CN113918624B - Turbine through-flow abnormity early warning method based on big data analysis - Google Patents

Turbine through-flow abnormity early warning method based on big data analysis Download PDF

Info

Publication number
CN113918624B
CN113918624B CN202111519717.3A CN202111519717A CN113918624B CN 113918624 B CN113918624 B CN 113918624B CN 202111519717 A CN202111519717 A CN 202111519717A CN 113918624 B CN113918624 B CN 113918624B
Authority
CN
China
Prior art keywords
data
data set
cylinder efficiency
real
cylinder
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202111519717.3A
Other languages
Chinese (zh)
Other versions
CN113918624A (en
Inventor
邵建宇
袁伟中
解剑波
吴斌
孙永平
张震伟
屠海彪
姜志锋
傅骏伟
郭鼎
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang Zheneng Taizhou No2 Power Generation Co ltd
Zhejiang Energy Group Research Institute Co Ltd
Original Assignee
Zhejiang Zheneng Taizhou No2 Power Generation Co ltd
Zhejiang Energy Group Research Institute Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang Zheneng Taizhou No2 Power Generation Co ltd, Zhejiang Energy Group Research Institute Co Ltd filed Critical Zhejiang Zheneng Taizhou No2 Power Generation Co ltd
Priority to CN202111519717.3A priority Critical patent/CN113918624B/en
Publication of CN113918624A publication Critical patent/CN113918624A/en
Application granted granted Critical
Publication of CN113918624B publication Critical patent/CN113918624B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/245—Query processing
    • G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
    • G06F16/2462—Approximate or statistical queries
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01—MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01D—NON-POSITIVE DISPLACEMENT MACHINES OR ENGINES, e.g. STEAM TURBINES
    • F01D21/00—Shutting-down of machines or engines, e.g. in emergency; Regulating, controlling, or safety means not otherwise provided for
    • F01D21/003—Arrangements for testing or measuring
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/24—Querying
    • G06F16/248—Presentation of query results

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Databases & Information Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computational Linguistics (AREA)
  • Fuzzy Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Mechanical Engineering (AREA)
  • Quality & Reliability (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Control Of Turbines (AREA)

Abstract

The invention relates to a turbine through-flow abnormity early warning method based on big data analysis, which comprises the following steps: collecting historical data of selected measuring points, setting upper and lower limits of normal values of the data of the selected measuring points, and cleaning the data by using a box line graph to obtain a normal data set; calculating the cylinder efficiency of each piece of data in the normal data set one by one, and putting the calculated cylinder efficiency into the data; and removing abnormal values of the cylinder efficiency values obtained by calculation to obtain an integrated data set. The invention has the beneficial effects that: according to the method, a turbine cylinder efficiency fitting surface model is constructed from multiple dimensions through a computer big data analysis technology based on historical data and steady state identification, the change of the cylinder efficiency is amplified through a layering and scoring method, the abnormity of the through-flow part of the turbine is early warned, the early problems of scaling of the through-flow part, damage of a steam inlet filter screen and the like are found, and sufficient intervention time is provided for the fault treatment of the turbine.

