CN111415070A - Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data - Google Patents
Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data Download PDFInfo
- Publication number
- CN111415070A CN111415070A CN202010139476.9A CN202010139476A CN111415070A CN 111415070 A CN111415070 A CN 111415070A CN 202010139476 A CN202010139476 A CN 202010139476A CN 111415070 A CN111415070 A CN 111415070A
- Authority
- CN
- China
- Prior art keywords
- wind turbine
- data
- value
- oil temperature
- early warning
- 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
- 238000000034 method Methods 0.000 title claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 54
- 238000000605 extraction Methods 0.000 claims abstract description 13
- 230000009467 reduction Effects 0.000 claims abstract description 10
- 239000013598 vector Substances 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 8
- 238000013461 design Methods 0.000 claims description 5
- 238000002790 cross-validation Methods 0.000 claims description 4
- 230000003247 decreasing effect Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 abstract description 7
- 238000012423 maintenance Methods 0.000 abstract description 4
- 239000003921 oil Substances 0.000 description 35
- 238000012544 monitoring process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 5
- 239000003086 colorant Substances 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 239000012208 gear oil Substances 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 238000005299 abrasion Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000010687 lubricating oil Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010792 warming Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Tourism & Hospitality (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data, which comprises an off-line training stage and an on-line application stage, wherein a related data set of a turbine generator is selected, the gearbox oil temperature is selected as a target variable, other variables are subjected to dimension reduction through an Autoencoder model for feature extraction, then an SVR is selected as a variable estimation model for training, real-time estimation of the target variable is realized, in the on-line application stage, a real-time running data set of the wind turbine generator is input into the SVR model after training is completed, a residual sequence of a target variable actual running value and a model estimation value is obtained, an EWMA is used for setting a residual sequence over-limit threshold, and a continuous over-limit early warning criterion is designed to realize the early sensing of the gearbox oil temperature over-temperature fault. The invention can realize the early sensing and early warning of the over-temperature fault of the oil temperature of the gearbox of the wind turbine generator, avoid the fault shutdown caused by the abnormal evolution of the components of the gearbox, reduce the shutdown time and the maintenance loss, and has stronger theoretical performance and practicability.
Description
Technical Field
The invention relates to the technical field of wind turbine generator fault early warning, in particular to a wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data.
Background
The wind power resource has large reserves and wide distribution, the cost of wind power generation is low, and the wind power generation is pollution-free, and is an environment-friendly clean energy, so in recent years, the research investment heat of the wind power generation in the global scope is high, and the installed capacity of a fan is also increased year by year. With the increase of the installed capacity, the operation and maintenance problems of the wind turbine are gradually emphasized.
The operation and maintenance problems of the fan mainly come from the high failure rate of the fan, and the reason of the high failure rate is that the fan parts which are put into use at the early stage are aged, so that the performance of the whole fan is reduced; secondly, with global warming, greenhouse effect and frequent extreme weather, the running environment of the fan is worse and worse, and a plurality of parts can be damaged and fail due to weather. The high failure rate of the fan can lead to frequent shutdown of the fan, and the generating efficiency is reduced, thereby causing great economic loss.
Among all subsystems of the wind turbine, the gear box subsystem plays a role of transmitting power of a wind wheel to a generator, is one of the most main components for converting wind energy into electric energy, and is also the subsystem with the highest failure rate in the wind turbine. The fan gear box is composed of a plurality of mechanical components, including easily damaged components such as bearings and gears, and the abrasion and the abnormality of the components can generally cause the abnormal rise of the oil temperature in the gear box, thereby affecting the operation of the whole fan.
Therefore, the early warning of the over-temperature fault of the oil temperature of the fan gear box is realized, and the great contribution is made to the improvement of the power generation benefit and the safety performance of the fan.
The traditional method for monitoring the fault of the gearbox is realized by analyzing vibration signals in a certain frequency range, but the vibration sensors are required to be additionally arranged on the monitored gearbox one by one, and the method is difficult to be widely applied to the wind power industry due to troublesome construction and overhigh cost.
At present, wind turbines are all provided with a supervisory control and data acquisition System (SCADA), the power and the rotating speed of the wind turbines and the temperature and other parameters of large components such as a gear box and a generator can be monitored, the running states of the main components are simply judged, but generally only a fixed threshold value is set for a single parameter, and the timeliness is poor.
