CN110298132A - A kind of turbine-generator units bearing throw trend forecasting method - Google Patents
A kind of turbine-generator units bearing throw trend forecasting method Download PDFInfo
- Publication number
- CN110298132A CN110298132A CN201910604666.0A CN201910604666A CN110298132A CN 110298132 A CN110298132 A CN 110298132A CN 201910604666 A CN201910604666 A CN 201910604666A CN 110298132 A CN110298132 A CN 110298132A
- Authority
- CN
- China
- Prior art keywords
- model
- turbine
- throw
- arma
- generator units
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The present invention is suitable for hydraulic generator technical group field, provide a kind of turbine-generator units bearing throw trend forecasting method, with the autoregressive moving-average model in time series models predicted method, can Accurate Prediction unit bearing throw development trend, realize the early warning of hydro-generating units initial failure.Arma modeling prediction bearing throw trend is compared with other predicted methods with strong to the annotation ability of linear model and stable data, precision of prediction is high simultaneously；Also have that data requirements is less, model structure simple computation amount is few, and it is more accurate to annotate in applying ability.This method can expand the development trends such as bearing liner temperature and the frame vibration of prediction turbine-generator units, improve unit operation stability.
Description
Technical field
The invention belongs to hydrogenerator technical group fields, provide a kind of turbine-generator units bearing throw trend prediction
Method.
Background technique
Turbine-generator units guide bearing is primarily subjected to the imbalance of the radial direction of unit rotating part mechanically and electrically magnetic force, makes machine
Group is run in defined throw and oscillating region.Studies have shown that mechanical factor failure is to lead to vibration of hydrogenerator set event
One of principal element of barrier, including rotor and stator rub, and rotor portion quality is unequal and shafting part misaligns all
Unit operation conditions can be caused abnormal.No matter the hydraulic generator of various types will overcome the unstable of operation, it at
To have to solve the problems, such as in unit design, manufacture, installation, operation and maintenance.Turbine-generator units, which break down, not only can
The power supply quality of unit is reduced, threatens unit safety operation and its service life, and the working environment of hydroelectric power plant can be deteriorated.
Fault Diagnosis of Hydroelectric Generating Set research achievement is not met by field demand at present, alarmed with static threshold based on
The Fault Diagnosis of Hydroelectric Generating Set of judgment basis is wanted to be difficult to meet initial failure pre-alerting ability.Therefore, how accurately to predict
Unit bearing throw and vibration signal development trend preferably judge the true operating status of unit, are to realize turbine-generator units
The important research content of equipment initial failure early warning, and realize the demand of hydro-generating units " preventative maintenance ".
Summary of the invention
The embodiment of the invention provides a kind of turbine-generator units bearing throw trend forecasting methods, are slided based on autoregression
Averaging model is predicted come the throw trend to turbine-generator units bearing.
The invention is realized in this way a kind of turbine-generator units bearing throw trend forecasting method, the method are specific
Include the following steps:
S1, the data sample for choosing sampling interval identical turbine-generator units bearing throw；
S2, it is based on timing { QtConstructing ARMA (n, m) model, ARMA (n, m) model is that the sliding of n rank autoregression m rank is flat
Equal model；
If S3, m or n are equal to 0, autoregression model AR (n) and moving average model MA (m) can be obtained respectively；
S4, the least-squares estimation for solving parameter matrix in autoregression model AR (n)
S5, it is based on same timing { QtIt is fitted autoregression model AR (p) and ARMA (n, m) model respectively；
S6, automatic returning coefficient is calculated based on autoregression model AR (p)And residual sequence { at, by the residual sequence
{atIt is assigned to ARMA (n, m) model；
S7, the model parameter matrix β for solving arma modeling, model parameter matrix β is by automatic returning coefficientIt is put down with mobile
Equal coefficient θ composition；
S8, detection residual sequence { atWhether be white noise, if testing result be it is yes, throw is determined based on AIC criterion
The order p of data sample prediction model, the i.e. order of arma modeling；
S9, turbine-generator units bearing throw trend is predicted based on throw data sample prediction model.
Further, if residual sequence { atAuto-correlation coefficient ρa,kMeet formula (1), even residual sequence { atIt is white
Noise, formula (1) are specific as follows
Wherein, Qt、Qt-kFor timing { QtIn sample data.
Further, determine that the order P of arma modeling, formula (2) are specific as follows based on formula (2):
P is the quantity of arma modeling parameter in formula, and N is the total length of sequence,For the variance of residual error, AIC (P) minimum value
Corresponding P value is the order of arma modeling.