Description

Turbine through-flow abnormity early warning method based on big data analysis
Technical Field
The invention belongs to the field of turbine fault early warning diagnosis, and particularly relates to a turbine through-flow abnormity early warning method based on big data analysis.
Background
The through-flow part of the steam turbine is a core component for converting steam heat energy into work, and the perfection degree of the through-flow part has important influence on the energy consumption level of a unit. When the performance of the through-flow part of the steam turbine deviates from a normal value (such as corrosion and scaling of the through-flow part, abrasion of a steam seal, breakage of a filter screen, breakage of a blade and the like), the thermal characteristic of the steam turbine set is inevitably changed, and the change of the thermal characteristic of the steam turbine set is reflected on the efficiency of the cylinder.
The design value of the cylinder efficiency (or the test value obtained by the test of scientific research institutes) given by a steam turbine manufacturer only relates to a plurality of working conditions generally, the operating working condition of the unit often deviates from the test working condition, and the cylinder efficiency can synchronously change along with the change of the opening degree and the load of the high-pressure regulating valve, so that the design value or the test value of the cylinder efficiency can not reflect the change of the through-flow part in the whole operating process of the unit.
At present, a mature algorithm for calculating a cylinder efficiency evaluation value based on historical data has poor sensitivity to small-amplitude fluctuation of the efficiency of a steam turbine cylinder, and the through-flow fault of the steam turbine cannot be found in time to influence the running safety of a unit, so that early warning of the through-flow fault of the steam turbine is significant for improving the thermal economy and safety of the steam turbine.
Disclosure of Invention
The invention aims to overcome the defects in the prior art and provides a turbine through-flow abnormity early warning method based on big data analysis.
The turbine through-flow abnormity early warning method based on big data analysis comprises the following steps:
step 1, collecting historical data of selected measuring points, setting upper and lower limits of normal values of the data of the selected measuring points, and cleaning the data by using a box line graph to obtain a normal data set
Figure 49083DEST_PATH_IMAGE001
The expert uses the trend graph to align the normal data set
Figure 565515DEST_PATH_IMAGE002
The data in the step (1) is manually and finely screened to obtain a normal data set
Figure 140853DEST_PATH_IMAGE003
；
Step 2, calculating normal data sets item by item
Figure 493468DEST_PATH_IMAGE004
The cylinder efficiency of each piece of data is calculated, and the calculated cylinder efficiency is put into the data; removing abnormal values of the cylinder efficiency values obtained through calculation to obtain an integrated data set
Figure 78033DEST_PATH_IMAGE005
And performing steady state identification according to the cylinder efficiency value through sliding a time window to obtain a steady state data set
Figure 448972DEST_PATH_IMAGE006
；
Step 3, setting the steady state data set
Figure 460790DEST_PATH_IMAGE007
Dividing according to the unit load and the opening of the regulating valve to obtain a divided data set
Figure 527798DEST_PATH_IMAGE008
(ii) a Then each divided data set is
Figure 916054DEST_PATH_IMAGE009
Performing equidistant layering according to the distance from the average value of the cylinder efficiency of the segmentation data set, and combining the data of the same layer into a layered data set
Figure 141499DEST_PATH_IMAGE010
(ii) a Respectively constructing cylinder efficiency fitting surface models corresponding to each layer of data, wherein the number of the surface layers of the fitting surface models is the same as that of the surface layers after equidistant layering; determining coefficient R of the obtained fitted surface model2If the fitting effect of the curved surface is smaller than the set value, judging that the fitting effect of the curved surface does not reach the standard, returning to the step 2, removing abnormal values of the calculated cylinder efficiency value, and making a decision coefficient R2Is closer to 1 (e.g., the original set value is 0.9, and the new set value is modified to 0.95) until the coefficient R is determined2Is greater than a set value;
step 4, collecting real-time data, wherein the collected data can also pass through a custom time window to obtain a real-time data set;
step 5, calculating the cylinder efficiency of each piece of data in the collected real-time data set one by one according to the step 2, and putting the calculated cylinder efficiency into the data to obtain a real-time integrated data set;
step 6, scoring the real-time integration data set; when the real-time integration data set is normal, the cylinder efficiency in each data falls in the top layer and the bottom layer of the fitted surface model obtained in the step 3; setting a score for each curved surface, wherein the real-time integration data corresponding to the cylinder efficiency falling into the curved surface is the score of the curved surface, the real-time integration data corresponding to the cylinder efficiency not falling onto the curved surface is the score of the curved surface closest to the curved surface, the real-time integration data corresponding to the cylinder efficiency exceeding the range of the curved surface is the score of the boundary curved surface, and counting the total score of all pieces of data in the real-time integration data set;
and 7, analyzing whether the cylinder efficiency is abnormally changed: collecting the total scores of all the data in the real-time integration data set obtained by statistics in the step 6 into a real-time integration data set score trend graph, and analyzing the score change trend; when the score rises to reach an upper threshold or falls to reach a lower threshold and lasts for a period of time, judging that the cylinder efficiency is abnormal; if the cylinder efficiency is abnormal after the cylinder efficiency is judged to be abnormal, triggering an alarm, and setting the upper and lower score thresholds and the triggering alarm times by professionals.
Preferably, step 1 specifically comprises the following steps:
step 1.1, collecting unit load, valve opening degree, steam inlet pressure, steam inlet temperature, steam exhaust pressure and steam exhaust temperature of a selected measuring point;