The gear box fault monitoring idea based on the SCADA data is that the real-time state data collected by the gear box sensor is subjected to model monitoring through an SCADA system connected with a wind turbine generator set, so that the purpose of gear box early warning is achieved.
The temperature of lubricating oil of the gearbox is a main monitoring index of the gearbox during monitoring, but the temperature of the gearbox is easily influenced by the surrounding environment and noise, so that the monitoring result is influenced.
Disclosure of Invention
The invention aims to overcome the technical problems that a vibration sensor is required to be additionally arranged for monitoring the fault of the gear box in the traditional method in the prior art, the construction is troublesome, the cost is overhigh, and the temperature of the gear oil is easily influenced by the surrounding environment and noise, and provides the over-temperature fault early warning method for the gear box oil temperature of the wind turbine generator based on the SCADA data.
In order to achieve the purpose, the invention adopts the following technical scheme:
a wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data comprises an off-line training stage and an on-line application stage, and comprises the following steps:
s1, performing an off-line training stage;
s2, data set collection, selection and marking are carried out: acquiring and recording data in a wind turbine SCADA system needing to perform over-temperature fault early warning of the oil temperature of the gearbox, selecting and marking related data sets, and selecting the oil temperature of the gearbox as a target variable;
s3, feature extraction: screening out features with strong correlation with target variables, selecting a self-coding Autoencoder model for dimension reduction and feature extraction based on all other variables of an SCADA system data set, and converting high-dimensional SCADA information into low-dimensional feature vectors;
s4, reconstructing a training set, and selecting a variable estimation model: constructing a training set by using the target variable and the characteristic variable after dimensionality reduction, and selecting a Support Vector Regression (SVR) as a variable estimation model;
s5, training the selected variable estimation model;
s6, carrying out residual error training and obtaining a residual error sequence;
s7, setting a threshold upper limit: counting indexes, setting a threshold upper limit of a residual sequence according to an EWMA model, judging whether the wind turbine needs to enter an online application stage, entering a step S8 when the wind turbine needs to enter the online application stage, and otherwise, ending the process;
s8, performing an online application stage;
s9, real-time estimation: repeating the step S2 to the step S5, inputting the characteristic variables into the trained variable estimation model, and obtaining the real-time estimation value of the target variable;
s10, obtaining a real-time operation residual error: repeating the residual training in step S6;
s11, real-time judgment, and design of judgment criteria according to the step S7: designing a continuous overrun judgment criterion and setting a judgment criterion threshold value C;
and S12, obtaining a final early warning result.
In the scheme of the invention, because other variables of the wind turbine SCADA system have a large amount of redundant information for early warning of the over-temperature fault of the oil temperature of the gearbox and the coupling relation among the parameters is complex, an Autoencoder model is used for reducing the dimension of the variables, the characteristics with strong correlation with the target variables are extracted, and the training efficiency of the variable estimation model is improved; an overrun threshold is set by using the EWMA, so that data fluctuation is smoothed, and the influence of an isolated abnormal value is avoided; and a continuous overrun judgment criterion is designed, so that false alarm caused by data fluctuation is reduced, and the accuracy of an early warning result is ensured.
Preferably, the related data set in step S2 is selected and marked as a data set in which the selected wind turbine is in a normal operation period, and is marked as a training set, and the related data set in step S9 is selected as a real-time operation data set of the selected wind turbine in step S2, and is marked as a test set.
Preferably, in step S4, the support vector regression SVR selects a radial basis function RBF, performs SVR parameter selection by using a cross validation grid search method, and determines a parameter model with the minimum error.
Preferably, the residual training in step S6 includes the following steps:
the method comprises the following steps of firstly, estimating the oil temperature of the gearbox in real time, and representing the normal operation state of the oil temperature of the gearbox;
subtracting the training model output of the variable estimation model from the actual operation value of the target variable, namely the target variable estimation value, to obtain a target variable estimation residual error;
and thirdly, calculating a residual error estimated value by using an exponential weighted moving average EWMA estimated value calculation formula.
Preferably, the calculation formula of the exponentially weighted moving average EWMA estimated value in the third step is as follows:
EWMA(i)＝βEWMA(i-1)+(1-β)T(i)；
β represents the weighted decreasing rate, EWMA (i-1) is the EWMA estimated value of the previous data point of the data point i, the initial value EWMA (0) ═ 0, and T (i) is the actual residual value of the data point i.
