CN110414155A - A kind of detection of fan part temperature anomaly and alarm method with single measuring point - Google Patents
A kind of detection of fan part temperature anomaly and alarm method with single measuring point Download PDFInfo
- Publication number
- CN110414155A CN110414155A CN201910702600.5A CN201910702600A CN110414155A CN 110414155 A CN110414155 A CN 110414155A CN 201910702600 A CN201910702600 A CN 201910702600A CN 110414155 A CN110414155 A CN 110414155A
- Authority
- CN
- China
- Prior art keywords
- temperature
- alarm
- measuring point
- data
- model
- 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.)
- Granted
Links
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
- Y02B10/00—Integration of renewable energy sources in buildings
- Y02B10/30—Wind power
Landscapes
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The fan part temperature anomaly with single measuring point that this application involves a kind of detects and alarm method, it is fitted the coefficient for obtaining multiple linear equation including the use of data of the training set data during equipment early stage fault-free to part temperatures measuring point to be measured, constructs temperature foh model；It is predicted using temperature point of the model of fit to test set data, the difference of the actual temperature and model prediction temperature that use test data is as prediction residual sequence；When on-line operation, the Wind turbines real-time running data of set time length T is collected, passes through the feature extracted using model and the temperature point of online data is predicted, obtain model prediction temperature；Use the difference of the parts to be tested real time temperature and model prediction temperature as prediction residual sequence, the statistical analysis of one group of online residual error numerical value based on acquisition, into the part temperatures early warning program.
Description
Technical field
The fan part temperature anomaly with single measuring point that this application involves a kind of detects and alarm method, and it is different to be suitable for blower
The technical field often detected.
Background technique
Existing portion of techniques includes that the included master control system of Wind turbines often limits the early warning logic of temperature anomaly
Determine the method for threshold value, i.e., discovery temperature below or above some pre-set threshold value i.e. generate alarm, such methods compared with
To be intuitive, simple, sensitive, but encounter when distinct temperature caused by some extraneous factors changes and often will appear false alarm, In
Performance is poor in terms of the Stability and veracity of early warning.It is usually all in blower to be measured by taking base bearing temperature pre-warning logic as an example
One or two temperature point is set on component, and wind power plant SCADA system (data acquisition and supervisor control) is by simple
The method of base bearing temperature threshold is set to be monitored to base bearing operating status, when base bearing temperature is higher than the threshold value of setting
When SCADA will to wind field owner issue alarm.This alarm mode is too simple, there is a high rate of false alarm, and component
Failure is often accumulation, a progressive process, the not lead of suitable length of alarming often of SCADA master control system, nothing
Method assists owner to carry out predictive maintenance.In the prior art to the detection of temperature anomaly and evaluation criterion mainly include the following types:
(1) early warning mechanism based on part temperatures characteristic quantity.Chinese patent application 201410355897.X,
In 201810698450.0, method by dividing operating condition, to the subset Extracting temperature characteristic quantity that unit data mark off, according to
The threshold value determined carries out the statistical method that real time temperature is more than threshold value and carries out fault pre-alarming.This alarm mode is more careful,
And temperature difference of the Wind turbines under different operation conditions is considered, but still belong to the method for absolute threshold, accurately
The disadvantage that rate is lower, alarm is unstable still remains, and depend heavilys on manually determining threshold value.
(2) based on the residual error early warning mechanism of normal behaviour model.Chinese patent application 201810146979.1,
201711477916.6, in 201710853212.8, by obtaining change relevant to the parts to be tested temperature in blower historical data
Amount, training neural network model, predicts the temperature of the parts to be tested, after line obtains real time temperature by actual temperature and
The residual error of predicted temperature deviates to determine failure.This alarm mode is more rigorous, has fully considered that the physics of component operation is special
Property and history run, when failure occurs, residual error is generally possible to embody fault signature, but due to training data work
The dependence of condition integrality and operating condition coverage rate, the operating condition differentiation of this method due to only relying on residual absolute value when changing, holds
It easily reports by mistake, stability is bad.
(3) based on cluster to target early warning mechanism.In Chinese patent application 201910027340.6, by full blast field wind
The mode of machine setting monitoring unit and reference unit, will monitor the temperature data and remaining reference that unit the parts to be tested obtains every time
The temperature data ensemble average value of unit compares, and determines if the temperature difference is more than certain proportion and has apparent ascendant trend
Monitor the temperature fault of unit.This alarm mode is more intuitive, but false to the type feature of full blast field blower, health status
And if physical characteristic coherence request when operation it is higher, once be unable to satisfy fan operation situation it is consistent if not can guarantee it is pre-
It is alert accurate.In addition especially in mountain wind field, due to the influence of turbulent flow, weather conditions locating for adjacent blower even have difference
Not.