step 1.2, data cleaning is carried out: setting upper and lower limits of normal data values of selected measuring points, and eliminating data exceeding the upper and lower limits of normal data values (unqualified quality) in the historical data of the measuring points acquired in the step 1.1 by utilizing a box diagram to obtain a normal data set
Figure 324218DEST_PATH_IMAGE011
Step 1.3, in order to guarantee the data quality, the steam turbine professional utilizes the trend chart to align the normal data set
Figure 651426DEST_PATH_IMAGE012
Then fine screening is carried out to obtain a normal data set
Figure 577793DEST_PATH_IMAGE013
。
Preferably, step 2 specifically comprises the following steps:
step 2.1, calculate the normal data set item by item
Figure 454482DEST_PATH_IMAGE014
The cylinder efficiency of each piece of data is calculated, and the calculated cylinder efficiency is put into the data, wherein the calculation formula of the cylinder efficiency is as follows:
Figure 322950DEST_PATH_IMAGE015
in the above formula, the first and second carbon atoms are,
Figure 121142DEST_PATH_IMAGE016
in order to be able to achieve a cylinder efficiency,
Figure 585621DEST_PATH_IMAGE017
which represents the enthalpy of the incoming steam,
Figure 785659DEST_PATH_IMAGE018
the expression of the enthalpy of the exhaust steam,
Figure 326492DEST_PATH_IMAGE019
representing the isentropic enthalpy found by finding a water vapor enthalpy entropy diagram from the steam inlet entropy and the steam outlet pressure;
step 2.2, eliminating abnormal values of the cylinder efficiency values obtained through calculation by utilizing the boxplot to obtain an integrated data set
Figure 346401DEST_PATH_IMAGE020
；
Step 2.3, screening out stable data in a set time length in the integrated data set to obtain a stable data set
Figure 880151DEST_PATH_IMAGE021
: sliding the time window according to the set time length to integrate the data set
Figure 934694DEST_PATH_IMAGE022
Separated into n lists
Figure 613806DEST_PATH_IMAGE023
Screening out given values of cylinder efficiency fluctuation smaller than fluctuation amplitude in each list
Figure 386590DEST_PATH_IMAGE024
The cylinder efficiency value is eliminated, and the cylinder efficiency fluctuation in each list is larger than the given value of fluctuation amplitude
Figure 724030DEST_PATH_IMAGE024
Merging the screened residual cylinder efficiency values into a new list to obtain a steady state data set
Figure 367501DEST_PATH_IMAGE025
(ii) a The screening formula is as follows:
Figure 244639DEST_PATH_IMAGE027
in the above formula, the first and second carbon atoms are,
Figure 120191DEST_PATH_IMAGE028
represents the maximum value of cylinder efficiency in the selected list,
Figure 149327DEST_PATH_IMAGE029
represents the minimum value of cylinder efficiency in the selected list,
Figure 435820DEST_PATH_IMAGE030
represents the average of the cylinder efficiencies in the selected list,
Figure 183197DEST_PATH_IMAGE031
representing a given value of fluctuation amplitude.
Preferably, step 3 specifically comprises the following steps:
step 3.1, set the steady state data
Figure 393598DEST_PATH_IMAGE032
Dividing according to the unit load and the opening of the regulating valve collected in the step 1 to obtain n divided data sets
Figure 480503DEST_PATH_IMAGE033
The set of all the segmented data sets is
Figure 236100DEST_PATH_IMAGE034
Wherein
Figure 408456DEST_PATH_IMAGE035
Representing the nth divided data set
Figure 422548DEST_PATH_IMAGE036
(ii) a Abandon
Figure 415824DEST_PATH_IMAGE037
Segmented data set with fewer than 50 data strips
Figure 60432DEST_PATH_IMAGE038
；
Step 3.2, divide each data set into
Figure 251242DEST_PATH_IMAGE038
Performing equidistant layering according to the distance from the average value of the cylinder efficiency of the segmentation data set, and combining the data of the same layer into a layered data set
Figure 6708DEST_PATH_IMAGE039
The set of all hierarchical data sets is
Figure 881254DEST_PATH_IMAGE040
The combination rule is as follows:
Figure 696764DEST_PATH_IMAGE041
in the above formula, the first and second carbon atoms are,
Figure 109290DEST_PATH_IMAGE042
represents the m-th level of the hierarchical data set,
Figure 668448DEST_PATH_IMAGE043
representing the 1 st segmented data set
Figure 427194DEST_PATH_IMAGE044
The data of the m-th layer in (1),
Figure 413605DEST_PATH_IMAGE045
representing the 2 nd segmented data set
Figure 579007DEST_PATH_IMAGE046
The data of the m-th layer in the (C),
Figure 427008DEST_PATH_IMAGE047
representing the nth divided data set
Figure 259835DEST_PATH_IMAGE035
The mth layer of data;
3.3, carrying out nonlinear fitting on each layer of hierarchical data set according to three dimensions of cylinder efficiency, opening of a regulating valve and unit load: regulating valve opening with cylinder efficiency as a dependent variable
Figure 417147DEST_PATH_IMAGE048
Load of the machine set
Figure 69845DEST_PATH_IMAGE049
As independent variable, pair
Figure 220073DEST_PATH_IMAGE050
、
Figure 907406DEST_PATH_IMAGE049
Carrying out third-order dimension expansion, wherein the third-order dimension expansion calculation formula is as follows:
Figure 235619DEST_PATH_IMAGE051
；
Figure 375613DEST_PATH_IMAGE052
；
Figure 565417DEST_PATH_IMAGE053
；
Figure 107257DEST_PATH_IMAGE054
；
Figure 340792DEST_PATH_IMAGE055
；
Figure 968083DEST_PATH_IMAGE056
；
Figure 460113DEST_PATH_IMAGE057
obtaining parameters
Figure 122038DEST_PATH_IMAGE058
、
Figure 792054DEST_PATH_IMAGE059
、
Figure 641061DEST_PATH_IMAGE060
、
Figure 438247DEST_PATH_IMAGE061
、
Figure 954679DEST_PATH_IMAGE062
、
Figure 795596DEST_PATH_IMAGE063
、
Figure 131900DEST_PATH_IMAGE064
(ii) a Will be provided with
Figure 780048DEST_PATH_IMAGE050
、
Figure 354249DEST_PATH_IMAGE049
、
Figure 366067DEST_PATH_IMAGE065
、
Figure 455246DEST_PATH_IMAGE059
、
Figure 328655DEST_PATH_IMAGE060
、
Figure 819679DEST_PATH_IMAGE066
、
Figure 2399DEST_PATH_IMAGE067
、
Figure 313295DEST_PATH_IMAGE068
、
Figure 488930DEST_PATH_IMAGE064
All 9 parameters are used as independent variables, linear fitting is carried out on the cylinder efficiency, and the cylinder efficiency is obtained respectively
Figure 631198DEST_PATH_IMAGE048
、
Figure 984819DEST_PATH_IMAGE049
、
Figure 533743DEST_PATH_IMAGE069
、
Figure 263802DEST_PATH_IMAGE059
、
Figure 198260DEST_PATH_IMAGE070
、
Figure 722782DEST_PATH_IMAGE071
、
Figure 54275DEST_PATH_IMAGE067
、
Figure 119183DEST_PATH_IMAGE068
、
Figure 924459DEST_PATH_IMAGE072
Obtaining a cylinder efficiency fitting surface model corresponding to each layer of data according to the weight of the data, and determining a coefficient R2And the mean square error MSE judges the fitting effect.