Turbine-generator units bearing throw trend forecasting method provided by the invention has the following beneficial effects:
1., can Accurate Prediction unit bearing throw with the autoregressive moving-average model in time series models predicted method
Development trend, realize the early warning of hydro-generating units initial failure；
2.ARMA model prediction bearing throw trend has the annotation to linear model and stable data compared with other predicted methods
The advantages that ability is strong, precision of prediction is high；In addition, also have the characteristics that data requirements is less, model structure simple computation amount is few, and
It is more accurate that ability is annotated in applying；
3. the development trends such as bearing liner temperature and the frame vibration of prediction turbine-generator units can be expanded, it is steady to improve unit operation
It is qualitative.
Detailed description of the invention
Fig. 1 is turbine-generator units bearing throw trend forecasting method flow chart provided in an embodiment of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Fig. 1 is turbine-generator units bearing throw trend forecasting method flow chart provided in an embodiment of the present invention, this method
Specifically comprise the following steps:
Step 1: first choosing the data sample of sampling interval identical turbine-generator units bearing throw, then pass through observation sample
The autocorrelogram and partial autocorrelation figure of notebook data carry out decision model classification；
In embodiments of the present invention, model classification includes: AR model, MA model, arma modeling, wherein AR model and MA
Model is two special cases as arma modeling m=0 or n=0 respectively.The correlation properties of time series are judgment models classifications
Important Theoretic Foundation.If the auto-correlation coefficient of data sample is truncated suddenly when lagging m+1 rank, that is, claim its truncation at m, that
MA (m) model can be established to the sequence.Equally, if the truncation at n of the PARCOR coefficients of data sample, so that it may
AR (n) model is established to the sequence.If truncation state is not presented for the auto-correlation coefficient and PARCOR coefficients of data sample,
It is all slowly drawn close from certain value to zero, that is, is in hangover state, so that it may which ARMA (n, m) model is established to the sequence.
Step 2: determining arma modeling.For the timing { Q of steady, normal state, zero-meant, if QtValue not only with preceding n
Each value Q of stept-1,Qt-2,…,Qt-nIt is related, but also each interference a with preceding m stept-1,at-2,…,at-mRelated (n, m
=1,2 ...) thought of linear regression, is then pressed, available arma modeling:
Formula (1) indicates a n rank autoregression m rank moving average model, is denoted as ARMA (n, m)；N, m respectively indicates the part AR
With the order of the part MA；Real parameterFor autoregressive coefficient, θj(j=1,2 ..., m) is rolling average coefficient,
It is all the parameter to be estimated of model；Random entry atIt is independent from each other white noise sequence, and obedience mean value is 0, variance isJust
State distribution.
When in ARMA (n, m) n or m be equal to 0 when, just obtained two special cases of arma modeling, i.e. autoregression model AR
(n) and moving average model MA (m).The mathematical description difference of the two models is as follows:
As m=0, AR (n) model:
As n=0, MA (m) model:
Qt=-θ1at-1-θ2at-2-…-θmat-m+at (3)
The estimation of step 3:AR model parameter.For AR model parameter estimation, using least squares estimate.According to polynary
The theory of linear regression forms several data before the one section of swing of water turbine generator set data and each data in timing
Multivariate linear equations are as follows:
It can be indicated with the form of matrix are as follows:
In formula:
The least-squares estimation of parameter matrix are as follows:
Wherein,Be solved by least square method come AR model in parameter to be askedThat is the matrix of autoregressive coefficient
Estimated value；
The estimation of step 4:ARMA model parameter.To timing { QtIt is fitted autoregression model AR (p) and ARMA (n, m) mould
Type, wherein p is the number of parameters of AR model, i.e., is only the number for returning member；Residual values a of the two kinds of models in synchronizationt
It is theoretically equal.It therefore, can be first to timing { QtFitting AR (p) model, calculate the sequence { a of residual errort, by this
{atIt is used as ARMA (n, m) model residual sequence { at, thus { a in ARMA (n, m) modeltBecome it is known that such ARMA mould
Shape parameter estimation procedure becomes linear and obtains regression process, then Least Square Method parameter can be used.
To timing { QtIt is first fitted AR (p) (p >=n+m) model, calculate model parameterAfterwards,It is all
Parameter to be askedThe matrix of composition,It is its corresponding numerical value.
Residual sequence is calculated by AR (p) model:
By { atIt is updated to ARMA (n, m) model, following matrix equation can be obtained:
Y=Q β+A (9)
In formula,
Least Square Method can also be used in model parameter β,
β=(QTQ)-1QTY (11)
Step 5: the applicability of model is examined.In theory, the basic condition that arma modeling is set up is residual sequence
{atIt is white noise, therefore, the most basic test criterion of model applicability is to examine { atIt whether is white noise.Examine residual error
Sequence { atAuto-correlation coefficient ρa,kFormula (12) can be met, if not satisfied, residual sequence { atIt is not white noise sequence, just
Illustrate not to be extracted there are also some important informations, model should be reset, pure randomness test can be carried out to residual error.