In addition, existing method can not form system closed loop, it is difficult to make system to the temperature anomaly of Wind turbines component
Monitors and alarm to property.Once temperature anomaly failure occurs in Wind turbines component from the point of view of experience, it can continue to that failure terminates, this
Requirement of one feature to alarm mechanism is high, it is desirable that quasi- to the early warning of the temperature monitoring systems of these components before failure generation
Really, stablize, and due to moving closer to failure, the grade of early warning can also be correspondinglyd increase, this is that the prior art faces most
Big challenge.
To sum up, problems of the prior art include: either the judgment method based on temperature threshold be still based on it is residual
The judgment method of differential mode type, accuracy, predictability and the stability of alarm are often difficult to meet simultaneously；Current existing temperature is different
Normal diagnostic techniques is usually not transitioned into the alarming logic at final user oriented end, and the demand and requirement to this technology all compare
It is higher, for wind field owner under the premise of reducing rate of false alarm, rate of failing to report, alarm signal is steadily exported as far as possible, risen
To the effect for preventing and reminding.
Summary of the invention
The application overcomes the irritability of absolute threshold differentiation, compared to simple residual error discrimination model accuracy and stabilization
Degree all increases, can to it is abnormal keep sensitive under the premise of reduce rate of false alarm, and ultimately form it is reasonable, have grade
The alarm of differentiation facilitates wind field owner to carry out predictive maintenance.
According to a kind of detection of fan part temperature anomaly and alarm method with single measuring point of the application, the blower portion
Part is equipped with single temperature point, and the detection and alarm method include mechanism driving data modeling program and the portion based on residual error
Part temperature pre-warning program；The mechanism driving data modeling program the following steps are included:
(1) monitoring data when fan operation are obtained, one group of equipment early stage and the without failure number of the parts to be tested are selected
According to as training data；
(2) according to the physical characteristic of the parts to be tested, original variable relevant to part temperatures is filtered out；
(3) pretreatment is carried out to original variable and obtains new processing variable, retouched according to original variable and processing variable construction
State the multiple linear equation of part temperatures；
(4) historical data that will acquire is divided into training set and test set, and training set is used to be fitted temperature prediction model,
The threshold value that test set is used to subsequent step determines；
(5) coefficient for obtaining multiple linear equation is fitted using data of the training set to part temperatures measuring point to be measured,
Temperature foh model is constructed, the model of fit of part temperatures measuring point measured value is obtained；
(6) it is predicted using temperature point of the model of fit to test set data, obtains model prediction temperature；Sentence simultaneously
The fitting precision of disconnected model reaches accuracy rate and requires then to save the model of fit；
(7) difference of the actual temperature and model prediction temperature that use test data is as prediction residual sequence；
(8) for single measuring point component, one group of residual error is obtained, is stored for on-line prediction use；
(9) when on-line operation, the Wind turbines real-time running data of set time length T is collected, according in step (3)
Method extracts feature to online data；
(10) pass through the feature extracted using model to predict the temperature point of online data, obtain model prediction
Temperature；
(11) difference of the parts to be tested real time temperature and model prediction temperature is used to obtain one group as prediction residual sequence
Online residual error；
(12) statistical analysis of one group of online residual error numerical value based on acquisition, into the part temperatures early warning program.
Preferably, the part temperatures early warning program based on residual error the following steps are included:
(1) the total sample point quantity N of online data in period T is recorded；For single measuring point component, pass through statistics first
Method carries out normal approach to test set residual error, the variance evaluation σ after being fitted2；
(2) weight term λ is added to the section n σ of the normal probability density curve of fitting, the weight closer to mean value center is got over
Small, the weight at more deviation mean value center is bigger；The number of each probability interval in right side is fallen on according to weight term and online residual sample
Interaction building health index HI
λkIndicate the weight term of setting, nkIndicate that online residual sample falls on the number of each probability interval in right side, i=1,2,
3,4；
(3) by online residual error slide window processing, four HI are calculated to the sample point in each small window, and count HI1, HI2,
HI3, HI4Value；If HI in the time window3+HI4Greater than HI1+HI2, then W is denoted as to the time window1, otherwise it is denoted as W0；
To the W in all time windows0And W1It is counted, if most classes are W1, then judge the Fans portion in time T
Temperature anomaly failure occurs for part, and output Warning Sign parameter is true.
Preferably, all Warning Sign parameters of the parts to be tested measuring point are exported into algorithm operation result database and is protected
It deposits, each algorithm operation calls the Warning Sign parameter value of historical storage to carry out logic judgment；If continuously algorithm operation is defeated three times
Warning Sign parameter out be true or history n times operation in Warning Sign parameter be really be more than half, then this operation triggering
Alarm；Wherein, the size of n is determined according to algorithm running frequency.
