CN111103136A - Fan gearbox fault detection method based on SCADA data analysis - Google Patents

Fan gearbox fault detection method based on SCADA data analysis Download PDF

Info

Publication number
CN111103136A
CN111103136A CN201911278929.XA CN201911278929A CN111103136A CN 111103136 A CN111103136 A CN 111103136A CN 201911278929 A CN201911278929 A CN 201911278929A CN 111103136 A CN111103136 A CN 111103136A
Authority
CN
China
Prior art keywords
data
gearbox
fan
prediction model
error
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
Application number
CN201911278929.XA
Other languages
Chinese (zh)
Inventor
宋哲
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Songshu Technology Co Ltd
Original Assignee
Nanjing Songshu Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing Songshu Technology Co Ltd filed Critical Nanjing Songshu Technology Co Ltd
Priority to CN201911278929.XA priority Critical patent/CN111103136A/en
Publication of CN111103136A publication Critical patent/CN111103136A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M13/00—Testing of machine parts
    • G01M13/02—Gearings; Transmission mechanisms
    • G01M13/021—Gearings

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
  • Wind Motors (AREA)

Abstract

The invention discloses a fan gear box fault detection method based on SCADA data analysis in the technical field of industrial energy-saving industry, which specifically comprises the following steps: s1: preprocessing SDACA data of the wind turbine generator by using a simple and effective algorithm, eliminating invalid data and combining all effective data to generate a training data set; s2: developing a prediction model based on a training data set by using a data mining method; s3: calculating the fitting error of the fan unit through a prediction model; s4: and compiling a statistical process control chart according to the fitting error of each fan, giving upper and lower control limit values, and activating the fault alarm of the gearbox if the installation error of the fitting error of the fan exceeds the control limit values. The performance of the ordinary gearbox is predicted by training a deep neural network through data of the ordinary gearbox, the established DNN model is verified through data of normal and abnormal gearboxes, abnormal behaviors of the gearbox are detected through fitting errors and statistical process control diagrams, and the fault condition of the fan gearbox is found in time.