Wherein, QtAnd Qt-kFor timing { QtIn data.
Model applicability inspection is carried out according to AIC criterion (information criterion), defines criterion function:
AIC=-2lnL+2P (13)
In formula, P is the quantity of arma modeling parameter, i.e., the sum of autoregressive order and sliding average order；L is timing
{QtLikelihood function, as { QtWhen being steady, normal state time series,
In formulaIt is { QtIn the conditional mathematical expectation estimated value of t moment, thereforeSubstitution formula (14) is gone forward side by side
Row is tired to multiply calculating, can obtain:
It enablesSubstitution formula (15) simultaneously takes natural logrithm to equal sign both sides, arrangement:
That is:
Formula (17) are substituted into formula (13), for given data length N, latter two are constant in this formula, have no effect on AIC
(P) comparison result, therefore can omit, thus:
In formula, N is the total length of sequence,For the variance of residual error.
Under conditions of model parameter estimation method is given, AIC (P) is the function of model order P, when P increases,
Decline, but latter 2P increases, and therefore, taking model order P when AIC (P) value minimum is applicable models order.
Step 6: by constantly analyzing, the model order of throw data sample and determination are finally accurately determined with AIC criterion
Model.
Step 7: turbine-generator units bearing throw trend being predicted according to fixed model.
Turbine-generator units bearing throw trend forecasting method provided by the invention has the following beneficial effects:
1., can Accurate Prediction unit bearing throw with the autoregressive moving-average model in time series models predicted method
Development trend, realize the early warning of hydro-generating units initial failure；
2.ARMA model prediction bearing throw trend has the annotation to linear model and stable data compared with other predicted methods
The advantages that ability is strong, precision of prediction is high；In addition, also have the characteristics that data requirements is less, model structure simple computation amount is few, and
It is more accurate that ability is annotated in applying；
3. the development trends such as bearing liner temperature and the frame vibration of prediction turbine-generator units can be expanded, it is steady to improve unit operation
It is qualitative.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (3)
1. a kind of turbine-generator units bearing throw trend forecasting method, which is characterized in that the method specifically includes following step
It is rapid:
S1, the data sample for choosing sampling interval identical turbine-generator units bearing throw；
S2, it is based on timing { QtConstructing ARMA (n, m) model, ARMA (n, m) model is n rank autoregression m rank sliding average mould
Type；
If S3, m or n are equal to 0, autoregression model AR (n) and moving average model MA (m) are obtained respectively；
S4, the least-squares estimation for solving parameter matrix in autoregression model AR (n)
S5, it is based on same timing { QtIt is fitted autoregression model AR (p) and ARMA (n, m) model respectively；
S6, automatic returning coefficient is calculated based on autoregression model AR (p)And residual sequence { at, by the residual sequence { at}
It is assigned to ARMA (n, m) model；
S7, the model parameter matrix β for solving arma modeling, model parameter matrix β is by automatic returning coefficientWith rolling average system
Number θ composition；
S8, detection residual sequence { atWhether be white noise, if testing result be it is yes, throw data sample is determined based on AIC criterion
The order P of this prediction model；
S9, turbine-generator units bearing throw trend is predicted based on throw data sample prediction model.
2. turbine-generator units bearing throw trend forecasting method as described in claim 1, which is characterized in that
If residual sequence { atAuto-correlation coefficient ρa,kMeet formula (1), even residual sequence { atIt is white noise, formula (1)
It is specific as follows
Wherein, Qt、Qt-kFor timing { QtIn sample data.
3. turbine-generator units bearing throw trend forecasting method as described in claim 1, which is characterized in that be based on formula (2)
To determine that the order P of arma modeling, formula (2) are specific as follows:
P is the quantity of arma modeling parameter in formula, and N is the total length of sequence,For the variance of residual error, AIC (P) minimum value is corresponded to
P value be arma modeling order.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910604666.0A CN110298132A (en) | 2019-07-05 | 2019-07-05 | A kind of turbine-generator units bearing throw trend forecasting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910604666.0A CN110298132A (en) | 2019-07-05 | 2019-07-05 | A kind of turbine-generator units bearing throw trend forecasting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110298132A true CN110298132A (en) | 2019-10-01 |
Family
ID=68030438
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910604666.0A Pending CN110298132A (en) | 2019-07-05 | 2019-07-05 | A kind of turbine-generator units bearing throw trend forecasting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110298132A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633583A (en) * | 2020-12-29 | 2021-04-09 | 南方电网调峰调频发电有限公司 | Generator set vibration prediction method and device, computer equipment and storage medium |
| CN113297680A (en) * | 2021-06-21 | 2021-08-24 | 中国航发沈阳发动机研究所 | Method for analyzing performance trend of small-bypass-ratio aircraft gas engine |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268424A (en) * | 2014-10-12 | 2015-01-07 | 刘岩 | Comprehensive subway energy consumption forecasting method based on time sequence |
| CN105703954A (en) * | 2016-03-17 | 2016-06-22 | 福州大学 | Network data flow prediction method based on ARIMA model |
| CN108875841A (en) * | 2018-06-29 | 2018-11-23 | 国家电网有限公司 | A kind of pumped storage unit vibration trend forecasting method |
-
2019
- 2019-07-05 CN CN201910604666.0A patent/CN110298132A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104268424A (en) * | 2014-10-12 | 2015-01-07 | 刘岩 | Comprehensive subway energy consumption forecasting method based on time sequence |
| CN105703954A (en) * | 2016-03-17 | 2016-06-22 | 福州大学 | Network data flow prediction method based on ARIMA model |
| CN108875841A (en) * | 2018-06-29 | 2018-11-23 | 国家电网有限公司 | A kind of pumped storage unit vibration trend forecasting method |
Non-Patent Citations (2)
| Title |
|---|
| 夏鑫: "水轮发电机组模型参数辨识与故障诊断方法研究", 《优秀博士学位论文全文数据库》 * |
| 沈冰等: "上海市原静安区成人流感样病例就诊百分比预测的自回归求和滑动平均模型构建与应用", 《上海预防医学》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112633583A (en) * | 2020-12-29 | 2021-04-09 | 南方电网调峰调频发电有限公司 | Generator set vibration prediction method and device, computer equipment and storage medium |
| CN113297680A (en) * | 2021-06-21 | 2021-08-24 | 中国航发沈阳发动机研究所 | Method for analyzing performance trend of small-bypass-ratio aircraft gas engine |
| CN113297680B (en) * | 2021-06-21 | 2023-08-08 | 中国航发沈阳发动机研究所 | Performance trend analysis method for aviation gas engine with small bypass ratio |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Moghadam et al. | Online condition monitoring of floating wind turbines drivetrain by means of digital twin | |
| Rezamand et al. | Improved remaining useful life estimation of wind turbine drivetrain bearings under varying operating conditions | |
| Dao | Condition monitoring and fault diagnosis of wind turbines based on structural break detection in SCADA data | |
| Teng et al. | DNN‐based approach for fault detection in a direct drive wind turbine<? show [AQ ID= Q1]?> | |
| Janjarasjitt et al. | Bearing condition diagnosis and prognosis using applied nonlinear dynamical analysis of machine vibration signal | |
| Wilkinson et al. | Comparison of methods for wind turbine condition monitoring with SCADA data | |
| Ocak et al. | Online tracking of bearing wear using wavelet packet decomposition and probabilistic modeling: A method for bearing prognostics | |
| Butler et al. | Exploiting SCADA system data for wind turbine performance monitoring | |
| CN111311059B (en) | Waterwheel house fault diagnosis method based on knowledge graph | |
| CN110298132A (en) | A kind of turbine-generator units bearing throw trend forecasting method | |
| CN112487910A (en) | Fault early warning method and system for nuclear turbine system | |
| Li et al. | Residual useful life estimation by a data‐driven similarity‐based approach | |
| CN107516279A (en) | A kind of method of network public-opinion automatic early-warning | |
| CN110361193A (en) | Method for distinguishing is known for wind generating set pitch control bearing fault | |
| Olsson et al. | A data-driven approach for predicting long-term degradation of a fleet of micro gas turbines | |
| Barbieri et al. | Sensor-based degradation prediction and prognostics for remaining useful life estimation: Validation on experimental data of electric motors | |
| Wang et al. | A hidden semi‐markov model with duration‐dependent state transition probabilities for prognostics | |
| Li et al. | Optimal Bayesian maintenance policy for a gearbox subject to two dependent failure modes | |
| Cevasco et al. | O&M cost-Based FMECA: Identification and ranking of the most critical components for 2-4 MW geared offshore wind turbines | |
| de N Santos et al. | Long-term fatigue estimation on offshore wind turbines interface loads through loss function physics-guided learning of neural networks | |
| Garlick et al. | A model-based approach to wind turbine condition monitoring using SCADA data | |
| CN108074045B (en) | Wind turbine generator complete machine vulnerability analysis and fault sequencing method and electronic terminal | |
| Haghani et al. | Data-driven multimode fault detection for wind energy conversion systems | |
| Alozie et al. | An adaptive model-based framework for prognostics of gas path faults in aircraft gas turbine engines | |
| Wang et al. | Maintenance decision based on data fusion of aero engines |
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 |