Preferably, when triggering alarm, set the alarm identification parameter of corresponding measuring point be it is true, by the corresponding alarm mark of measuring point
Parameter is exported into algorithm operation result database and is saved, the operation of each algorithm call the alarm identification parameter of historical storage into
Row logic judgment, if this alarm identification parameter for running corresponding measuring point is very, to take identification parameter of alarming in history k times operation
The value-at-risk of this alarm is determined for genuine number, and ratio is mapped as alarm level；The size of k runs frequency according to algorithm
Rate is determining, k > > n.
Wherein, after maintenance, the history early warning library of corresponding component is reset fan part；The mechanism driving data modeling
Original variable in the step of program (2) includes wind speed, active power, generator speed, the measured value of temperature point and cabin
At least one of temperature；In the step of mechanism driving data modeling program (6), the judgment criteria of the fitting precision of model
It is carried out by residual analysis and fitting precision analysis；If fitting precision is less than the 10% of all data mean values, and residual error is logical
QQ-Norm is crossed to examine and Jarque-Bera inspection, obedience normality distribution, then it is assumed that reach the requirement of accuracy rate.
The application's has the beneficial effect that
1. temperature anomaly monitoring and alarm method method proposes the component for Wind turbines with temperature point,
It can be suitable for the monitoring temperature of single measuring point component；
2. the physics law that this method is run according to different components, using normal behaviour model modeling obtain predicted temperature and
Actual temperature residual error fully takes into account the characteristic and history run of independent every Fans, for the blower with single measuring point
Component, the method analyzed on the basis of obtained one-dimensional residual error using statistical distribution sufficiently improve the accuracy of alarm, and phase
There is certain lead compared with absolute threshold early warning；
3. this method on the basis of normal behaviour model, fully takes into account the alarm condition of independent every Fans history simultaneously
It brings into the alarming logic of real time monitoring, sufficiently improves the stability of alarm, and meet the alarm level for closing on failure generation
It is incremented by.
Detailed description of the invention
Fig. 1 is the flow diagram of fan part the temperature anomaly detection and alarm method with single measuring point of the application.
Fig. 2 is the probability density curve of the test set residual error normal approach in the application.
Specific embodiment
For the purposes, technical schemes and advantages of the application are more clearly understood, below in conjunction with attached drawing to the application
Embodiment be described in detail.It should be noted that in the absence of conflict, in the embodiment and embodiment in the application
Feature can mutual any combination.
As shown in Figure 1, according to a kind of detection of fan part temperature anomaly and alarm method with single measuring point of the application,
The fan part is equipped with a temperature point, and the detection and alarm method include the modeling of mechanism driving data and be based on residual
Two processes of part temperatures early warning of difference,
The modeling of mechanism driving data includes following processing step:
1. the monitoring data of SCADA system when obtaining fan operation, select one group of equipment early stage and the parts to be tested does not occur
The data of failure are as training data；
2. according to the physical characteristic of the parts to be tested, original variable relevant to part temperatures is filtered out, including wind speed, active
Power, generator speed, the measured value of temperature point and cabin temperature and data timestamp；
3. pair original variable carries out pretreatment and obtains new processing variable, according to original variable and processing variable construction description
The multiple linear equation of part temperatures；
4. data are divided into training set and survey according to the ratio of such as 7:3 according to the quantity of the historical data got
Examination collection, training set are used to be fitted temperature prediction model, and the threshold value that test set is used to subsequent step determines；
5. using such as homing method (including but not limited to linear regression, Ridge are returned, LASSO recurrence) to be measured
Part temperatures measuring point is fitted the coefficient for obtaining multiple linear equation, and the homing method based on tree-model is (including but not limited to
Random forests algorithm, XGBoost algorithm etc.) and homing method neural network based building temperature foh model, obtain portion
The model of part temperature point measured value；
6. passing through the feature extracted using model to predict the temperature point of test data, model prediction temperature is obtained
Degree；The fitting precision of judgment models simultaneously reaches accuracy rate and requires then preservation model；Specific judgment criteria can for example pass through
Residual analysis and fitting precision analysis, if fitting precision (RMSE) is less than the 10% of all data mean values, and residual error passes through
QQ-Norm is examined and Jarque-Bera is examined, Normal Distribution；
7. the difference of the actual temperature for using test data and model prediction temperature is as prediction residual sequence；
8. obtaining one group of residual error RES for single measuring point component, it is stored for on-line prediction use；
9. when on-line operation, the Wind turbines real-time running data of set time length T is collected, according to the side in step 3
Method extracts feature to online data；
10. passing through the feature extracted using model to predict the temperature point of online data, model prediction is obtained
Temperature；
11. using the difference of the parts to be tested real time temperature and model prediction temperature as prediction residual sequence (following shorthand
For online residual error), single measuring point obtains one group of online residual error.