Description

Fan gearbox fault detection method based on SCADA data analysis
Technical Field
The invention relates to the technical field of industrial energy-saving industry, in particular to a fan gear box fault detection method based on SCADA data analysis.
Background
Due to the aging of wind farms, the Operating and Maintenance (OM) costs become very important. The main subsystems of a wind turbine generator set, such as the gearbox, generator, bearings, etc., are the most interesting parts of condition monitoring and fault detection research. If a fault can be detected in advance so that the operator has sufficient time to adjust the power generation schedule and prepare for equipment replacement, the operating costs will be greatly reduced. Since the gearbox represents a significant portion of the total cost and failure thereof can result in excessive downtime, the present invention requires the establishment of an effective wind gearbox monitoring model to reduce excessive downtime and thus reduce the total cost.
The traditional gearbox monitoring method is to analyze the vibration signal in the frequency domain. Mohanty et al (2006) detect a multi-stage transmission using a discrete wavelet transform of the current signal. Roman et al (2014) extract gear damage features by using spectral analysis and acceleration envelope technology, and accurately detect specific damage features by using a synchronous analysis method. In recent years, data mining algorithms have also been introduced in the detection of abnormal performance of gearboxes, and rafie et al (2007) model different gear states using neural networks based on the standard deviation of the wavelet packet coefficients. Zhang et al (2012) combines a data mining algorithm with a control map to monitor vibration excitation of the transmission.
Since previous research usually requires the installation of additional sensors, it is difficult to acquire vibration data for practical industrial applications. However, SCADA systems connecting wind turbines and meteorological stations are currently being built in most modern wind farms, and compared to vibration data, SCADA data is relatively cheap and non-intrusive in the time domain. In addition, the monitoring model can be applied to different operation conditions. Von et al (2013) derive a robust relationship between temperature and power output and use SDACA oil temperature to predict gearbox faults. Gacia et al (2006) simulated the normal behavior of the gearbox at the bearing oil temperature using a neural network algorithm and applied the model to detect initial anomalies in the gearbox. The King et al (2013) adopt a nonlinear state estimation technology to establish an oil temperature model, and consider Welch t test for fault detection.
According to the invention, the gear box is preliminarily researched by utilizing SCADA data, and the oil temperature of the gear box is taken as a monitoring target. However, the oil temperature of the gearbox is easily influenced by the environment, and the noise of measured data is high, so that an optional monitoring target, namely the lubricating oil pressure of the gearbox, which is less influenced by the environment, is considered, a prediction model of the lubricating oil pressure of the gearbox is established by adopting a deep neural network algorithm, and a statistical control chart is introduced to detect abnormal behaviors existing in the wind power gearbox, so that the fan gearbox fault monitoring method based on SCADA system data analysis is established.
Disclosure of Invention
The invention aims to provide a fan gearbox fault detection method based on SCADA data analysis, which aims to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme: the fan gear box fault detection method based on SCADA data analysis specifically comprises the following steps:
s1: preprocessing SDACA data of the wind turbine generator by using a simple and effective algorithm, eliminating invalid data and combining all effective data to generate a training data set;
s2: developing a prediction model based on a training data set by using a data mining method;
s3: calculating the fitting error of the fan unit through a prediction model;
s4: and compiling a statistical process control chart according to the fitting error of each fan, giving upper and lower control limit values, and activating the fault alarm of the gearbox if the installation error of the fitting error of the fan exceeds the control limit values.
3. Preferably, the prediction model takes the lubricating oil pressure as a monitoring target, and establishes a gearbox prediction model, and the method comprises the following two stages:
the first stage is as follows: establishing a lubricating oil pressure prediction model by utilizing a DNN algorithm;
and a second stage: and (4) constructing an EMWA control chart, and alarming the gearbox fault by using the UCL and the LCL.
Preferably, in the first stage, the prediction model is trained based on SCADA data of all normal wind turbines representing normal operation of the wind turbines, and fitting errors of normal and abnormal wind turbines are estimated;
the fitting error is e, and the calculation method of e is
Figure BDA0002316151480000031
Wherein
Figure BDA0002316151480000032
In order to predict the oil pressure,
Figure BDA0002316151480000033
is the actual oil pressure.
Preferably, in the first stage, a DNN with three hidden layers is further included To simulate the mapping from the inputs To, Po and Ts To the output Pl, and the DNN training process estimates parameters Wn and bn by minimizing the mean square fitting error, which are the weights and deviations of n layers, respectively, as shown below:
the oil pressure calculation formula is:
Figure BDA0002316151480000034
the parameter calculation formula is as follows:
Figure BDA0002316151480000035
l represents the number of hidden layers.
Preferably, the second stage comprises the following specific steps:
static test value z of EMWAtThe calculation formula is as follows: z is a radical oft＝λet+(1-λ)zt-1；
Wherein etIs a reconstruction error at time t, and λ is 0<λ<1 is a constant;
from the above formula, z istThe mean and variance of (a) are the mean and standard deviation of the standard error e:
Figure BDA0002316151480000036