Preferably, the coefficient R is determined in step 32The value range of the set value is 0.9-1.
Preferably, the coefficient R is determined in step 32Is set toThe value was 0.95.
Preferably, the real-time data collected in step 4 includes unit load, opening of the regulating valve, steam inlet pressure, steam inlet temperature, steam exhaust pressure and steam exhaust temperature.
Preferably, when the score is set for each curved surface in step 6, each layer of curved surface above the average value of cylinder efficiency is set as a positive score, each layer of curved surface below the average value of cylinder efficiency is set as a negative score, and the scores are symmetrically distributed, for example, the score of the 1 st layer is + e score, and the score of the last layer is-e score.
The invention has the beneficial effects that:
according to the method, a turbine cylinder efficiency fitting surface model is constructed from multiple dimensions through a computer big data analysis technology based on historical data and steady state identification, the change of the cylinder efficiency is amplified through a layering and scoring method, the abnormity of the through-flow part of the turbine is early warned, the early problems of scaling of the through-flow part, damage of a steam inlet filter screen and the like are found, and sufficient intervention time is provided for the fault treatment of the turbine.
When the unit actually operates, the normal value of the cylinder efficiency also changes according to the change of the working condition because the load of the unit changes in real time; the cylinder efficiency fitting curved surface model is constructed through historical data, the upper limit and the lower limit (falling into the curved surface at the uppermost layer and the curved surface at the lowermost layer) of the normal value of the actual operating cylinder efficiency are determined, and the sensitivity to small-amplitude fluctuation of the turbine cylinder efficiency is improved by utilizing a layering scoring method; whether the efficiency of the cylinder is abnormal or not is analyzed and judged through the real-time data fractional trend graph, so that whether the through-flow part is abnormal or not is judged, and the heat economy and the safety of the steam turbine are improved. Through a real-time data score trend graph, the total scores of the day and the last day, the total scores of the week and the last week, and even the cylinder efficiencies of different time spans such as the total scores of the year and the last year can be compared more visually, and the cylinder efficiency test after through-flow modification can be replaced to judge the through-flow modification effect.
Drawings
FIG. 1 is a flow chart of a turbine through-flow anomaly early warning method based on big data analysis according to the present invention;
FIG. 2 is a graph of one layer of cylinder efficiency layered curved surfaces according to example 2 of the present invention;
FIG. 3 is a load line diagram of a training data set in embodiment 2 of the present invention;
FIG. 4 is a line graph of the load of the unit after the steady-state screening of the training data in embodiment 2 of the present invention;
fig. 5 is a cylinder efficiency scoring trend chart in embodiment 2 of the present invention.
Detailed Description
The present invention will be further described with reference to the following examples. The following examples are set forth merely to aid in the understanding of the invention. It should be noted that, for a person skilled in the art, several modifications can be made to the invention without departing from the principle of the invention, and these modifications and modifications also fall within the protection scope of the claims of the present invention.
Example one
The embodiment of the application provides a turbine through-flow abnormity early warning method based on big data analysis as shown in fig. 1:
step 1, collecting historical data of selected measuring points, setting upper and lower limits of normal values of the data of the selected measuring points, and cleaning the data by using a box line graph to obtain a normal data set
Figure 619883DEST_PATH_IMAGE073
The expert uses the trend graph to align the normal data set
Figure 392667DEST_PATH_IMAGE074
Re-screening the data to obtain a normal data set
Figure 198949DEST_PATH_IMAGE075
；
Step 2, calculating normal data sets item by item
Figure 357267DEST_PATH_IMAGE076
The cylinder efficiency of each piece of data is calculated, and the calculated cylinder efficiency is put into the data; removing abnormal values of the cylinder efficiency values obtained through calculation to obtain an integrated data set
Figure 489171DEST_PATH_IMAGE077
By slidingA time window for performing steady state identification according to the cylinder efficiency value to obtain a steady state data set
Figure 280409DEST_PATH_IMAGE078
；
Step 3, setting the steady state data set
Figure 906694DEST_PATH_IMAGE079
Dividing according to the unit load and the opening of the regulating valve to obtain a divided data set
Figure 404671DEST_PATH_IMAGE080
(ii) a Then each divided data set is
Figure 707476DEST_PATH_IMAGE081
Performing equidistant layering according to the distance from the average value of the cylinder efficiency of the segmentation data set, and combining the data of the same layer into a layered data set
Figure 189273DEST_PATH_IMAGE082
(ii) a Respectively constructing cylinder efficiency fitting surface models corresponding to each layer of data, wherein the number of the surface layers of the fitting surface models is the same as that of the surface layers after equidistant layering; determining coefficient R of the obtained fitted surface model2If the fitting effect of the curved surface is smaller than the set value, judging that the fitting effect of the curved surface does not reach the standard, returning to the step 2, removing abnormal values of the calculated cylinder efficiency value, and making a decision coefficient R2Is closer to 1 to screen out the steady state data set until the coefficient R is determined2Is greater than a set value;
step 4, collecting real-time data to obtain a real-time data set;
step 5, calculating the cylinder efficiency of each piece of data in the collected real-time data set one by one according to the step 2, and putting the calculated cylinder efficiency into the data to obtain a real-time integrated data set;
