring point. is the total number of data points in the sorted data sequence.
[0041] In this embodiment, the confidence ratio is preset =0.1%, and remove 0.1% of the data in each power range from the smallest end of each stability measurement point data.
[0042] Arrange the data of the same measuring point in descending order, remove the first r×m data in the sorted data sequence, and count as one if there are less than one. Read the first value of the data sequence after elimination, which is the maximum value of the measuring point after taking the confidence level, where m is the total number of data points in the sorted data sequence. In this embodiment, m=20000, then the first 20 points are eliminated.
[0043] In a preferred embodiment, the warning threshold of the first area is smaller than the warning threshold of the second area.
[0044] In this embodiment, the first area warning threshold , the second area warning threshold , thereby dividing the multi-level warning interval, and a represents the median of the corresponding measuring point.
[0045] Step S6 is to judge the stability measurement point data and perform corresponding early warning operations.
[0046] In the preferred solution, the comparison with the first area warning threshold and the second area warning threshold of the corresponding measuring point in step S6 triggers the corresponding warning operation, including: When the maximum value of the confidence level of the stability measurement point in the second data set is less than the warning threshold of the first area, the corresponding stability measurement point is in a normal operating state.
[0047] When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the warning threshold of the first area and less than the warning threshold of the second area, the corresponding stability measurement point is in a state of concern and triggers a first-level warning.
[0048] When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the second area warning threshold, the corresponding stability measurement point is in a dangerous state and triggers a second-level warning.
[0049] This method was run on real-time data for one year, during which the upper guide swing and upper frame vibration generated 6 first-level warnings and 2 second-level warnings, while the traditional monitoring system of the power station did not trigger the warning. During the unit maintenance, the upper frame and the upper guide bearing were inspected and loose supports were found. This shows that the early warning method of this embodiment can detect abnormalities in advance at an early stage of equipment defects.
[0050] Example 2 In combination with Example 1, a stability early warning system for a hydro-turbine generator set is further described, and is used to implement a stability early warning method for a hydro-turbine generator set in Example 1, comprising: The historical data module is used to obtain historical data of stability measurement points of all units of the same type in the power station under operating conditions within a preset first time period.
[0051] The first data set module is used to perform data preprocessing and stability data screening on historical data, and then draw the distribution curve of each measurement point in turn according to the data of each stability measurement point, and use the Kolmogorov-Smirnov test to determine whether each measurement point conforms to the Burr distribution, eliminate the measurement points that do not conform to the Burr distribution, and save the data set after elimination as the first data set.
[0052] The warning threshold module is used to use the first data set and calculate the median of each stability measurement point data in the first data set, and preset the first area warning threshold and the second area warning threshold of the corresponding stability measurement point according to the median.
[0053] The second data set module is used to obtain the stability measurement point data under the unit operation state within a preset second time period, and after extracting features based on all the features of the data in the first data set, perform data preprocessing and stability data screening, and save it as the second data set. The features of the second data set are exactly the same as those of the first data set.
[0054] The early warning module is used to calculate the maximum value of the confidence level of each stability measurement point data in the second data set, and compare it with the first area early warning threshold and the second area early warning threshold of the corresponding measurement point to trigger the corresponding early warning operation.
[0055] The system of this embodiment uses historical data to calculate the warning threshold for stability warning, which can timely discover and solve potential problems of the hydro-generator set and improve equipment safety and reliability.
[0056] Example 3 In combination with Example 1, a computer-readable storage medium includes: the storage medium stores a computer program, and when the computer program is executed by a processor, the early warning method for stability of a hydro-turbine generator set in Example 1 is executed.
[0057] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.
Claims
1. A method for early warning of stability of a hydro-generator set, characterized in that: The following steps are involved: S1: obtaining historical data of stability measurement points of all units of the same type in the power station under operating conditions within a preset first time period; S2: After data preprocessing and stability data screening of the historical data, distribution curves of each measurement point are drawn in turn according to the data of each stability measurement point, and the Kolmogorov-Smirnov test is used to determine whether each measurement point conforms to the Burr distribution; S3: Eliminate the measurement points that do not conform to the Burr distribution in the data obtained after the data preprocessing and stability data screening in S2, and save the eliminated data set as the first data set; S4: Calculate the median of each stability measurement point data in the first data set, and preset a first area warning threshold and a second area warning threshold corresponding to the stability measurement point according to the median; S5: within a preset second time period, obtaining stability measurement point data under the running state of the unit, and extracting features according to all features of the data in the first data set, and then performing the data preprocessing and stability data screening, and saving the data as a second data set, the features of the second data set are completely the same as those of the first data set; S6: Calculate the maximum value of the confidence level of the data of each stability measurement point in the second data set, and compare it with the first area warning threshold and the second area warning threshold of the corresponding measurement point to trigger a corresponding warning operation.