the control limit of the EWMA control map is based on a sigma limit of L, where L is typically equal to 3, and the upper and lower control limits of the EWMA are dependent on time t, which is given by the formula:
Figure BDA0002316151480000041
Figure BDA0002316151480000042
compared with the prior art, the invention has the beneficial effects that:
(1) the method has low cost, does not need to install an additional sensor, is directly used in the existing SDACA system, and can accurately obtain SCADA data of industrial application;
(2) non-intrusive, since SCADA systems connecting wind turbines and meteorological stations have been built in most wind farms, SCADA data is non-intrusive in the time domain compared to vibration data used in traditional monitoring;
(3) the method has strong pertinence, and aims at different operating conditions, the semi-supervised learning technology is utilized to establish the online monitoring of the prediction model, the fault alarm of the gearbox is activated once the abnormal turbine generates errors exceeding the monitoring limit value, corresponding measures are taken, the downtime of the gearbox caused by faults is reduced, and therefore the cost is reduced.
Detailed Description
The technical solutions in the embodiments of the present invention are clearly and completely described below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Fan gear box fault detection method based on SCADA data analysis, its characterized in that: the method specifically comprises the following steps:
s1: preprocessing SDACA data of the wind turbine generator by using a simple and effective algorithm, eliminating invalid data and combining all effective data to generate a training data set;
s2: developing a prediction model based on a training data set by using a data mining method;
s3: calculating the fitting error of the fan unit through a prediction model;
s4: and compiling a statistical process control chart according to the fitting error of each fan, giving upper and lower control limit values, and activating the fault alarm of the gearbox if the installation error of the fitting error of the fan exceeds the control limit values.
The prediction model takes the lubricating oil pressure as a monitoring target, and establishes a gearbox prediction model, and comprises the following two stages:
the first stage is as follows: establishing a lubricating oil pressure prediction model by utilizing a DNN algorithm;
the prediction model is trained on the basis of SCADA data of all normal wind turbines representing normal operation of the wind turbines, and fitting errors of normal and abnormal fans are estimated;
the fitting error is e, and the calculation method of e is
Figure BDA0002316151480000051
Wherein
Figure BDA0002316151480000052
In order to predict the oil pressure,
Figure BDA0002316151480000053
is the actual oil pressure.
A DNN with three hidden layers is also included To model the mapping from the inputs To, Po and Ts To the output Pl, and the training process of DNN is To estimate the parameters Wn and bn by minimizing the mean square fitting error, which are the weights and the deviations of the n layers, respectively, as shown below:
the oil pressure calculation formula is:
Figure BDA0002316151480000054
the parameter calculation formula is as follows:
Figure BDA0002316151480000055
l represents the number of hidden layers.
And a second stage: and (4) constructing an EMWA control chart, and alarming the gearbox fault by using the UCL and the LCL.
6. The method comprises the following specific steps:
static test value z of EMWAtThe calculation formula is as follows: z is a radical oft＝λet+(1-λ)zt-1；
Wherein etIs a reconstruction error at time t, and λ is 0<λ<1 is a constant;
from the above formula, z istThe mean and variance of (a) are the mean and standard deviation of the standard error e:
Figure BDA0002316151480000061
the control limit of the EWMA control map is based on a sigma limit of L, where L is typically equal to 3, and the upper and lower control limits of the EWMA are dependent on time t, which is given by the formula:
Figure BDA0002316151480000062
Figure BDA0002316151480000063
the invention provides a case: and analyzing gearbox fault cases of the Liaoning wind power plant and the Hebei wind power plant.
In Liaoning wind farm, 9 wind driven generators are considered, one fan has a gearbox fault record, and the rest are common fans. SCADA data of 4 fans are collected in a wind power plant in the north of the river, wherein 1 fan is an abnormal fan, 3 fans are normal fans, all wind power generation sets are provided with SCADA systems, and the sampling frequency of the SCADA systems is 10 minutes. SCADA data of Liaoning and Hebei wind power plants are respectively collected from 1 day 4-17 days 5-2015 and from 1 day 4-6 months 2 days 2015. Details of the two wind farms are given in the table below.
Figure BDA0002316151480000064
In order to ensure that the prediction model can reflect the characteristics of a normal wind turbine generator, four parameters are selected from original SCADA data, data are preprocessed, a DNN prediction model is trained, and fitting errors of normal and abnormal fans are calculated. These fit errors may be used to measure the difference between the current behavior of the wind turbine and the expected normal behavior. In the conventional turbine EWMA diagram, the fitting error is within the boundary range of the UCL and the LCL. In addition, a failure may alert two or three days before the actual failure occurs, providing wind farm personnel with sufficient time to check the gearbox condition and plan replacement. If an abnormal gearbox could be repaired or replaced before failure occurred, additional costs would be avoided. Therefore, it provides enough time for wind farm operators to check the gearbox condition and plan maintenance, and furthermore, the whole monitoring process can pass through the parallel superoxide dismutase algorithm within minutes, the results show that the proposed monitoring model is feasible for industrial application.
In the description herein, references to the description of "one embodiment," "an example," "a specific example" or the like are intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims (5)