Preferably, the setting of the threshold upper limit in step S7 includes the following steps:
step one, calculating standard deviation sigma of training set residual sequencetrain；
Step two, for the data point i, setting the upper limit of the estimated residual error threshold of the data point i to be equal to the EWMA estimated value EWMA (i) plus three times of standard deviation sigmatrainI.e. u (i) ═ ewma (i) +3 σtrainWherein U (i) is a threshold upper limit.
Because the oil temperature in the over-temperature fault of the oil temperature of the gearbox is abnormally increased, only the upper limit of the threshold value is considered, and the lower limit of the threshold value is not considered;
preferably, the real-time residual value R is obtained by subtracting the model estimation value from the actual operation value of the target variable in step S10i。
Preferably, the step S11 of designing the criterion includes the following steps:
step one, when the estimated residual error value of a data point i is larger than the upper threshold limit of the data point i, namely RiWhen the value is greater than U (i), calculating the continuous number Q (i) of data points of which the estimated residual errors appearing before the data points i are greater than the corresponding threshold upper limit;
step two, when the condition Q (i) of the discrimination criterion is satisfied and is not less than C, alarming is carried out at a data point i;
step three, when the condition Q (i) < C is satisfied, or the estimated residual error of the data point i is not more than the upper threshold limit of the data point i, namely RiWhen the value is less than or equal to U (i), no alarm is given at a data point i.
Because the data of the wind turbine generator has volatility, in order to avoid false alarm caused by isolated abnormal points, a continuous overrun judgment criterion needs to be designed and a judgment criterion threshold value C needs to be set,
the invention has the beneficial effects that:
1) aiming at the condition that a large amount of redundant information for early warning of the over-temperature fault of the oil temperature of the gearbox exists in the variables recorded by the SCADA system, dimension reduction and feature extraction are carried out on the basis of an Autoencoder model, features with strong correlation with target variables are screened out, and the efficiency and accuracy of variable estimation model training are guaranteed;
2) aiming at the fluctuation of a target variable, an EWMA is used for calculating a residual error estimation value and estimating the upper limit of a residual error threshold, the fluctuation of the variable is smoothed, and the influence of an isolated abnormal value is avoided;
3) based on the estimated residual sequence, a continuous overrun judgment criterion is designed, so that false alarms caused by variable fluctuation are reduced, and the accuracy of an early warning result is ensured.
Drawings
FIG. 1 is a flow chart of a fault early warning method for over-temperature of oil temperature of a gearbox of a wind turbine generator system.
FIG. 2 is a graph of variable estimation model fit results in an embodiment of the invention.
FIG. 3 is a diagram of a target variable estimation residual in an embodiment of the invention.
Fig. 4 is a diagram of an early warning result in an embodiment of the invention.
Detailed Description
The technical scheme of the invention is further specifically described by the following embodiments and the accompanying drawings.
Example 1: according to the wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data, as shown in FIG. 1, the method comprises an offline training stage and an online application stage, and the method comprises the following steps:
s1, performing an off-line training stage;
s2, data set collection, selection and marking are carried out: acquiring and recording data in an SCADA system of the wind turbine generator, which needs to perform over-temperature fault early warning of the oil temperature of the gear box, selecting and marking related data sets, selecting the oil temperature of the gear box as a target variable, selecting and marking the related data sets to represent that the data sets of the wind turbine generator in a normal operation period are selected and marked as training sets,
s3, feature extraction: and screening out features with strong correlation with the target variable, selecting a self-coding Autoencoder model for dimension reduction and feature extraction based on all other variables of the SCADA system data set, and converting high-dimensional SCADA information into low-dimensional feature vectors.
S4, reconstructing a training set, and selecting a variable estimation model: constructing a training set by using the target variable and the characteristic variable after dimensionality reduction, selecting a Support Vector Regression (SVR) as a variable estimation model, selecting a Radial Basis Function (RBF) by the Support Vector Regression (SVR), selecting SVR parameters by using a cross validation grid search method, and determining a parameter model with the minimum error.
And S5, training the selected variable estimation model.
And S6, residual error training is carried out, and a residual error sequence is obtained.
The residual training in step S6 includes the following steps:
the method comprises the following steps of firstly, estimating the oil temperature of the gearbox in real time, and representing the normal operation state of the oil temperature of the gearbox;
subtracting the training model output of the variable estimation model from the actual operation value of the target variable, namely the target variable estimation value, to obtain a target variable estimation residual error;
and thirdly, calculating a residual error estimated value by using an exponential weighted moving average EWMA estimated value calculation formula.
The calculation formula of the exponentially weighted moving average EWMA estimated value in the third step is as follows:
EWMA(i)＝βEWMA(i-1)+(1-β)T(i)；
β represents the weighted decreasing rate, EWMA (i-1) is the EWMA estimated value of the previous data point of the data point i, the initial value EWMA (0) ═ 0, and T (i) is the actual residual value of the data point i.
S7, setting a threshold upper limit: and counting indexes, setting a threshold upper limit of the residual sequence according to the EWMA model, judging whether the wind turbine needs to enter an online application stage, entering a step S8 when the wind turbine needs to enter the online application stage, and otherwise, ending the process.
The threshold upper limit setting in step S7 includes the steps of:
step one, calculating standard deviation sigma of training set residual sequencetrain；
Step two, for the data point i, setting the upper limit of the estimated residual error threshold of the data point i to be equal to the EWMA estimated value EWMA (i) plus three times of standard deviation sigmatrainI.e. u (i) ═ ewma (i) +3 σtrainWherein U (i) is a threshold upper limit.
S8, performing an online application stage;
s9, real-time estimation: repeating the steps S2-S5, selecting the relevant data set as the real-time operation data set of the wind turbine generator set in the step S2, marking the data set as a test set, obtaining the real-time operation data set of the wind turbine generator set, selecting the oil temperature of the gearbox as a target variable, performing feature extraction on other variables through an Autoencoder model, inputting the feature variables into a trained variable estimation model to obtain a real-time estimation value of the target variable, and subtracting the model estimation value from the actual operation value of the target variable to obtain a real-time residual value Ri；
S10, obtaining a real-time operation residual error: repeating the residual training in step S6;
s11, real-time judgment, and design of judgment criteria according to the step S7: designing a continuous overrun judgment criterion and setting a judgment criterion threshold value C;
the design of the judgment criterion in the step S11 comprises the following steps:
step one, when the estimated residual error value of a data point i is larger than the upper threshold limit of the data point i, namely RiWhen the value is greater than U (i), calculating the continuous number Q (i) of data points of which the estimated residual errors appearing before the data points i are greater than the corresponding threshold upper limit;
step two, when the condition Q (i) of the discrimination criterion is satisfied and is not less than C, alarming is carried out at a data point i;
step three, when the condition Q (i) < C is satisfied, or the estimated residual error of the data point i is not more than the upper threshold limit of the data point i, namely RiWhen the value is less than or equal to U (i), no alarm is given at a data point i.
And S12, obtaining a final early warning result.
Example 2: the wind turbine generator system fault early warning method is characterized in that fault early warning is carried out on a certain wind turbine generator system which has a fault of over-temperature of the oil temperature of the gear box, the wind turbine generator system is not connected with the gear box system at 2019.01.30, the diagnosis result is that the oil temperature of the gear box is too high, the time range of data collected by an SCADA system of the wind turbine generator system is 2018.08.0800: 00: 00-2019.01.3014: 00:00, and the data sampling interval is 1 min.
The implementation data set of the wind turbine generator gearbox oil temperature over-temperature fault early warning method in the embodiment is wind turbine generator SCADA operation data, and the implementation steps of the method are as follows:
1) performing data set collection, selection and labeling: selecting data of the previous 4 months of the data set, namely 2018.08.0800: 00: 00-2018.11.3023: 59:00, as a training set for model training, selecting the rest data, namely 2018.12.0100: 00: 00-2019.01.3014: 00:00, as a test set, as a real-time operation data set, and selecting the oil temperature of the gearbox as a target variable;
2) and (3) carrying out feature extraction: in the embodiment, the data set is free of time information and target variables, and comprises 39 other variables, 39 variable information is used as a coding Autoencoder model to be input into a hidden layer, namely, the dimension of a feature vector is set to be 10, and the 39 dimensional variable information is extracted to form a 10 dimensional feature vector;
3) reconstructing a training set, and selecting a variable estimation model: constructing a training set by using the target variable and the characteristic variable after dimensionality reduction, selecting a Support Vector Regression (SVR) as a variable estimation model, selecting a Radial Basis Function (RBF) by the Support Vector Regression (SVR), selecting SVR parameters by using a cross validation grid search method, and determining a parameter model with the minimum error.
4) And training the selected variable estimation model, then performing residual error training, and obtaining a residual error sequence. And estimating the oil temperature of the gearbox in real time, and representing the normal running state of the oil temperature of the gearbox.
And subtracting the training model output of the variable estimation model from the actual operation value of the target variable, namely the target variable estimation value, to obtain a target variable estimation residual error.
The target variable estimation result is shown in fig. 2, a blue line is a target variable actual operation value of the training set plus the test set, a yellow line is a target variable model estimation value of the training set, a green line is a target variable model estimation value of the test set, a red line is an estimation residual value obtained by subtracting the model estimation value from the target variable actual operation value of the training set plus the test set, and the fitting index is selected as root mean square error RMSE, wherein the training set RMSE is 1.8158, and the test set RMSE is 8.2407. Since colors cannot be represented in the drawings, the colors are replaced with labels.
The residual estimate is calculated using an exponentially weighted moving average EWMA estimate calculation formula.
The calculation formula of the exponentially weighted moving average EWMA estimation value is as follows:
EWMA(i)＝βEWMA(i-1)+(1-β)T(i)；
β represents the weighted decreasing rate, EWMA (i-1) is the EWMA estimated value of the previous data point of the data point i, the initial value EWMA (0) ═ 0, and T (i) is the actual residual value of the data point i.
Estimating residuals based on target variables, calculating residual estimates using an exponentially weighted moving average EWMA, which for data point i has an exponentially weighted moving average EWMA estimate EWMA (i) ═ β EWMA (i-1) + (1- β) t (i), in this example β ═ 0.9, EWMA (0) ═ 0;
5) setting a threshold upper limit: counting indexes, setting a threshold upper limit of a residual sequence according to an EWMA model, and calculating a standard deviation sigma of the residual sequence of a training settrainFor data point i, the upper threshold is set equal to the EWMA estimate EWMA (i) plus three times the standard deviation σtrainI.e. u (i) ═ ewma (i) +3 σtrain，
Because the oil temperature in the over-temperature fault of the oil temperature of the gearbox abnormally rises, only the upper threshold limit is considered, and the lower threshold limit is not considered, as shown in fig. 3, a schematic diagram of the target variable estimation residual error in this embodiment is shown, where a green line is a training set residual error, a yellow line is a test set residual error, a blue line is an EWMA estimation value, and a black line is an upper threshold limit, and since colors in the drawing cannot be expressed, the color is replaced with a label.
6) Performing feature extraction on other variables of a real-time operation data point, namely a data point j in a test set, through an Autoencoder model, inputting a feature vector into a trained variable estimation model to obtain a real-time estimation value of a target variable, and subtracting the model estimation value from an actual operation value of the target variable to obtain a real-time residual value Rj；
7) Designing continuous out-of-limit criterion and setting criterion threshold value C, in this embodiment C is 30, for data point j, when the data point residual value is greater than the upper threshold value, that is, RjU (j), calculating that continuity occurs before data point jThe number q (j) by which the data point residual value is greater than the upper threshold limit.
When the condition Q (i) ≧ C is met, alarming is carried out at the data point j, and when the condition Q (i) < C is met or the estimated residual error of the data point j is not more than the upper threshold limit of the data point, namely RjAnd when the value is less than or equal to U (j), no alarm is given at the data point j, and the operation is carried out on all the data points in the test set.
Fig. 4 is a diagram of an early warning result in the embodiment of the present invention, where a red point is a data point corresponding to an alarm time, the earliest alarm time is 2019.01.0106: 25:00, and early warning is implemented 29 days before a fault occurs, and since colors in the diagram cannot be represented, the color is replaced with a label.
The result shows that the alarm device can realize accurate alarm before the over-temperature fault of the oil temperature of the gearbox occurs, and the result has validity and reliability.
The invention discloses a wind turbine generator gearbox oil temperature over-temperature fault early warning method which mainly comprises the steps of carrying out feature extraction based on an Autoencoder, training a variable estimation model, obtaining real-time operation residual errors, designing a criterion and the like.
The method comprises the steps of selecting a data set of a unit in a normal operation state period based on real-time operation data recorded by a data acquisition and monitoring control (SCADA) system installed in the wind turbine generator, selecting the oil temperature of a gearbox as a target variable, performing feature extraction by using an Autoencoder model for other variables for dimensionality reduction, then selecting an SVR as a variable estimation model for training, realizing real-time estimation of the target variable, inputting the real-time operation data set of the wind turbine generator into the trained SVR model in online application, obtaining a residual sequence of an actual operation value and a model estimation value of the target variable, considering variable volatility based on the residual sequence, setting an over-limit threshold value of the residual sequence by using an EWMA (equal weighted average) and designing a continuous over-limit early warning judgment criterion to realize the early sensing of the over-temperature fault of the oil temperature of the gearbox.
In the fault early warning method, the Autoencoder model is used for reducing the dimension of the variable, the characteristic with strong correlation with the target variable is extracted, and the training efficiency of the variable estimation model is improved; an overrun threshold is set by using the EWMA, so that data fluctuation is smoothed, and the influence of an isolated abnormal value is avoided; and a continuous overrun judgment criterion is designed, so that false alarm caused by data fluctuation is reduced, and the accuracy of an early warning result is ensured.
The method can overcome the technical problems that in the prior art, a vibration sensor needs to be additionally arranged for monitoring the fault of the gearbox, the construction is troublesome, the cost is too high, and the temperature of the gear oil is easily influenced by the surrounding environment and noise, can realize the early sensing and early warning of the over-temperature fault of the oil temperature of the gearbox of the wind turbine generator system, avoids the fault shutdown caused by the abnormal evolution of the components of the gearbox, reduces the shutdown time and the maintenance loss, reduces the influence of the surrounding environment and the noise on the oil temperature data, ensures the accuracy of the early warning result, has low cost, convenience, practical value and stronger theoretical performance and practicability.
Claims (8)
1. A wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data is characterized by comprising an off-line training stage and an on-line application stage, and the method comprises the following steps:
s1, performing an off-line training stage;
s2, data set collection, selection and marking are carried out: acquiring and recording data in a wind turbine SCADA system needing to perform over-temperature fault early warning of the oil temperature of the gearbox, selecting and marking related data sets, and selecting the oil temperature of the gearbox as a target variable;
s3, feature extraction: screening out features with strong correlation with target variables, selecting a self-coding Autoencoder model for dimension reduction and feature extraction based on all other variables of an SCADA system data set, and converting high-dimensional SCADA information into low-dimensional feature vectors;
s4, reconstructing a training set, and selecting a variable estimation model: constructing a training set by using the target variable and the characteristic variable after dimensionality reduction, and selecting a Support Vector Regression (SVR) as a variable estimation model;
s5, training the selected variable estimation model;
s6, carrying out residual error training and obtaining a residual error sequence;
s7, setting a threshold upper limit: counting indexes, setting a threshold upper limit of a residual sequence according to an EWMA model, judging whether the wind turbine needs to enter an online application stage, entering a step S8 when the wind turbine needs to enter the online application stage, and otherwise, ending the process;
s8, performing an online application stage;
s9, real-time estimation: repeating the step S2 to the step S5, inputting the characteristic variables into the trained variable estimation model, and obtaining the real-time estimation value of the target variable;
s10, obtaining a real-time operation residual error: repeating the residual training in step S6;
s11, real-time judgment, and design of judgment criteria according to the step S7: designing a continuous overrun judgment criterion and setting a judgment criterion threshold value C;
and S12, obtaining a final early warning result.
2. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 1, wherein the related data set in the step S2 is selected and marked as a data set in which the wind turbine generator is in a normal operation period, and is marked as a training set, and the related data set in the step S9 is selected as a real-time operation data set of the wind turbine generator in the step S2, and is marked as a test set.
3. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as recited in claim 1, wherein in the step S4, a radial basis kernel function (RBF) is selected for a Support Vector Regression (SVR), a cross validation grid search method is used for SVR parameter selection, and a parameter model with the minimum error is determined.
4. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 1, wherein the residual error training in the step S6 comprises the following steps:
the method comprises the following steps of firstly, estimating the oil temperature of the gearbox in real time, and representing the normal operation state of the oil temperature of the gearbox;
subtracting the training model output of the variable estimation model from the actual operation value of the target variable, namely the target variable estimation value, to obtain a target variable estimation residual error;
and thirdly, calculating a residual error estimated value by using an exponential weighted moving average EWMA estimated value calculation formula.
5. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 4, wherein the calculation formula of the exponential weighted moving average EWMA estimated value in the third step is as follows:
EWMA(i)＝βEWMA(i-1)+(1-β)T(i)；
β represents the weighted decreasing rate, EWMA (i-1) is the EWMA estimated value of the previous data point of the data point i, the initial value EWMA (0) ═ 0, and T (i) is the actual residual value of the data point i.
6. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 1, wherein the setting of the threshold upper limit in the step S7 includes the following steps:
step one, calculating standard deviation sigma of training set residual sequencetrain；
Step two, for the data point i, setting the upper limit of the estimated residual error threshold of the data point i to be equal to the EWMA estimated value EWMA (i) plus three times of standard deviation sigmatrainI.e. u (i) ═ ewma (i) +3 σtrainWherein U (i) is a threshold upper limit.
7. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 1, wherein the real-time residual value R is obtained by subtracting the model estimation value from the actual operation value of the target variable in the step S10i。
8. The wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data as claimed in claim 1, wherein the criterion design in the step S11 includes the following steps:
step one, when the estimated residual error value of a data point i is larger than the upper threshold limit of the data point i, namely RiWhen the value is greater than U (i), calculating the continuous number Q (i) of data points of which the estimated residual errors appearing before the data points i are greater than the corresponding threshold upper limit;
step two, when the condition Q (i) of the discrimination criterion is satisfied and is not less than C, alarming is carried out at a data point i;
step three, when the condition Q (i) < C is satisfied, or the estimated residual error of the data point i is not more than the upper threshold limit of the data point i, namely RiWhen the value is less than or equal to U (i), no alarm is given at a data point i.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010139476.9A CN111415070A (en) | 2020-03-03 | 2020-03-03 | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010139476.9A CN111415070A (en) | 2020-03-03 | 2020-03-03 | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111415070A true CN111415070A (en) | 2020-07-14 |
Family
ID=71494103
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010139476.9A Pending CN111415070A (en) | 2020-03-03 | 2020-03-03 | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111415070A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111950505A (en) * | 2020-08-24 | 2020-11-17 | 湖南科技大学 | State evaluation method for wind driven generator sensor of SSA-AANN |
| CN112232405A (en) * | 2020-10-13 | 2021-01-15 | 中车青岛四方机车车辆股份有限公司 | Fault prediction, monitoring and diagnosis method of gearbox and corresponding device |
| CN112834211A (en) * | 2020-12-31 | 2021-05-25 | 江苏国科智能电气有限公司 | Fault early warning method for transmission system of wind turbine generator |
| CN113408199A (en) * | 2021-06-16 | 2021-09-17 | 华电山东新能源有限公司 | Gearbox oil temperature fault early warning method based on multilayer perception neural network |
| CN113570146A (en) * | 2021-07-30 | 2021-10-29 | 国电电力内蒙古新能源开发有限公司 | LightGBM-based wind turbine generator gearbox oil temperature fault early warning method |
| CN114295367A (en) * | 2021-11-12 | 2022-04-08 | 华能新能源股份有限公司 | Wind turbine generator gearbox working condition online monitoring method |
| CN116883396A (en) * | 2023-09-06 | 2023-10-13 | 天津医科大学 | Resting state magnetic resonance image analysis method and system based on artificial intelligence |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101263499A (en) * | 2005-07-11 | 2008-09-10 | 布鲁克斯自动化公司 | Intelligent condition monitoring and fault diagnostic system |
| CN110378042A (en) * | 2019-07-23 | 2019-10-25 | 山东大学 | Gearbox of wind turbine oil temperature method for detecting abnormality and system based on SCADA data |
-
2020
- 2020-03-03 CN CN202010139476.9A patent/CN111415070A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101263499A (en) * | 2005-07-11 | 2008-09-10 | 布鲁克斯自动化公司 | Intelligent condition monitoring and fault diagnostic system |
| CN110378042A (en) * | 2019-07-23 | 2019-10-25 | 山东大学 | Gearbox of wind turbine oil temperature method for detecting abnormality and system based on SCADA data |
Non-Patent Citations (3)
| Title |
|---|
| 王洪斌;王红;何群;王跃灵;周振;: "基于深度信念网络的风机主轴承状态监测方法" * |
| 王炜超;袁逸萍;孙文磊;赵琴;樊盼盼;贾依达尔・热孜别克;: "融合SCADA数据的风电机组齿轮箱状态评估" * |
| 黄小光;潘东浩;史晓鸣;王杏;王欣;: "风电机组齿轮箱系统的故障预警研究" * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111950505A (en) * | 2020-08-24 | 2020-11-17 | 湖南科技大学 | State evaluation method for wind driven generator sensor of SSA-AANN |
| CN111950505B (en) * | 2020-08-24 | 2023-08-29 | 湖南科技大学 | SSA-AANN wind driven generator sensor state evaluation method |
| CN112232405A (en) * | 2020-10-13 | 2021-01-15 | 中车青岛四方机车车辆股份有限公司 | Fault prediction, monitoring and diagnosis method of gearbox and corresponding device |
| CN112232405B (en) * | 2020-10-13 | 2022-09-02 | 中车青岛四方机车车辆股份有限公司 | Fault prediction, monitoring and diagnosis method of gearbox and corresponding device |
| CN112834211A (en) * | 2020-12-31 | 2021-05-25 | 江苏国科智能电气有限公司 | Fault early warning method for transmission system of wind turbine generator |
| CN113408199A (en) * | 2021-06-16 | 2021-09-17 | 华电山东新能源有限公司 | Gearbox oil temperature fault early warning method based on multilayer perception neural network |
| CN113570146A (en) * | 2021-07-30 | 2021-10-29 | 国电电力内蒙古新能源开发有限公司 | LightGBM-based wind turbine generator gearbox oil temperature fault early warning method |
| CN114295367A (en) * | 2021-11-12 | 2022-04-08 | 华能新能源股份有限公司 | Wind turbine generator gearbox working condition online monitoring method |
| CN116883396A (en) * | 2023-09-06 | 2023-10-13 | 天津医科大学 | Resting state magnetic resonance image analysis method and system based on artificial intelligence |
| CN116883396B (en) * | 2023-09-06 | 2023-11-28 | 天津医科大学 | Resting state magnetic resonance image analysis method and system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111415070A (en) | Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data | |
| CN106682814B (en) | Wind turbine generator fault intelligent diagnosis method based on fault knowledge base | |
| CN111597682B (en) | Method for predicting remaining life of bearing of gearbox of wind turbine | |
| CN111539553B (en) | Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree | |
| CN110362045B (en) | Marine doubly-fed wind turbine generator fault discrimination method considering marine meteorological factors | |
| Butler et al. | A feasibility study into prognostics for the main bearing of a wind turbine | |
| CN111814849B (en) | DA-RNN-based wind turbine generator set key component fault early warning method | |
| CN110969185A (en) | Equipment abnormal state detection method based on data reconstruction | |
| CN111103137A (en) | Wind turbine gearbox fault diagnosis method based on deep neural network | |
| CN111814848B (en) | Self-adaptive early warning strategy design method for temperature faults of wind turbine generator | |
| CN109356798B (en) | Wind driven generator gearbox state monitoring method based on co-integration analysis | |
| CN114841580A (en) | Generator fault detection method based on hybrid attention mechanism | |
| CN111878322A (en) | Wind power generator device | |
| CN111396266A (en) | GBRT-based wind turbine generator bearing fault early warning method | |
| CN115578084A (en) | Wind turbine generator set frequency converter fault early warning method based on deep convolution self-encoder | |
| CN117365869A (en) | Self-adaptive early warning strategy design method for wind turbine blade tower sweeping faults | |
| CN116609055A (en) | Method for diagnosing wind power gear box fault by using graph convolution neural network | |
| CN113468728B (en) | Variable pitch system fault prediction method based on neural network | |
| Pandit et al. | Using Gaussian process theory for wind turbine power curve analysis with emphasis on the confidence intervals | |
| CN117151684A (en) | Wind power fan data analysis early warning method, system, device and readable storage medium | |
| Souza et al. | Evaluation of data based normal behavior models for fault detection in wind turbines | |
| Song et al. | Framework of designing an adaptive and multi-regime prognostics and health management for wind turbine reliability and efficiency improvement | |
| Zhang | Comparison of data-driven and model-based methodologies of wind turbine fault detection with SCADA data | |
| CN114841061A (en) | Wind power gear box health assessment method and system integrating timing sequence information | |
| CN114607571A (en) | Offshore wind power gear box fault identification method and system based on lubricating system monitoring |
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 |