2. The early warning method for the stability of a hydro-generator set according to claim 1 is characterized in that: The data preprocessing comprises the following steps: A1: Null value processing: the method is to use spline interpolation to fill in the null value so that the data of each measuring point are aligned; A2: Dead value elimination. The method is: preset the number of dead value judgment points k. If the values of the measured point data do not change for more than k points in a row, the corresponding data are judged as dead values and eliminated.
3. The early warning method for the stability of a hydro-generator set according to claim 2, characterized in that: The stability data screening comprises the following steps: B1: Thermal steady-state process screening: according to the unit temperature change curve, eliminate the data where the temperature does not reach a stable state during the unit startup and shutdown process; B2: Data fluctuation screening, the method is: Preset data fluctuation tracing time , select the measuring point , read the measuring point For each value in the data set, calculate the measurement point in turn At each time point in the data set Fluctuation value , the formula is: ; In the formula, For measuring point exist The data value at the moment, For measuring point In the time interval ( , ), For measuring point In the time interval ( , ) Setting volatility thresholds ,when When The data fluctuates too much. The data is concentrated at a certain time point. All data are removed. Iterate through all the time points in the data set in sequence, and iterate through all the measurement points in the data set in sequence.
4. The early warning method for the stability of a hydro-generator set according to claim 1, characterized in that: The calculation method of the maximum value after taking the confidence level of each stability measurement point in S6 is: Preset the confidence ratio r, read the data of each stability measurement point in the data set, arrange the data of the same measurement point in descending order, eliminate the first r×m data in the sorted data sequence, and count as one if there are less than 1 data. Read the first value of the data sequence after elimination, which is the maximum value of the measurement point after taking the confidence, where m is the total number of data points in the sorted data sequence.
5. The early warning method for the stability of a hydro-generator set according to claim 1, characterized in that: The preset time span of the first time period is greater than 3 years, and the preset time span of the second time period is less than 1 month.
6. The early warning method for the stability of a hydro-generator set according to claim 5, characterized in that: The first area warning threshold is less than the second area warning threshold.
7. The early warning method for the stability of a hydro-generator set according to claim 6, characterized in that: The comparison in S6 with the first area warning threshold and the second area warning threshold of the corresponding measuring point to trigger the corresponding warning operation includes: When the maximum value of the confidence level of the stability measurement point in the second data set is less than the warning threshold of the first area, the corresponding stability measurement point is in a normal operating state; When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the warning threshold of the first area and less than the warning threshold of the second area, the corresponding stability measurement point is in a state of concern, triggering a first-level warning; When the maximum value of the confidence level of the stability measurement point in the second data set is greater than the second area warning threshold, the corresponding stability measurement point is in a dangerous state and triggers a second-level warning.
8. An early warning system for the stability of a hydro-generator set, characterized in that: The method for realizing the stability early warning of a hydro-generator set as claimed in claim 1 comprises: A historical data module, used to obtain historical data of stability measurement points of all units of the same type in the power station under operating conditions within a preset first time period; A first data set module is used to perform data preprocessing and stability data screening on the historical data, and then draw the distribution curve of each measurement point in turn according to the data of each stability measurement point, and use the Kolmogorov-Smirnov test to determine whether each measurement point conforms to the Burr distribution, remove the measurement points that do not conform to the Burr distribution, and save the data set after the removal as the first data set; A warning threshold module, used to use a first data set, calculate the median of each stability measurement point data in the first data set, and preset a first area warning threshold and a second area warning threshold corresponding to the stability measurement point according to the median; A second data set module is used to obtain stability measurement point data under the running state of the unit within a preset second time period, and perform feature extraction based on all features of the data in the first data set, and then perform data preprocessing and stability data screening, and save it as a second data set, and the features of the second data set are exactly the same as those of the first data set; The early warning module is used to calculate the maximum value of the confidence level of each stability measurement point data in the second data set, and compare it with the first area early warning threshold and the second area early warning threshold of the corresponding measurement point to trigger the corresponding early warning operation.
9. A computer-readable storage medium, characterized in that: include: The storage medium stores a computer program, and when the computer program is executed by the processor, the early warning method for the stability of the hydro-generator set described in claims 1-7 is executed.