1. Fan gear box fault detection method based on SCADA data analysis, its characterized in that: the method specifically comprises the following steps:
s1: preprocessing SDACA data of the wind turbine generator by using a simple and effective algorithm, eliminating invalid data and combining all effective data to generate a training data set;
s2: developing a prediction model based on a training data set by using a data mining method;
s3: calculating the fitting error of the fan unit through a prediction model;
s4: and compiling a statistical process control chart according to the fitting error of each fan, giving upper and lower control limit values, and activating the fault alarm of the gearbox if the installation error of the fitting error of the fan exceeds the control limit values.
2. The wind turbine gearbox fault detection method based on SCADA data analysis of claim 1, wherein: the prediction model takes the lubricating oil pressure as a monitoring target, and establishes a gearbox prediction model, and comprises the following two stages:
the first stage is as follows: establishing a lubricating oil pressure prediction model by utilizing a DNN algorithm;
and a second stage: and (4) constructing an EMWA control chart, and alarming the gearbox fault by using the UCL and the LCL.
3. The wind turbine gearbox fault detection method based on SCADA data analysis of claim 1, wherein: in the first stage, the prediction model is trained on the basis of SCADA data of all normal wind turbines representing normal operation of the wind turbines, and fitting errors of normal and abnormal fans are estimated;
the fitting error is e, and the calculation method of e is
Figure FDA0002316151470000011
Wherein
Figure FDA0002316151470000012
In order to predict the oil pressure,
Figure FDA0002316151470000013
is the actual oil pressure.
4. The wind turbine gearbox fault detection method based on SCADA data analysis of claim 1, wherein: in the first stage, a DNN with three hidden layers is also included To model the mapping from the inputs To, Po and Ts To the output Pl, and the DNN training process estimates the parameters Wn and bn by minimizing the mean square fitting error, which are the weights and the deviations of the n layers, respectively, as shown below:
the oil pressure calculation formula is:
Figure FDA0002316151470000021
the parameter calculation formula is as follows:
Figure FDA0002316151470000022
l represents the number of hidden layers.
5. The wind turbine gearbox fault detection method based on SCADA data analysis of claim 1, wherein: the second stage comprises the following specific steps:
static test value z of EMWAtThe calculation formula is as follows: z is a radical oft＝λet+(1-λ)zt-1；
Wherein etIs a reconstruction error at time t, and λ is 0<λ<1 is a constant;
from the above formula, z istThe mean and variance of (a) are the mean and standard deviation of the standard error e:
Figure FDA0002316151470000023
the control limit of the EWMA control map is based on a sigma limit of L, where L is typically equal to 3, and the upper and lower control limits of the EWMA are dependent on time t, which is given by the formula:
Figure FDA0002316151470000024
Figure FDA0002316151470000025
CN201911278929.XA 2019-12-13 2019-12-13 Fan gearbox fault detection method based on SCADA data analysis Pending CN111103136A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911278929.XA CN111103136A (en) 2019-12-13 2019-12-13 Fan gearbox fault detection method based on SCADA data analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911278929.XA CN111103136A (en) 2019-12-13 2019-12-13 Fan gearbox fault detection method based on SCADA data analysis