Part temperatures early warning based on residual error includes following processing step:
1. recording the total sample point quantity N of online data in period T；For single measuring point component, pass through statistics first
Method carries out normal approach to test set residual error, the variance evaluation σ after being fitted2；
Var (X)=σ2=∫ (x- μ)2F (x) dx=∫ x2f(x)dx-μ2
2. weight term λ is added in the section n σ of the normal probability density curve of pair fitting, the weight closer to mean value center is got over
Small, the weight at more deviation mean value center is bigger, and curve in Fig. 2 is the probability density curve of test set residual error normal approach, and 1,
Sample point in 2,3,4 four probability intervals respectively corresponds weight term from small to large, only focuses on and falls on the right side of mean value (i.e. right half
Part) sample；The number interaction building health index of each probability interval in right side is fallen on according to weight term and online residual sample
HI
λkIndicate the weight term of setting, nkIndicate that online residual sample falls on the number of each probability interval in right side, i=1,2,
3,4；
3. online residual error slide window processing is calculated four HI to the sample point in each small window, and count HI1, HI2,
HI3, HI4Value；If HI in the time window3+HI4Greater than HI1+HI2, then W is denoted as to the time window1, otherwise it is denoted as W0；
To the W in all time windows0And W1It is counted, if most classes are W1, then judge the Fans portion in time T
Temperature anomaly failure occurs for part, and output Warning Sign parameter WarningFlag is True (true)；
4. exporting all Warning Sign parameter WarningFlag of the parts to be tested measuring point to algorithm operation result database
Middle preservation, each algorithm operation call the WarningFlag value of historical storage to carry out logic judgment；Single measuring point is treated, if continuously
It is more than half that the WarningFlag of algorithm operation output, which is that WarningFlag is True in True or the operation of history n times, three times
Number, then this operation triggering alarm, the alarm identification parameter AlarmFlag for concurrently setting corresponding measuring point is True, and the size of n takes
Certainly in algorithm running frequency；
The corresponding alarm identification parameter AlarmFlag of measuring point is exported into algorithm operation result database and is saved, every time
Algorithm operation calls the AlarmFlag value of historical storage to carry out logic judgment, if this alarm for running corresponding measuring point identifies ginseng
Number AlarmFlag is True (true), then taking AlarmFlag in history k times operation is the number of True to determine this alarm
Value-at-risk, and ratio is mapped as alarm level AlarmLevel, i.e., the number that AlarmFlag is True in history k times operation
More, alarm level is higher；The size of k depends on algorithm running frequency, usual k > > n；
5. if algorithm operation does not trigger alarm identification parameter WarningFlag or alarm identification parameter AlarmFlag, no
Alarm；
6. after wind field owner overhauls, the history early warning library of corresponding component is reset.
Mechanism and data combination drive are based in the application, using the data predictions such as energy conservation equation and feature construction
Method, establishes the corresponding temperature foh model of different measuring points namely normal behaviour model, lift scheme forecast quality reach pre-
The purpose of the property surveyed maintenance；For the big component of single measuring point, temperature prediction residual error is used and targetedly post-processes mechanism, in which: is logical
The mode for crossing probability density interal separation and sliding window ensure that the accuracy of alarm, rate of false alarm be also effectively reduced；It is returned using oneself
Alarm mechanism, join algorithm model and result database are presented, the operation alarm condition of history algorithm is fully considered and brings into pre-
In the alert logic of alarm, the stability of alarm is improved, and realizes and has the alarm feedback that grade is gradually incremented by when failure.
Embodiment
Algorithm ann test is carried out using certain wind field fan operation data in 2018, testing single measuring point is wind turbine power generation
Machine stator winding temperature, because temperature anomaly failure does not occur, data simulation is taken in this experiment, and runs and tie with initial data
Fruit compares.
Verification step:
1. acquiring 10 months operation datas of the wind field；By the last one month collected generator unit stator winding temperature people
Random noise and tendency noise is added in work, simulates as abnormal data；
2. being subject to manual construction as input parameter by load parameter, environmental parameter, other reference measuring point temperature etc. and building
The normal behaviour model of vertical generator unit stator winding temperature prediction；
3. having directlyed adopt the regression algorithm based on tree-model for generator unit stator winding, it is subject to Feature Selection, uses
Input feature vector mainly have temperature of engine bearings, generator speed, electric current, generator active power etc., use grid-search
Tuning is carried out to the hyper parameter of XGBoost model, determines optimal feature weight, the depth capacity of tree, learning rate etc..
4. choosing above-mentioned variable using the data before the last one month, returning based on model is carried out to the equation in step 3
Return fitting；
5. observing fitting precision, test set residual error Normal Distribution, preservation model.According to the distribution feelings of test set residual error
Condition has determined probability density function estimated by the residual distribution of generator unit stator winding (p.d.f), obtains 0 He of mean value of residual error
Estimate variance 0.2；Gradient weight is arranged to different sections in section n_sigma (n σ) that residual distribution is divided according to abovementioned steps
, represent the sensitivity to failures at different levels；
6. the data using the fan trouble the first half carry out online verification；
7. daily algorithm operation is primary as unit of one day, the data on the primary input same day, every operation once exports this time
The value-at-risk and alarm level of operation；
8. sample frequency is 5 seconds for generator unit stator winding, the sample point number once inputted is 17280, by sample
Point is window width according to every 360 points, and every 100 points are step-length, and sliding window constructs 170 small windows.To the sample in each small window, system
It counts and falls in the sample point number that each residual error divides region in all samples, in conjunction with four etc. in the corresponding small window of weight term building
The health value HI of grade, judgement obtain the output result (0/1) of each small window；
9. the health value of pair all small windows counts, if whole is more than that the health of half window refers to that output result is 1, meet
Trigger early-warning conditions；
10. if generating early warning more than half occur continuously running three times all generating in early warning or nearest ten operations, most
Nearly primary operation triggering alarm；
11. determining the value-at-risk of this alarm according to the alarm times of historical accumulation, and it is mapped to alarm level, value-at-risk
40 the above are more serious；
12. in online verification, time span is two months.In the emulation of the last one month fault data, this algorithm is from mould
Quasi- daystart continuously generates the constantly incremental alarm of 40 or more, grade, and control group test 1 is that residual error is directly used n_
The mode early warning of sigma threshold value is not having faulty first month to occur reporting by mistake for several times；Control group test 2 is using original
Data execute algorithm, do not alarm in two months.Prove that this method can in advance, accurately, steadily predict failure,
Have the function that prevent trouble before it happens.
Although embodiment disclosed by the application is as above, the content is only to facilitate understanding the application and adopting
Embodiment is not limited to the application.Technical staff in any the application technical field is not departing from this
Under the premise of the disclosed spirit and scope of application, any modification and change can be made in the implementing form and in details,
But the scope of patent protection of the application, still should be subject to the scope of the claims as defined in the appended claims.
Claims (7)
1. a kind of detection of fan part temperature anomaly and alarm method with single measuring point, the fan part are equipped with single temperature
Spend measuring point, which is characterized in that described detect with alarm method includes mechanism driving data modeling program and the component based on residual error
Temperature pre-warning program；
The mechanism driving data modeling program the following steps are included:
(1) monitoring data when fan operation are obtained, one group of equipment early stage is selected and the without failure data of the parts to be tested is made
For training data；
(2) according to the physical characteristic of the parts to be tested, original variable relevant to part temperatures is filtered out；
(3) pretreatment is carried out to original variable and obtains new processing variable, description portion is constructed according to original variable and processing variable
The multiple linear equation of part temperature；
(4) historical data that will acquire is divided into training set and test set, and training set is used to be fitted temperature prediction model, test
The threshold value that collection is used to subsequent step determines；
(5) coefficient for obtaining multiple linear equation, building are fitted using data of the training set to part temperatures measuring point to be measured
Temperature foh model obtains the model of fit of part temperatures measuring point measured value；
(6) it is predicted using temperature point of the model of fit to test set data, obtains model prediction temperature；Mould is judged simultaneously
The fitting precision of type reaches accuracy rate and requires then to save the model of fit；
(7) difference of the actual temperature and model prediction temperature that use test data is as prediction residual sequence；
(8) for single measuring point component, one group of residual error is obtained, is stored for on-line prediction use；
(9) when on-line operation, the Wind turbines real-time running data of set time length T is collected, according to the method in step (3)
Feature is extracted to online data；
(10) pass through the feature extracted using model to predict the temperature point of online data, obtain model prediction temperature
Degree；
(11) use the difference of the parts to be tested real time temperature and model prediction temperature as prediction residual sequence, obtain one group it is online
Residual error；
(12) statistical analysis of one group of online residual error numerical value based on acquisition, into the part temperatures early warning program.
2. the detection of fan part temperature anomaly and alarm method, feature according to claim 1 with single measuring point exists
In, the part temperatures early warning program based on residual error the following steps are included:
(1) the total sample point quantity N of online data in period T is recorded；For single measuring point component, pass through statistical method first
Normal approach is carried out to test set residual error, the variance evaluation σ after being fitted2；
(2) weight term λ is added to the section n σ of the normal probability density curve of fitting, the weight closer to mean value center is smaller,
The weight for more deviateing mean value center is bigger；The number interaction of each probability interval in right side is fallen on according to weight term and online residual sample
Construct health index HI
λkIndicate the weight term of setting, nkIndicate that online residual sample falls on the number of each probability interval in right side, i=1,2,3,4；
(3) by online residual error slide window processing, four HI are calculated to the sample point in each small window, and count HI1, HI2, HI3,
HI4Value；If HI in the time window3+HI4Greater than HI1+HI2, then W is denoted as to the time window1, otherwise it is denoted as W0；
To the W in all time windows0And W1It is counted, if most classes are W1, then judge Fans component hair in time T
Raw temperature anomaly failure, output Warning Sign parameter is true.
3. the detection of fan part temperature anomaly and alarm method, feature according to claim 2 with single measuring point exists
In, all Warning Sign parameters of the parts to be tested measuring point are exported into algorithm operation result database and are saved, each algorithm fortune
Row calls the Warning Sign parameter value of historical storage to carry out logic judgment；If continuously the Warning Sign of algorithm operation output is joined three times
Number be true or history n times operation in Warning Sign parameter be really be more than half, then this operation triggering alarm；Wherein, n
Size is determined according to algorithm running frequency.
4. the detection of fan part temperature anomaly and alarm method, feature according to claim 3 with single measuring point exists
In, trigger alarm when, set correspondence measuring point alarm identification parameter be it is true, the corresponding alarm identification parameter of measuring point is exported to calculation
It being saved in method operation result database, each algorithm operation calls the alarm identification parameter of historical storage to carry out logic judgment,
If this alarm identification parameter for running corresponding measuring point is that very, taking alarm identification parameter in history k times operation is genuine number
It determines the value-at-risk of this alarm, and ratio is mapped as alarm level；The size of k is determining according to algorithm running frequency, and k > >
n。
5. the detection of fan part temperature anomaly and alarm method, feature according to claim 4 with single measuring point exists
In after maintenance, the history early warning library of corresponding component is reset fan part.
6. the detection of fan part temperature anomaly and alarm side according to any one of claims 1-5 with single measuring point
Method, which is characterized in that the original variable the step of mechanism driving data modeling program in (2) include wind speed, active power,
At least one of generator speed, the measured value of temperature point and cabin temperature.
7. the detection of fan part temperature anomaly and alarm side according to claim 1 to 6 with single measuring point
Method, which is characterized in that the step of the mechanism driving data modeling program in (6), the judgment criteria of the fitting precision of model is logical
It crosses residual analysis and fitting precision analysis carries out；If fitting precision is less than the 10% of all data mean values, and residual error passes through
QQ-Norm is examined and Jarque-Bera is examined, and obeys normality distribution, then it is assumed that reach the requirement of accuracy rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910702600.5A CN110414155B (en) | 2019-07-31 | 2019-07-31 | Fan component temperature abnormity detection and alarm method with single measuring point |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910702600.5A CN110414155B (en) | 2019-07-31 | 2019-07-31 | Fan component temperature abnormity detection and alarm method with single measuring point |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110414155A true CN110414155A (en) | 2019-11-05 |
| CN110414155B CN110414155B (en) | 2022-09-30 |
Family
ID=68364808
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910702600.5A Active CN110414155B (en) | 2019-07-31 | 2019-07-31 | Fan component temperature abnormity detection and alarm method with single measuring point |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110414155B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110992205A (en) * | 2019-11-28 | 2020-04-10 | 中国船舶重工集团海装风电股份有限公司 | State detection method and system for generator winding of wind turbine generator and related components |
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| CN111075661A (en) * | 2019-12-25 | 2020-04-28 | 明阳智慧能源集团股份公司 | Method for judging health condition of main shaft bearing of wind turbine generator based on temperature change trend |
| CN111581072A (en) * | 2020-05-12 | 2020-08-25 | 国网安徽省电力有限公司信息通信分公司 | Disk failure prediction method based on SMART and performance log |
| CN111766514A (en) * | 2020-06-19 | 2020-10-13 | 南方电网调峰调频发电有限公司 | Data analysis method for equipment detection points |
| CN111781498A (en) * | 2020-06-19 | 2020-10-16 | 南方电网调峰调频发电有限公司 | Data analysis system of equipment detection point |
| CN111814848A (en) * | 2020-06-22 | 2020-10-23 | 浙江大学 | Self-adaptive early warning strategy design method for temperature fault of wind turbine generator |
| CN112211794A (en) * | 2020-09-02 | 2021-01-12 | 五凌电力有限公司新能源分公司 | Cabin temperature abnormity early warning method, device, equipment and medium of wind turbine generator |