step 6, scoring the real-time integration data set; when the real-time integration data set is normal, the cylinder efficiency in each data falls in the top layer and the bottom layer of the fitted surface model obtained in the step 3; setting a score for each curved surface, wherein the real-time integration data corresponding to the cylinder efficiency falling into the curved surface is the score of the curved surface, the real-time integration data corresponding to the cylinder efficiency not falling onto the curved surface is the score of the curved surface closest to the curved surface, the real-time integration data corresponding to the cylinder efficiency exceeding the range of the curved surface is the score of the boundary curved surface, and counting the total score of all pieces of data in the real-time integration data set;
and 7, analyzing whether the cylinder efficiency is abnormally changed: collecting the total scores of all the data in the real-time integration data set obtained by statistics in the step 6 into a real-time integration data set score trend graph, and analyzing the score change trend; when the score rises to reach an upper threshold or falls to reach a lower threshold and lasts for a period of time, judging that the cylinder efficiency is abnormal; if the cylinder efficiency is abnormal, the alarm is triggered after the cylinder efficiency is judged to be abnormal.
Example two
On the basis of the first embodiment, the second embodiment of the present application verifies the effectiveness of the turbine through-flow abnormality early warning method based on big data analysis in the first embodiment by using data from a certain power plant high-pressure cylinder:
model training is carried out by using data from 1 month and 1 day 00:00 in 2018 to 12 month and 31 day and 24:00 in 2018, and data from 1 month and 1 day 00 in 2019 to 11 month and 30 days in 2019 and 24:00 in 24 months are used as test data (specific time when an accident is found during maintenance and not confirmed).
Step 1.1, the collected parameters comprise: the system comprises a unit load, a high-pressure regulating valve opening, main steam pressure, main steam temperature, high-pressure cylinder exhaust pressure and high-pressure cylinder exhaust temperature. The counting time interval is 5 seconds, and 6307200 pieces of 2018 annual data are obtained. The load line graph of the training data set in the embodiment is shown in fig. 3;
step 1.2, rejecting quality-unqualified data such as shutdown, dead spots, abnormality, mutation and the like in the collected data, rejecting 1011200 quality-unqualified data such as mutation, abnormality, shutdown and the like in the collected data by utilizing a box line diagram, and obtaining 5233600 data of a normal data set in actual operation after fine screening by professional personnel.
Step 2.1, alignmentCalculating the cylinder efficiency one by one according to the processed historical data, and putting the calculated cylinder efficiency into the data to obtain an integrated data set
Figure 858064DEST_PATH_IMAGE083
And eliminating abnormal values of the data with the cylinder efficiency values larger than 100 and smaller than 80 obtained through calculation to obtain 5210470 pieces of data.
And 2.2, screening stable data of the high-pressure cylinder efficiency within a period of time to obtain cylinder efficiency stable data, wherein a load line graph of the unit after the stable screening of the training data in the embodiment is shown in fig. 4. Obtaining a plurality of lists by sliding a time window of 5 minutes, screening data according to whether the ratio of the difference between the maximum value and the minimum value of the cylinder efficiency in the lists to the average value of the lists is greater than 0.15, merging the lists after screening to obtain 2747489 steady-state data sets
Figure 538444DEST_PATH_IMAGE084
。
And 3, constructing a cylinder efficiency fitting surface model. Steady state cylinder efficiency data set
Figure 746572DEST_PATH_IMAGE085
Dividing the data according to the load of 10MW and the opening of the high-pressure regulating valve of 1 percent to obtain a plurality of divided data sets
Figure 466397DEST_PATH_IMAGE086
Removing a data set with data smaller than 50, and then dividing each batch of data into 10 layers at equal intervals according to (maximum-minimum)/9 by being far from the average value of the cylinder efficiency, wherein a surface map of one layer of the cylinder efficiency layered surface is shown in fig. 2; and carrying out nonlinear fitting on each layer to obtain an equation, a fitting surface and decision coefficients of Mean Square Error (MSE) and R2 of each layer. As in table 1 below:
TABLE 1 equation, fitted surface and MSE and R2 decision coefficient table for each layer after equidistant layering
Figure 480489DEST_PATH_IMAGE087
And 4, acquiring real-time data, wherein the embodiment adopts test data of a fault time period to perform simulation test, the acquisition time interval is 1 second, and the real-time data set time window is set to be 5 minutes. In the embodiment, real-time data of 30 days are collected for testing, and 779100 pieces of data are obtained after screening.
And 5, calculating the cylinder efficiency of the collected real-time data, and putting the calculated cylinder efficiency into the data to obtain a real-time integrated data set.
And 6, grading the real-time integration data set, setting scores for each curved surface when the data of the real-time integration data set is normal, wherein the cylinder efficiency of each piece of data is within the uppermost curved surface and the lowermost curved surface, the scores from the first layer to the tenth layer are respectively 5, 4, 3, 2, 1, -2, -3, -4 and-5, the score of the curved surface is obtained for the point falling into the curved surface, the score of the nearest curved surface is obtained for the point not falling onto the curved surface, the score of the point above the first layer is 5, the score of the point below the tenth layer is-5, and the total score of all the points of the real-time integration data set is counted, wherein a score trend graph is shown in fig. 5.
And 7, analyzing whether the cylinder efficiency is abnormally changed or not. And (4) when the alarm score deviation is set by a professional to be more than 2 and continuously appears for 6 times, performing through-flow abnormity alarm. In the test example, the through-flow abnormal alarm appears about 13 days 11, 20 and 2019, and the score changes obviously after 20 days 11, 20 and 2019 through the analysis of the integrated score trend graph of the real-time integrated data for a period of 5 minutes. The score is larger than the previous score by 3 points, and the score is continuously larger until the test time is over, and the judgment of the occurrence time of the through-flow abnormity is accurate.