Publications (1)

Publication Number Publication Date
CN111103136A true CN111103136A (en) 2020-05-05

Family

ID=70422367

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911278929.XA Pending CN111103136A (en) 2019-12-13 2019-12-13 Fan gearbox fault detection method based on SCADA data analysis

Country Status (1)

Country Link
CN (1) CN111103136A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN115264054A (en) * 2022-09-28 2022-11-01 北谷电子有限公司 Method and system for monitoring whether gearbox is abnormal or not
CN116910570A (en) * 2023-09-13 2023-10-20 华能新能源股份有限公司山西分公司 Wind turbine generator system fault monitoring and early warning method and system based on big data

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103268366A (en) * 2013-03-06 2013-08-28 辽宁省电力有限公司电力科学研究院 Combined wind power prediction method suitable for distributed wind power plant
CN107153929A (en) * 2017-07-10 2017-09-12 龙源（北京）风电工程技术有限公司 Gearbox of wind turbine fault monitoring method and system based on deep neural network
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

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103268366A (en) * 2013-03-06 2013-08-28 辽宁省电力有限公司电力科学研究院 Combined wind power prediction method suitable for distributed wind power plant
CN107153929A (en) * 2017-07-10 2017-09-12 龙源（北京）风电工程技术有限公司 Gearbox of wind turbine fault monitoring method and system based on deep neural network
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

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN115264054A (en) * 2022-09-28 2022-11-01 北谷电子有限公司 Method and system for monitoring whether gearbox is abnormal or not
CN115264054B (en) * 2022-09-28 2022-12-13 北谷电子有限公司 Method and system for monitoring whether gearbox is abnormal or not
CN116910570A (en) * 2023-09-13 2023-10-20 华能新能源股份有限公司山西分公司 Wind turbine generator system fault monitoring and early warning method and system based on big data
CN116910570B (en) * 2023-09-13 2023-12-15 华能新能源股份有限公司山西分公司 Wind turbine generator system fault monitoring and early warning method and system based on big data

Similar Documents

Publication Publication Date Title
EP3638900B1 (en) Independent monitoring system for a wind turbine
Dao et al. Condition monitoring and fault detection in wind turbines based on cointegration analysis of SCADA data
Uluyol et al. Power curve analytic for wind turbine performance monitoring and prognostics
Kim et al. Use of SCADA data for failure detection in wind turbines
CN111103136A (en) Fan gearbox fault detection method based on SCADA data analysis
Schlechtingen et al. Wind turbine condition monitoring based on SCADA data using normal behavior models. Part 2: Application examples
Yang et al. Wind turbine condition monitoring by the approach of SCADA data analysis
Nejad et al. A prognostic method for fault detection in wind turbine drivetrains
Zhang et al. SCADA-data-based wind turbine fault detection: A dynamic model sensor method
Cambron et al. Control chart monitoring of wind turbine generators using the statistical inertia of a wind farm average
Gonzalez et al. On the use of high-frequency SCADA data for improved wind turbine performance monitoring
KR20140127915A (en) Method and system for real time gas turbine performance advisory
Cui et al. An anomaly detection approach using wavelet transform and artificial neural networks for condition monitoring of wind turbines' gearboxes
CN114372504A (en) Wind turbine generator fault early warning method based on graph neural network
Du et al. A SCADA data based anomaly detection method for wind turbines
CN111103137A (en) Wind turbine gearbox fault diagnosis method based on deep neural network
Sharma et al. Condition monitoring of wind turbines: a review
Dao Condition monitoring of wind turbines based on cointegration analysis of gearbox and generator temperature data
Wiggelinkhuizen et al. CONMOW: Condition monitoring for offshore wind farms
Li et al. A condition monitoring approach of multi-turbine based on VAR model at farm level
Tchakoua et al. New trends and future challenges for wind turbines condition monitoring
Koukoura et al. Wind turbine intelligent gear fault identification
Huang et al. Remaining useful life estimation of hydrokinetic turbine blades using power signal
Sharma et al. Condition monitoring of wind turbine gear box
CN115929569A (en) Fault diagnosis method for variable pitch system of wind turbine generator

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
RJ01 Rejection of invention patent application after publication
RJ01 Rejection of invention patent application after publication

Application publication date: 20200505