| CN112734130A (en) * | 2021-01-21 | 2021-04-30 | 河北工业大学 | Fault early warning method for double-fed fan main shaft |
| CN113221453A (en) * | 2021-04-30 | 2021-08-06 | 华风数据(深圳)有限公司 | Fault monitoring and early warning method for output shaft of gearbox of wind turbine generator |
| CN113325824A (en) * | 2021-06-02 | 2021-08-31 | 三门核电有限公司 | Regulating valve abnormity identification method and system based on threshold monitoring |
| WO2022012137A1 (en) * | 2020-07-15 | 2022-01-20 | 上海电气风电集团股份有限公司 | Method and system for monitoring wind turbine generator set, and computer-readable storage medium |
| CN114151291A (en) * | 2021-11-18 | 2022-03-08 | 华能新能源股份有限公司 | Early fault monitoring method for wind turbine generator |
| CN114186666A (en) * | 2021-11-29 | 2022-03-15 | 中电华创(苏州)电力技术研究有限公司 | Generator coil temperature anomaly monitoring method based on self-standardization encoding and decoding |
| CN114251238A (en) * | 2021-11-30 | 2022-03-29 | 北京金风慧能技术有限公司 | Variable pitch motor temperature anomaly detection method and equipment |
| WO2022121298A1 (en) * | 2020-12-11 | 2022-06-16 | 青岛海尔空调器有限总公司 | Air conditioner alarm control method and apparatus, and electronic device and storage medium |
| WO2023005467A1 (en) * | 2021-07-29 | 2023-02-02 | 东方电气集团东方电机有限公司 | Temperature early warning method for stator winding |
| CN116167250A (en) * | 2023-04-23 | 2023-05-26 | 南京群顶科技股份有限公司 | Machine room environment assessment method based on temperature difference weighting and time sequence algorithm |
| US11920562B2 (en) | 2020-06-04 | 2024-03-05 | Vestas Wind Systems A/S | Temperature estimation in a wind turbine |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107977508A (en) * | 2017-11-29 | 2018-05-01 | 北京优利康达科技股份有限公司 | A kind of dynamo bearing failure prediction method |
| CN108680358A (en) * | 2018-03-23 | 2018-10-19 | 河海大学 | A kind of Wind turbines failure prediction method based on bearing temperature model |
| WO2018228648A1 (en) * | 2017-06-14 | 2018-12-20 | Kk Wind Solutions A/S | Independent monitoring system for a wind turbine |
-
2019
- 2019-07-31 CN CN201910702600.5A patent/CN110414155B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018228648A1 (en) * | 2017-06-14 | 2018-12-20 | Kk Wind Solutions A/S | Independent monitoring system for a wind turbine |
| CN107977508A (en) * | 2017-11-29 | 2018-05-01 | 北京优利康达科技股份有限公司 | A kind of dynamo bearing failure prediction method |
| CN108680358A (en) * | 2018-03-23 | 2018-10-19 | 河海大学 | A kind of Wind turbines failure prediction method based on bearing temperature model |
Non-Patent Citations (5)
| Title |
|---|
| XIAOHANG JIN等: "Anomaly Detection and Fault Prognosis for Bearings", 《 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| 丁佳煜等: "基于轴承温度模型的风电机组故障预测研究", 《可再生能源》 * |
| 何群等: "基于相关主成分分析和极限学习机的风电机组主轴承状态监测研究", 《计量学报》 * |
| 姚万业等: "基于Hadoop平台的风机群落故障预警", 《电力科学与工程》 * |
| 孙鹏等: "采用预测模型与模糊理论的风电机组状态参数异常辨识方法", 《电力自动化设备》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110991666A (en) * | 2019-11-25 | 2020-04-10 | 远景智能国际私人投资有限公司 | Fault detection method, model training method, device, equipment and storage medium |
| CN110991666B (en) * | 2019-11-25 | 2023-09-15 | 远景智能国际私人投资有限公司 | Fault detection method, training device, training equipment and training equipment for model, and storage medium |
| CN110992205A (en) * | 2019-11-28 | 2020-04-10 | 中国船舶重工集团海装风电股份有限公司 | State detection method and system for generator winding of wind turbine generator and related components |
| CN111075661A (en) * | 2019-12-25 | 2020-04-28 | 明阳智慧能源集团股份公司 | Method for judging health condition of main shaft bearing of wind turbine generator based on temperature change trend |
| CN111075661B (en) * | 2019-12-25 | 2021-11-09 | 明阳智慧能源集团股份公司 | Method for judging health condition of main shaft bearing of wind turbine generator based on temperature change trend |
| CN111581072A (en) * | 2020-05-12 | 2020-08-25 | 国网安徽省电力有限公司信息通信分公司 | Disk failure prediction method based on SMART and performance log |
| CN111581072B (en) * | 2020-05-12 | 2023-08-15 | 国网安徽省电力有限公司信息通信分公司 | Disk fault prediction method based on SMART and performance log |
| US11920562B2 (en) | 2020-06-04 | 2024-03-05 | Vestas Wind Systems A/S | Temperature estimation in a wind turbine |
| CN111766514A (en) * | 2020-06-19 | 2020-10-13 | 南方电网调峰调频发电有限公司 | Data analysis method for equipment detection points |
| CN111781498A (en) * | 2020-06-19 | 2020-10-16 | 南方电网调峰调频发电有限公司 | Data analysis system of equipment detection point |
| CN111814848A (en) * | 2020-06-22 | 2020-10-23 | 浙江大学 | Self-adaptive early warning strategy design method for temperature fault of wind turbine generator |
| CN111814848B (en) * | 2020-06-22 | 2024-04-09 | 浙江大学 | Self-adaptive early warning strategy design method for temperature faults of wind turbine generator |
| WO2022012137A1 (en) * | 2020-07-15 | 2022-01-20 | 上海电气风电集团股份有限公司 | Method and system for monitoring wind turbine generator set, and computer-readable storage medium |
| CN112211794A (en) * | 2020-09-02 | 2021-01-12 | 五凌电力有限公司新能源分公司 | Cabin temperature abnormity early warning method, device, equipment and medium of wind turbine generator |
| WO2022121298A1 (en) * | 2020-12-11 | 2022-06-16 | 青岛海尔空调器有限总公司 | Air conditioner alarm control method and apparatus, and electronic device and storage medium |
| CN112734130B (en) * | 2021-01-21 | 2022-06-10 | 河北工业大学 | Fault early warning method for double-fed fan main shaft |
| CN112734130A (en) * | 2021-01-21 | 2021-04-30 | 河北工业大学 | Fault early warning method for double-fed fan main shaft |
| CN113221453A (en) * | 2021-04-30 | 2021-08-06 | 华风数据(深圳)有限公司 | Fault monitoring and early warning method for output shaft of gearbox of wind turbine generator |
| CN113325824A (en) * | 2021-06-02 | 2021-08-31 | 三门核电有限公司 | Regulating valve abnormity identification method and system based on threshold monitoring |
| WO2023005467A1 (en) * | 2021-07-29 | 2023-02-02 | 东方电气集团东方电机有限公司 | Temperature early warning method for stator winding |
| CN114151291A (en) * | 2021-11-18 | 2022-03-08 | 华能新能源股份有限公司 | Early fault monitoring method for wind turbine generator |
| CN114186666A (en) * | 2021-11-29 | 2022-03-15 | 中电华创(苏州)电力技术研究有限公司 | Generator coil temperature anomaly monitoring method based on self-standardization encoding and decoding |
| CN114186666B (en) * | 2021-11-29 | 2023-10-13 | 中电华创(苏州)电力技术研究有限公司 | Generator coil temperature anomaly monitoring method based on self-standardized encoding and decoding |
| CN114251238A (en) * | 2021-11-30 | 2022-03-29 | 北京金风慧能技术有限公司 | Variable pitch motor temperature anomaly detection method and equipment |
| CN116167250A (en) * | 2023-04-23 | 2023-05-26 | 南京群顶科技股份有限公司 | Machine room environment assessment method based on temperature difference weighting and time sequence algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110414155B (en) | 2022-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110414155A (en) | A kind of detection of fan part temperature anomaly and alarm method with single measuring point | |
| CN106872657B (en) | A kind of multivariable water quality parameter time series data accident detection method | |
| CN108460144B (en) | Coal equipment fault early warning system and method based on machine learning | |
| CN104573850B (en) | A kind of Power Plant Equipment state evaluating method | |
| CN107949812B (en) | Method for detecting anomalies in a water distribution system | |
| CN110414154A (en) | A kind of detection of fan part temperature anomaly and alarm method with double measuring points | |
| CN109543210A (en) | A kind of Wind turbines failure prediction system based on machine learning algorithm platform | |
| CN107844067B (en) | A kind of gate of hydropower station on-line condition monitoring control method and monitoring system | |
| CN109186813A (en) | A kind of temperature sensor self-checking unit and method | |
| CN106407589B (en) | Fan state evaluation and prediction method and system | |
| CN106019195B (en) | Electric power measurement automation verification assembly line fault diagnosis system | |
| CN109298697A (en) | Thermal power plant's various parts working state evaluation method based on DBM Dynamic Baseline Model | |
| CN112561736A (en) | Fault diagnosis system and method for relay protection device of intelligent substation | |
| Butler et al. | Exploiting SCADA system data for wind turbine performance monitoring | |
| CN113705924B (en) | Intelligent diagnosis method and system for thermal control equipment | |
| CN102789676B (en) | Method for designing industrial alarm on basis of alarm evidence fusion | |
| CN107038453A (en) | The fault early warning system and determination methods of a kind of Wind turbines | |
| CN109583520A (en) | A kind of state evaluating method of cloud model and genetic algorithm optimization support vector machines | |
| CN102062832A (en) | Power system online prewarning method based on microdisturbance signal low-frequency oscillation mode identification | |
| CN108846511A (en) | A kind of defect of transformer equipment trend analysis based on regulation big data platform | |
| CN110580492A (en) | Track circuit fault precursor discovery method based on small fluctuation detection | |
| CN116292367B (en) | Power plant fan system abnormal condition detection method based on one-dimensional convolution | |
| CN105910806A (en) | Filling pump early health status monitoring method | |
| CN114004059B (en) | Health portrait method for hydroelectric generating set | |
| CN207992717U (en) | A kind of gate of hydropower station on-line condition monitoring system |
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 |