Claims (3)

1. A turbine through-flow abnormity early warning method based on big data analysis is characterized by comprising the following steps:
step 1, collecting historical data of selected measuring points, setting upper and lower limits of normal values of the data of the selected measuring points, and cleaning the data by using a box line graph to obtain a normal data set
Figure DEST_PATH_IMAGE001
The expert uses the trend graph to align the normal data set
Figure DEST_PATH_IMAGE002
The data in the step (1) is manually and finely screened to obtain a normal data set
Figure DEST_PATH_IMAGE003
；
Step 1.1, collecting unit load, valve opening degree, steam inlet pressure, steam inlet temperature, steam exhaust pressure and steam exhaust temperature of a selected measuring point;
step 1.2, data cleaning is carried out: setting upper and lower limits of normal data values of selected measuring points, and eliminating data exceeding the upper and lower limits of the normal data values in the historical data of the measuring points acquired in the step 1.1 by utilizing a box line diagram to obtain a normal data set
Figure 312481DEST_PATH_IMAGE001
；
Step 1.3, utilizing the trend graph to align the normal data set
Figure DEST_PATH_IMAGE004
Then fine screening is carried out to obtain a normal data set
Figure DEST_PATH_IMAGE005
；
Step 2, calculating normal data sets item by item
Figure 400261DEST_PATH_IMAGE003
The cylinder efficiency of each piece of data is calculated, and the calculated cylinder efficiency is put into the data; removing abnormal values of the cylinder efficiency values obtained through calculation to obtain an integrated data set
Figure DEST_PATH_IMAGE006
And performing steady state identification according to the cylinder efficiency value through sliding a time window to obtain a steady state data set
Figure DEST_PATH_IMAGE007
；
Step 2.1, calculate the normal data set item by item
Figure DEST_PATH_IMAGE008
The cylinder efficiency of each piece of data is calculated, and the calculated cylinder efficiency is put into the data, wherein the calculation formula of the cylinder efficiency is as follows:
Figure DEST_PATH_IMAGE009
in the above formula, the first and second carbon atoms are,
Figure DEST_PATH_IMAGE010
in order to be able to achieve a cylinder efficiency,
Figure DEST_PATH_IMAGE011
which represents the enthalpy of the incoming steam,
Figure DEST_PATH_IMAGE012
the expression of the enthalpy of the exhaust steam,
Figure DEST_PATH_IMAGE013
representing the isentropic enthalpy found by finding a water vapor enthalpy entropy diagram from the steam inlet entropy and the steam outlet pressure;
step 2.2, eliminating abnormal values of the cylinder efficiency values obtained through calculation by utilizing the boxplot to obtain an integrated data set
Figure DEST_PATH_IMAGE014
；
Step 2.3, screening out stable data in a set time length in the integrated data set to obtain a stable data set
Figure DEST_PATH_IMAGE015
: sliding the time window according to the set time length to integrate the data set
Figure DEST_PATH_IMAGE016
Separated into n lists
Figure DEST_PATH_IMAGE017
Screening out given values of cylinder efficiency fluctuation smaller than fluctuation amplitude in each list
Figure DEST_PATH_IMAGE018
The cylinder efficiency value is eliminated, and the cylinder efficiency fluctuation in each list is larger than the given value of fluctuation amplitude
Figure DEST_PATH_IMAGE019
Merging the screened residual cylinder efficiency values into a new list to obtain a steady state data set
Figure DEST_PATH_IMAGE020
(ii) a The screening formula is as follows:
Figure DEST_PATH_IMAGE021
in the above formula, the first and second carbon atoms are,
Figure DEST_PATH_IMAGE022
represents the maximum value of cylinder efficiency in the selected list,
Figure DEST_PATH_IMAGE023
represents the minimum value of cylinder efficiency in the selected list,
Figure DEST_PATH_IMAGE024
represents the average of the cylinder efficiencies in the selected list,
Figure 136922DEST_PATH_IMAGE018
representing a given value of fluctuation amplitude;
step 3, setting the steady state data set
Figure DEST_PATH_IMAGE025
Dividing according to the unit load and the opening of the regulating valve to obtain a divided data set
Figure DEST_PATH_IMAGE026
(ii) a Then each divided data set is
Figure DEST_PATH_IMAGE027
Performing equidistant layering according to the distance from the average value of the cylinder efficiency of the segmentation data set, and combining the data of the same layer into a layered data set
Figure DEST_PATH_IMAGE028
(ii) a Respectively constructing cylinder efficiency fitting surface models corresponding to each layer of data, wherein the number of the surface layers of the fitting surface models is the same as that of the surface layers after equidistant layering; determining coefficient R of the obtained fitted surface model2If the fitting effect of the curved surface is smaller than the set value, judging that the fitting effect of the curved surface does not reach the standard, returning to the step 2, removing abnormal values of the calculated cylinder efficiency value, and determining a coefficient R2The value range of the set value is 0.9-1, and a steady-state data set is screened out until a coefficient R is determined2Is greater than a set value;
step 3.1, set the steady state data
Figure DEST_PATH_IMAGE029
Dividing according to the unit load and the opening of the regulating valve collected in the step 1 to obtain n divided data sets
Figure DEST_PATH_IMAGE030
The set of all the segmented data sets is
Figure DEST_PATH_IMAGE031
Wherein
Figure DEST_PATH_IMAGE032
Representing the nth divided data set
Figure 719082DEST_PATH_IMAGE030
(ii) a Abandon
Figure DEST_PATH_IMAGE033
Segmented data set with fewer than 50 data strips
Figure DEST_PATH_IMAGE034
；
Step 3.2, divide each data set into
Figure DEST_PATH_IMAGE035
Performing equidistant layering according to the distance from the average value of the cylinder efficiency of the segmentation data set, and combining the data of the same layer into a layered data set
Figure DEST_PATH_IMAGE036
The set of all hierarchical data sets is
Figure DEST_PATH_IMAGE037
The combination rule is as follows:
Figure DEST_PATH_IMAGE038
in the above formula, the first and second carbon atoms are,
Figure DEST_PATH_IMAGE039
represents the m-th level of the hierarchical data set,
Figure DEST_PATH_IMAGE040
representing the 1 st segmented data set
Figure DEST_PATH_IMAGE041
The data of the m-th layer in (1),
Figure DEST_PATH_IMAGE042
representing the 2 nd segmented data set
Figure DEST_PATH_IMAGE043
The data of the m-th layer in the (C),
Figure DEST_PATH_IMAGE044
representing the nth divided data set
Figure DEST_PATH_IMAGE045
The mth layer of data;
3.3, carrying out nonlinear fitting on each layer of hierarchical data set according to the cylinder efficiency, the opening of the regulating valve and the unit load: regulating valve opening with cylinder efficiency as a dependent variable
Figure DEST_PATH_IMAGE046
Load of the machine set
Figure DEST_PATH_IMAGE047
As independent variable, pair
Figure 987251DEST_PATH_IMAGE046
、
Figure 777090DEST_PATH_IMAGE047
Carrying out third-order dimension expansion, wherein the third-order dimension expansion calculation formula is as follows:
Figure DEST_PATH_IMAGE048
；
Figure DEST_PATH_IMAGE049
；
Figure DEST_PATH_IMAGE050
；
Figure DEST_PATH_IMAGE051
；
Figure DEST_PATH_IMAGE052
；
Figure DEST_PATH_IMAGE053
；
Figure DEST_PATH_IMAGE054
obtaining parameters
Figure DEST_PATH_IMAGE055
、
Figure DEST_PATH_IMAGE056
、
Figure DEST_PATH_IMAGE057
、
Figure DEST_PATH_IMAGE058
、
Figure DEST_PATH_IMAGE059
、
Figure DEST_PATH_IMAGE060
、
Figure DEST_PATH_IMAGE061
(ii) a Will be provided with
Figure DEST_PATH_IMAGE062
、
Figure 220622DEST_PATH_IMAGE047
、
Figure 888363DEST_PATH_IMAGE055
、
Figure 760504DEST_PATH_IMAGE056
、
Figure 222710DEST_PATH_IMAGE057
、
Figure DEST_PATH_IMAGE063
、
Figure DEST_PATH_IMAGE064
、
Figure DEST_PATH_IMAGE065
、
Figure DEST_PATH_IMAGE066
As independent variables, linear fitting is carried out on the cylinder efficiency to respectively obtain
Figure 288624DEST_PATH_IMAGE062
、
Figure 291215DEST_PATH_IMAGE047
、
Figure 283441DEST_PATH_IMAGE055
、
Figure 182127DEST_PATH_IMAGE056
、
Figure 627015DEST_PATH_IMAGE057
、
Figure 869515DEST_PATH_IMAGE058
、
Figure DEST_PATH_IMAGE067
、
Figure 450669DEST_PATH_IMAGE065
、
Figure 51415DEST_PATH_IMAGE066
Obtaining a cylinder efficiency fitting surface model corresponding to each layer of data according to the weight of the data, and determining a coefficient R2And mean square error MSE to judge the fitting effect;
step 4, collecting real-time data to obtain a real-time data set;
step 5, calculating the cylinder efficiency of each piece of data in the collected real-time data set one by one according to the step 2, and putting the calculated cylinder efficiency into the data to obtain a real-time integrated data set;
step 6, scoring the real-time integration data set; the specific way of integrating the data sets in real time for scoring is as follows: when the real-time integration data set is normal, the cylinder efficiency in each data falls in the top layer and the bottom layer of the fitted surface model obtained in the step 3; setting a score for each curved surface, wherein the real-time integration data corresponding to the cylinder efficiency falling into the curved surface is the score of the curved surface, the real-time integration data corresponding to the cylinder efficiency not falling onto the curved surface is the score of the curved surface closest to the curved surface, the real-time integration data corresponding to the cylinder efficiency exceeding the range of the curved surface is the score of the boundary curved surface, and counting the total score of all pieces of data in the real-time integration data set;
and 7, analyzing whether the cylinder efficiency is abnormally changed: collecting the total scores of all the data in the real-time integration data set obtained by statistics in the step 6 into a real-time integration data set score trend graph, and analyzing the score change trend; when the score rises to reach an upper threshold or falls to reach a lower threshold and lasts for a period of time, judging that the cylinder efficiency is abnormal; if the cylinder efficiency is abnormal, the alarm is triggered after the cylinder efficiency is judged to be abnormal.
2. The turbine through-flow abnormity early warning method based on big data analysis according to claim 1, characterized in that: determining the coefficient R in step 32The set value of (a) is 0.95.
3. The turbine through-flow abnormity early warning method based on big data analysis according to claim 1, characterized in that: the real-time data collected in the step 4 comprise unit load, adjusting valve opening, steam inlet pressure, steam inlet temperature, steam exhaust pressure and steam exhaust temperature.
CN202111519717.3A 2021-12-14 2021-12-14 Turbine through-flow abnormity early warning method based on big data analysis Active CN113918624B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111519717.3A CN113918624B (en) 2021-12-14 2021-12-14 Turbine through-flow abnormity early warning method based on big data analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111519717.3A CN113918624B (en) 2021-12-14 2021-12-14 Turbine through-flow abnormity early warning method based on big data analysis

Publications (2)

Publication Number Publication Date
CN113918624A CN113918624A (en) 2022-01-11
CN113918624B true CN113918624B (en) 2022-03-22

Family

ID=79249135

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111519717.3A Active CN113918624B (en) 2021-12-14 2021-12-14 Turbine through-flow abnormity early warning method based on big data analysis

Country Status (1)

Country Link
CN (1) CN113918624B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114460470A (en) * 2022-01-26 2022-05-10 上海玫克生智能科技有限公司 Battery pack state analysis method and system based on voltage and terminal

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103850726A (en) * 2014-03-28 2014-06-11 苏州华瑞能泰发电技术有限公司 Method for quickly determining stationary sliding pressing optimization curve of steam turbine
CN109948847A (en) * 2019-03-18 2019-06-28 河海大学 A kind of multi-objective Evolutionary Algorithm applied to multi-reservoir scheduling
CN111434115A (en) * 2017-12-08 2020-07-17 华为技术有限公司 Clustering correction for texture synthesis in video coding
CN112766724A (en) * 2021-01-20 2021-05-07 中国工商银行股份有限公司 Service monitoring method, device and equipment

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103850726A (en) * 2014-03-28 2014-06-11 苏州华瑞能泰发电技术有限公司 Method for quickly determining stationary sliding pressing optimization curve of steam turbine
CN111434115A (en) * 2017-12-08 2020-07-17 华为技术有限公司 Clustering correction for texture synthesis in video coding
CN109948847A (en) * 2019-03-18 2019-06-28 河海大学 A kind of multi-objective Evolutionary Algorithm applied to multi-reservoir scheduling
CN112766724A (en) * 2021-01-20 2021-05-07 中国工商银行股份有限公司 Service monitoring method, device and equipment

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
基于多级非线性回归的低压缸性能在线评价方法;秦刚华等;《电站***工程》;20210131;第37卷(第1期);第10-13页 *
智慧电厂建设中设备特性认知及能效提升关键问题研究;周慎学等;《电力与能源》;20171231;第38卷(第6期);第718-723页 *
计及大规模全钒液流电池储能***外特性建模与仿真;葛维春等;《电力***保护与控制》;20190901;第47卷(第17期);第171-179页 *

Also Published As

Publication number Publication date
CN113918624A (en) 2022-01-11

Similar Documents

Publication Publication Date Title
JP5091604B2 (en) Distribution evaluation method, product manufacturing method, distribution evaluation program, and distribution evaluation system
CN113918624B (en) Turbine through-flow abnormity early warning method based on big data analysis
US5838561A (en) Automatic control loop monitoring and diagnostics
CN104573850A (en) Method for evaluating state of thermal power plant equipment
CN112179655B (en) Turbo generator fault early warning method based on threshold classification
CN107851294B (en) State-based preventive maintenance device and method for large-scale operation system
CN113806346A (en) Turbine degradation trend measuring method and terminal based on big data analysis
CN113176081B (en) Historical data-based turbine blade wear monitoring method
CN117314244B (en) Process flow data supervision system and method based on data analysis
CA2252868C (en) Automatic control loop monitoring and diagnostics
CN115905974B (en) Method for detecting abnormal furnace condition of blast furnace
CN110928248B (en) Method for determining performance degradation degree of gas turbine
CN112685216A (en) Equipment abnormity monitoring system and method based on trend analysis
CN112215493A (en) Method for judging and evaluating process quality of cigarette shredding workshop section
CN110441063B (en) Method for monitoring and diagnosing cracks of large high-speed rotor shaft
CN108182306B (en) Method for determining degradation failure threshold of abrasive particle characteristic parameters of vehicle power transmission device
CN112628132B (en) Water pump key index prediction method based on machine learning
CN115062979A (en) Metal roof performance evaluation method and system based on hierarchical analysis and fuzzy evaluation
Ding et al. Research on seasonal correlation of wavelet packet energy spectrum and temperature of Runyang Suspension Bridge
CN112508350A (en) Vibration early warning method and system for steam turbine generator unit
CN115539220B (en) Gas turbine blade channel temperature sensor fault detection method
CN114969140B (en) Method for detecting and analyzing performance data of fluent strip products
CN117391312B (en) Intelligent alarm processing method and system for intelligent building
CN116943329A (en) Wind turbine generator gearbox filter blockage early warning diagnosis method and system
Purba et al. Gas Turbine Hidden Capacity Recovery by Inlet Air Filter Variation Method to Produce Clean and Efficient Energy

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant