CN113989555A - Application method for early warning of fan yaw fault based on random forest algorithm - Google Patents

Application method for early warning of fan yaw fault based on random forest algorithm Download PDF

Info

Publication number
CN113989555A
CN113989555A CN202111250893.1A CN202111250893A CN113989555A CN 113989555 A CN113989555 A CN 113989555A CN 202111250893 A CN202111250893 A CN 202111250893A CN 113989555 A CN113989555 A CN 113989555A
Authority
CN
China
Prior art keywords
yaw
random forest
data
characteristic
fan
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.)
Withdrawn
Application number
CN202111250893.1A
Other languages
Chinese (zh)
Inventor
赵鹏东
杜保华
黄凤旗
石伟
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.)
Xian TPRI Power Station Information Technology Co Ltd
Original Assignee
Xian TPRI Power Station Information 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 Xian TPRI Power Station Information Technology Co Ltd filed Critical Xian TPRI Power Station Information Technology Co Ltd
Priority to CN202111250893.1A priority Critical patent/CN113989555A/en
Publication of CN113989555A publication Critical patent/CN113989555A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/243—Classification techniques relating to the number of classes
    • G06F18/24323—Tree-organised classifiers
    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
    • F03D—WIND MOTORS
    • F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
    • 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
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02B—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO BUILDINGS, e.g. HOUSING, HOUSE APPLIANCES OR RELATED END-USER APPLICATIONS
    • Y02B10/00—Integration of renewable energy sources in buildings
    • Y02B10/30—Wind power

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Sustainable Development (AREA)
  • Sustainable Energy (AREA)
  • Chemical & Material Sciences (AREA)
  • Combustion & Propulsion (AREA)
  • Mechanical Engineering (AREA)
  • Wind Motors (AREA)

Abstract

The invention discloses an application method for early warning yaw faults of a fan based on a random forest algorithm, which comprises the following steps: selecting a yaw characteristic set of the wind turbine generator; acquiring historical operation data of the feature set, and classifying and sorting the data according to a normal state, a fault state and an alarm state; selecting a Gini coefficient characteristic parameter for the data; constructing a random forest model algorithm according to the CART algorithm; evaluating the constructed algorithm; substituting the new real-time data into the constructed algorithm model to obtain a predicted result; and displaying the prediction result. The method can predict the fault of the wind turbine yaw system in real time, is resistant to overfitting, high in prediction accuracy, concise and easy to implement, reliable in result and free of the influence of human factors. The method has the advantages that the early characteristic excavation of yaw faults is realized, so that the adverse effect of mechanical faults of the yaw system of the wind generating set is reduced, the hidden danger of fan operation is reduced, and the method has important significance for the safe and stable operation of the wind generating set.

Description

Application method for early warning of fan yaw fault based on random forest algorithm
Technical Field
The invention belongs to the technical field of early warning of yaw faults of a wind turbine generator by using a random forest algorithm based on big data, and particularly relates to an application method for early warning of yaw faults of a fan based on the random forest algorithm.
Background
At present, a wind power centralized control base is built on a large scale in China, and wind power generation increasingly permeates into a power system. The running stability of the wind turbine generator can be effectively improved by accurately predicting the running state of the wind turbine generator and early warning the fault. At present, most wind turbines are equipped with a Data Acquisition and monitoring Control System (SCADA), which can acquire and record all-directional operation state information of the wind turbines in real time and provide Data support for analyzing the operation state of the wind turbines by big Data. The wind turbine yaw system has three main functions. Firstly, follow the wind direction and change, mutually support with wind turbine generator system's control system, make wind turbine generator system's impeller be in the windward state all the time, make full use of wind energy improves wind turbine generator system's generating efficiency. And secondly, the protection of cable untwisting and cable twisting is realized, because the wind generating set can continuously yaw towards one direction, in order to ensure that the cable of the suspension part of the set cannot generate excessive twisting to cause cable breakage, the cable can be automatically untwisted when the cable reaches the designed winding value. Meanwhile, when the yaw reaches the design limit value, the unit can be triggered to stop. And thirdly, the unit is positioned, and when the wind direction is relatively stable, the locking torque can be provided according to a set value, so that the engine room is stabilized in one direction, and the safe and stable operation of the unit is guaranteed. In view of the important role of a wind turbine yaw system in the safe operation of a wind turbine, in order to carry out fault early warning on the wind turbine yaw system in real time, the method utilizes historical and real-time SCADA system data as samples, selects characteristic parameters of the yaw system, and constructs a random forest prediction model based on a CART algorithm. Due to the characteristics of overfitting resistance of the random forest algorithm, high prediction accuracy, easiness in implementation and the like, the method is excellent in yaw fault early warning of the fan.
Disclosure of Invention
The invention aims to provide an application method for early warning yaw faults of a fan based on a random forest algorithm.
An application method for early warning of fan yaw faults based on a random forest algorithm comprises the following steps:
1) creating a characteristic data parameter table of a yaw system of the wind generating set, 2) obtaining a characteristic data set of a plurality of marker post fans in a yaw normal state period of the wind power plant, and recording each piece of recorded data as (x)i1,xi2,xi3，...,xin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval data as XzI.e. by
Figure BDA0003322466650000021
Wherein m is the number of the benchmark fans; simultaneously constructing one-dimensional scalar result sets with the same quantity, and recording the result sets as Yz；
3) Acquiring a characteristic data set of a yaw fault state period of a plurality of marker post fans of a wind power plant, and recording each recorded data as (g)i1,gi2,gi3，...,gin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval data as GzI.e. by
Figure BDA0003322466650000022
Wherein m is the number of the benchmark fans; simultaneously constructing one-dimensional scalar result sets with the same quantity, and recording the result sets as Yg；
4) Acquiring a characteristic data set of a yaw warning state period of a plurality of marker post fans in the wind power plant, and recording each recorded data as (e)i1,ei2,ei3，...,ein) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all the time interval data of the fan with normal yaw as EzI.e. by
Figure BDA0003322466650000023
Wherein m is the number of the benchmark fans; simultaneously, a one-dimensional scalar (scalar value is '2') result set with the same quantity is constructed and recorded as Ye；
5) The above feature data set [ X ]z，Gz，EzMerging and constructing a representation mode of a two-dimensional feature matrix, and recording the representation mode as X; namely, it is
Figure BDA0003322466650000031
At the same time, the above result set [ Y ]z，Yz，YzRepresentation of the merged construction into a one-dimensional result matrix, denoted y, i.e.
Figure BDA0003322466650000032
6) The characteristic selection of the damping coefficient is carried out on the constructed characteristic data set (X) and a result set (y) corresponding to the characteristic data set (X), the selection standard of the damping coefficient is that each child node achieves the highest purity, namely all observations falling into the child nodes belong to the same classification, and at the moment, the damping coefficient is the smallest, the purity is the highest, and the uncertainty is the smallest; converting the constructed characteristic data set (X) And result set (y) into a random forest model according to a Classification And Regression Tree (namely, a Classification Regression Tree algorithm, namely a CART algorithm for short); acquiring a new wind turbine generator yaw characteristic parameter set, and constructing new wind turbine generator yaw two-dimensional characteristic matrix data;
9) according to the random forest model constructed in the step 7), transmitting the data into the new wind turbine generator yaw characteristic two-dimensional matrix data constructed in the step 8) to calculate a new result, namely predicting whether yaw faults or yaw alarms need to be generated in the future; and voting the new calculation result by the independent decision tree in the random forest model, namely selecting the final calculation result with the maximum independent decision tree prediction result in the random forest model.
In a further development of the invention, in step 1), the characteristic data parameters are as follows:
TABLE A characteristic data parameter table of fan yaw system
Serial number Parameter name Unit of Serial number Parameter name Unit of
1 Wind speed m/s 14 Yaw brake full release state -
2 Hub temperature ℃ 15 Yaw brake partial release state -
3 Ambient temperature ℃ 16 Operating state of hydraulic pump -
4 Wind direction ° 17 Yaw encoder reset state -
5 Nacelle position ° 18 Running state of yaw lubricating pump -
6 Angle of wind ° 19 Heating operation state of hydraulic oil -
7 Yaw mode - 20 Left untwisted state -
8 Yaw cable twisting angle ° 21 State of right untwisting -
9 Yaw lock - 22 Cable untwisting activation state -
10 Yaw in clockwise mode of operation - 23 Yaw lubricating system alarm -
11 Yaw in counter-clockwise mode of operation - 24 Yaw CW -
12 Yaw hydraulic brake on state - 25 Yaw CCW
13 Yaw motor braking state - 26 Yaw limit switch actuation
The invention has the further improvement that in the step 2), the scalar value of the one-dimensional scalar is 0.
A further improvement of the present invention is that, in step 3), the scalar value of the one-dimensional scalar is 1.
A further improvement of the present invention is that, in step 4), the scalar value of the one-dimensional scalar is 2.
A further improvement of the present invention is that, in step 6), if there is a total of K-class result sets, the probability that the sample belongs to the K-th class is: p is a radical ofkThen the kini index of the probability distribution is:
Figure BDA0003322466650000041
the further improvement of the invention is that in the step 7), the process of constructing the random forest according to the CART algorithm is as follows:
71) selecting a characteristic variable X from a characteristic data set [ X ]iA value of (3), then x is selectediA value v ofi，viDividing the n-dimensional space into two parts, all points of one part satisfying xi＜＝viAll the points of the other part satisfy xi>viFor non-continuous variables, the attribute values have only two values, namely equal to the value or not equal to the value;
72) recursion processing, namely, reselecting an attribute from the two part value domains obtained in the previous step to continue dividing until the whole n-dimensional space is divided;
73) the criteria for the division are determined according to equation (1).
The invention has the further improvement that the method also comprises a step 10) of collecting data for multiple rounds of calculation or carrying out optimized reconstruction on the random forest model in order to avoid prediction errors, namely increasing the number of random forests and adjusting the threshold value of the hyperparameter.
The invention has at least the following beneficial technical effects:
1. high-dimensional (multi-feature) data can be constructed without dimension reduction and feature selection
2. The degree of importance of the feature can be judged
3. The mutual influence among different characteristics can be judged
4. Is not easy to overfit
5. The characteristic set is fast in algorithm training speed and easy to make into a parallel method
6. Is relatively simple to realize
7. For unbalanced data sets, errors can be balanced.
8. Accuracy can still be maintained if a significant portion of the features are missing.
In conclusion, the application method for early warning the yaw fault of the fan based on the random forest algorithm provided by the invention relates to how to select the characteristic set of the yaw system of the wind turbine generator, preprocess data, select characteristic parameters, construct the random forest algorithm, evaluate the algorithm and predict the result. The method can predict the fault of the wind turbine yaw system in real time, is resistant to overfitting, high in prediction accuracy, concise and easy to implement, reliable in result and free of the influence of human factors. The method has the advantages that the early characteristic excavation of yaw faults is realized, so that the adverse effect of mechanical faults of the yaw system of the wind generating set is reduced, the hidden danger of fan operation is reduced, and the method has important significance for the safe and stable operation of the wind generating set.
Drawings
FIG. 1 is a process flow diagram of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art. It should be noted that the embodiments and features of the embodiments may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.
The working principle of the yaw system of the wind generating set is as follows:
the working principle of the yaw system of the wind generating set is as follows: the yaw detection mechanism transmits the wind speed and wind direction signals of the current unit to the PLC; the PLC calculates the current wind angle of the unit, the PLC internal control logic judges whether the unit starts yawing, and when the yawing starting condition is reached, the PLC sends out a control instruction. Firstly, yaw motor electronic brake is opened, a yaw brake system hydraulic station is decompressed, then the PLC sends yaw soft start enabling signals and yaw contactor control signals (left deviation and right deviation), then sends yaw soft start enabling signals, then the yaw motor starts to operate, the yaw motor drives the engine room to rotate on a yaw fluted disc through a yaw pinion after being decelerated by a yaw reducer, yaw action is completed, and the yaw fluted disc is fixed on a fan tower cylinder and is connected with the engine room through a yaw bearing.
Wind generating set driftage system architecture:
the yawing system of the wind generating set comprises three parts, namely a yawing detection mechanism (an anemoscope, a wind vane and a yawing encoder), a yawing control mechanism (a PLC and an electric control system) and a yawing execution mechanism (a yawing driving motor, a yawing speed reducer, a yawing pinion, a yawing bearing fluted disc, a yawing bearing, a lubricating system, a yawing braking system and the like).
Based on the working principle and the system structure of the wind turbine yaw system, the invention clarifies the application technical method by the following steps:
the data source of the invention can adopt a real-time database of a fan SCADA system, a regional production centralized control system or a group production real-time supervision system. The method comprises the steps of firstly determining yaw characteristic data set parameters of a fan of a certain type, then selecting historical data meeting conditions, constructing a characteristic data set (X) and a result set (y), constructing a random forest according to a CART algorithm, and finally early warning new wind turbine generator yaw faults according to a constructed random forest model. The implementation steps of the specific application method for the random forest early warning wind turbine yaw fault are introduced as follows:
(1) the data source is investigated, and the general data are from a regional production centralized control system with three areas for safe production.
(2) Data acquisition
The communication protocol for data acquisition is determined, and data is generally acquired directly from a production centralized control system mirror image to a three-region real-time database. The default data sampling frequency is around 3 seconds. And conventional mature data communication protocols such as OPC or 104 can also be adopted to ensure the stability, safety and reliable quality of data.
(3) Selection of yaw characteristic parameters of wind turbine generator
According to the yaw system monitoring data of most of domestic wind turbines, the basic characteristic parameters are determined as follows:
TABLE A characteristic parameter table of fan yaw system
Figure BDA0003322466650000071
Figure BDA0003322466650000081
(4) Data pre-processing
Wind power plants are generally distributed in places with little smoke, such as coastal islands, remote mountain areas, gobi beach areas and the like, and are in severe environments with large climate temperature difference, much sand and dust, variable wind conditions, high turbulence, low density and the like. Therefore, the collected data must be preprocessed to remove some obviously wrong data, and the following method is generally adopted:
1. null processing
Due to the fact that the high-altitude wind direction is unstable, the unit frequently and intermittently drifts or communication is interrupted, the SCADA data can have null values, and the data are directly removed.
2. Outlier processing
The abnormal values in the SCADA data mainly refer to the condition that certain uncontrollable parameters such as wind speed, air temperature and the like exceed a normal range, for example, the wind speed is less than 0, and the like, and the data are generally deleted. In addition, for data which obviously deviates from normal values, abnormal values can be estimated by using a probability statistical analysis method, and values at nearby moments are used for replacing the abnormal values.
3. Abnormal condition data processing
And discarding abnormal working condition data such as wind-limit electricity, fault shutdown, storm shutdown and the like, wherein the operation data of the abnormal working condition data seriously deviate from a normal design value, and screening and deleting the abnormal working condition data according to logic judgment of the SCADA.
4. Data normalization processing
And carrying out normalization processing on the screened data. The following formula is adopted to carry out linear transformation on the original data, and the original data are mapped into a range of [ 0, 1 ] in a centralized manner.
Figure BDA0003322466650000091
Wherein x ismaxIs the maximum value, x, in the sample dataminIs the minimum value in the sample data.
(5) Constructing a data feature matrix
1. Acquiring a characteristic data set of a plurality of marker post fans in the wind power plant in a yawing normal state period, and recording each recorded data as (x)i1,xi2,xi3，...,xin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval data as XzI.e. by
Figure BDA0003322466650000092
Wherein m is the number of the benchmark fans; simultaneously, a one-dimensional scalar (scalar value is '0') result set with the same quantity is constructed and recorded as Yz。
2. Acquiring a characteristic data set of a yaw fault state period of a plurality of marker post fans of a wind power plant, and recording each recorded data as (g)i1,gi2,gi3，...,gin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval data as GzI.e. by
Figure BDA0003322466650000093
Wherein m is the number of the benchmark fans; simultaneously, a one-dimensional scalar (scalar value is 1) result set with the same quantity is constructed and recorded as Yg。
3. Acquiring a characteristic data set of a yaw warning state period of a plurality of marker post fans in the wind power plant, and recording each recorded data as (e)i1,ei2,ei3，...,ein) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all the time interval data of the fan with normal yaw as EzI.e. by
Figure BDA0003322466650000094
Wherein m is the number of the benchmark fans; simultaneously, a one-dimensional scalar (scalar value is '2') result set with the same quantity is constructed and recorded as Ye。
4. The above feature data set [ X ]z，Gz，EzMerging and constructing a representation mode of a two-dimensional feature matrix, and recording the representation mode as X; namely, it is
Figure BDA0003322466650000101
At the same time, the above result set [ Y ]z，Yz，YzAnd the expression mode of merging and constructing the one-dimensional result matrix is marked as y. Namely, it is
Figure BDA0003322466650000102
(6) Selection of characteristic parameters
And (4) selecting characteristics of the Gini coefficient (GINI) by the constructed characteristic data set [ X ] and a result set [ y ] corresponding to the characteristic data set [ X ]. The criteria for selection of the kini coefficients is that each child node reaches the highest purity, i.e., all observations falling within the child node belong to the same class, at which time the kini coefficient is the smallest, the highest purity, and the smallest uncertainty. Given a total of K classes of result sets, the probability that a sample belongs to class K is: p is a radical ofkThen the kini index of the probability distribution is:
Figure BDA0003322466650000103
(7) construction of random forest algorithm model
And converting the constructed characteristic data set (X) And result set (y) into a random forest model according to a Classification And Regression Tree (namely, a Classification Regression Tree algorithm, namely a CART algorithm for short). The process of constructing a random forest according to the CART algorithm is roughly as follows:
1. selecting a characteristic variable X from a characteristic data set [ X ]iA value of (3), then x is selectediA value v ofi，viDividing the n-dimensional space into two parts, all points of one part satisfying xi＜＝viAll the points of the other part satisfy xi＞viFor non-continuous variables, the attribute values have only two values, i.e., equal to or not equal to the value.
2. And performing recursive processing, namely reselecting an attribute for continuous division according to the two parts of value domains obtained in the previous step until the whole n-dimensional space is completely divided.
3. The division criterion is determined according to equation (2).
(8) Algorithm model evaluation
The method mainly uses Mean Square Error (MSE), Mean Absolute Error (MAE), mean absolute error percentage (MAPE) and a decision coefficient R2 as evaluation indexes.
The smaller the value of MSE, the better, the formula is:
Figure BDA0003322466650000111
in the formula, yiData No. i of the test data set;
Figure BDA0003322466650000112
the predicted value of the No. i data of the test data set is obtained.
The smaller the value of MAE, the better, the formula is:
Figure BDA0003322466650000113
the smaller the MAPE value, the better, the formula:
Figure BDA0003322466650000114
the value of R2 is between 0-1, and the larger the value in the range, the better, the formula is:
Figure BDA0003322466650000115
(9) prediction and analysis
1. And (5) acquiring a new wind turbine generator yaw characteristic parameter set according to the mode in the step (5), and constructing new wind turbine generator yaw two-dimensional characteristic matrix data.
2. And (4) according to the random forest model constructed in the step (7), transmitting newly constructed fan yaw two-dimensional characteristic matrix data to calculate a new result, namely predicting whether yaw faults or yaw alarms need to be generated in the future. And voting the new calculation result by the independent decision tree in the random forest model, namely selecting the final calculation result with the maximum independent decision tree prediction result in the random forest model.
3. In order to avoid prediction errors, data can be collected for multiple times to carry out multiple rounds of calculation, and the random forest model can also be optimized and reconstructed, for example, the number of random forests is increased, and the threshold value of the hyper-parameter is adjusted.
Although the invention has been described in detail hereinabove with respect to a general description and specific embodiments thereof, it will be apparent to those skilled in the art that modifications or improvements may be made thereto based on the invention. Accordingly, such modifications and improvements are intended to be within the scope of the invention as claimed.

Claims (8)

1. An application method for early warning of fan yaw faults based on a random forest algorithm is characterized by comprising the following steps:
1) creating a characteristic data parameter table of a yaw system of the wind generating set, 2) obtaining a characteristic data set of a plurality of marker post fans in a yaw normal state period of the wind power plant, and recording each piece of recorded data as (x)i1,xi2,xi3，...,xin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval dataIs XzI.e. by
Figure FDA0003322466640000011
Wherein m is the number of the benchmark fans; simultaneously constructing one-dimensional scalar result sets with the same quantity, and recording the result sets as Yz；
3) Acquiring a characteristic data set of a yaw fault state period of a plurality of marker post fans of a wind power plant, and recording each recorded data as (g)i1,gi2,gi3，...,gin) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all normal yaw fan time interval data as GzI.e. by
Figure FDA0003322466640000012
Wherein m is the number of the benchmark fans; simultaneously constructing one-dimensional scalar result sets with the same quantity, and recording the result sets as Yg；
4) Acquiring a characteristic data set of a yaw warning state period of a plurality of marker post fans in the wind power plant, and recording each recorded data as (e)i1,ei2,ei3，...,ein) I is expressed as the ith station; n-26, representing the characteristic parameters in table a; recording all the time interval data of the fan with normal yaw as EzI.e. by
Figure FDA0003322466640000013
Wherein m is the number of the benchmark fans; simultaneously, a one-dimensional scalar (scalar value is '2') result set with the same quantity is constructed and recorded as Ye；
5) The above feature data set [ X ]z，Gz，EzMerging and constructing a representation mode of a two-dimensional feature matrix, and recording the representation mode as X; namely, it is
Figure FDA0003322466640000014
At the same time, the above result set [ Y ]z，Yz，YzRepresentation of the merged construction into a one-dimensional result matrix, denoted y, i.e.
Figure FDA0003322466640000021
6) The characteristic selection of the damping coefficient is carried out on the constructed characteristic data set (X) and a result set (y) corresponding to the characteristic data set (X), the selection standard of the damping coefficient is that each child node achieves the highest purity, namely all observations falling into the child nodes belong to the same classification, and at the moment, the damping coefficient is the smallest, the purity is the highest, and the uncertainty is the smallest; converting the constructed characteristic data set (X) And result set (y) into a random forest model according to a Classification And Regression Tree (namely, a Classification Regression Tree algorithm, namely a CART algorithm for short); acquiring a new wind turbine generator yaw characteristic parameter set, and constructing new wind turbine generator yaw two-dimensional characteristic matrix data;
9) according to the random forest model constructed in the step 7), transmitting the data into the new wind turbine generator yaw characteristic two-dimensional matrix data constructed in the step 8) to calculate a new result, namely predicting whether yaw faults or yaw alarms need to be generated in the future; and voting the new calculation result by the independent decision tree in the random forest model, namely selecting the final calculation result with the maximum independent decision tree prediction result in the random forest model.
2. The application method for early warning of fan yaw fault based on random forest algorithm according to claim 1, wherein in the step 1), the characteristic data parameters are as follows:
TABLE A characteristic data parameter table of fan yaw system
Figure FDA0003322466640000022
Figure FDA0003322466640000031
3. The application method for early warning of the yaw fault of the fan based on the random forest algorithm is characterized in that in the step 2), the scalar value of the one-dimensional scalar is 0.
4. The application method for early warning of the yaw fault of the fan based on the random forest algorithm is characterized in that in the step 3), the scalar value of the one-dimensional scalar is 1.
5. The application method for early warning of the yaw fault of the fan based on the random forest algorithm is characterized in that in the step 4), the scalar value of the one-dimensional scalar is 2.
6. The application method for early warning of fan yaw fault based on random forest algorithm as claimed in claim 1, wherein in step 6), if there are K types of result sets in total, the probability that the sample belongs to the kth type is: p is a radical ofkThen the kini index of the probability distribution is:
Figure FDA0003322466640000032
7. the application method for early warning of fan yaw fault based on random forest algorithm as claimed in claim 6, wherein in step 7), the process of constructing random forest according to CART algorithm is as follows:
71) selecting a characteristic variable X from a characteristic data set [ X ]iA value of (3), then x is selectediA value v ofi，viDividing the n-dimensional space into two parts, all points of one part satisfying xi＜＝viAll the points of the other part satisfy xi>viFor non-continuous variables, the attribute values have only two values, namely equal to the value or not equal to the value;
72) recursion processing, namely, reselecting an attribute from the two part value domains obtained in the previous step to continue dividing until the whole n-dimensional space is divided;
73) the criteria for the division are determined according to equation (1).
8. The application method for early warning of fan yaw fault based on random forest algorithm as claimed in claim 1, further comprising step 10) of collecting data for multiple rounds of calculation or performing optimized reconstruction on the random forest model, namely increasing the number of random forests and adjusting the threshold of the hyper-parameter, in order to avoid prediction error.
CN202111250893.1A 2021-10-26 2021-10-26 Application method for early warning of fan yaw fault based on random forest algorithm Withdrawn CN113989555A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111250893.1A CN113989555A (en) 2021-10-26 2021-10-26 Application method for early warning of fan yaw fault based on random forest algorithm

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111250893.1A CN113989555A (en) 2021-10-26 2021-10-26 Application method for early warning of fan yaw fault based on random forest algorithm

Publications (1)

Publication Number Publication Date
CN113989555A true CN113989555A (en) 2022-01-28

Family

ID=79741980

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111250893.1A Withdrawn CN113989555A (en) 2021-10-26 2021-10-26 Application method for early warning of fan yaw fault based on random forest algorithm

Country Status (1)

Country Link
CN (1) CN113989555A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114384885A (en) * 2022-03-23 2022-04-22 希望知舟技术(深圳)有限公司 Process parameter adjusting method, device, equipment and medium based on abnormal working conditions

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114384885A (en) * 2022-03-23 2022-04-22 希望知舟技术(深圳)有限公司 Process parameter adjusting method, device, equipment and medium based on abnormal working conditions

Similar Documents

Publication Publication Date Title
CN111237134B (en) Offshore double-fed wind driven generator fault diagnosis method based on GRA-LSTM-stacking model
CN108680358A (en) A kind of Wind turbines failure prediction method based on bearing temperature model
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
CN110594107B (en) Wind turbine generator fault detection method and device based on rapid gradient elevator
CN114215706B (en) Early warning method and device for cracking faults of wind turbine generator blades
CN111091298B (en) Wind power plant flow field coupling characteristic evaluation and intelligent grouping method and system
CN112994101B (en) Neural network-based wind power plant generated power post-evaluation and monitoring method
CN111415070A (en) Wind turbine generator gearbox oil temperature over-temperature fault early warning method based on SCADA data
CN111209934A (en) Fan fault prediction and alarm method and system
CN113107785A (en) Real-time monitoring method and device for power performance abnormity of wind turbine generator
CN111814848B (en) Self-adaptive early warning strategy design method for temperature faults of wind turbine generator
CN113989555A (en) Application method for early warning of fan yaw fault based on random forest algorithm
CN113468728B (en) Variable pitch system fault prediction method based on neural network
CN116771610A (en) Method for adjusting fault evaluation value of variable pitch system of wind turbine
CN113946977A (en) Application method for early warning of fan variable pitch fault based on decision tree algorithm
CN115842408A (en) Wind power plant operation state detection system and method based on SCADA
CN115450864A (en) Wind turbine blade icing diagnosis method based on synthesis of minority samples
CN114065618A (en) Method for fitting fan power curve based on parameter model of differential evolution
CN114912807A (en) Method and system for evaluating generated energy improving effect of technically improved wind turbine generator
CN112801352A (en) Power prediction method for cooperative control in offshore wind farm
Zhang et al. An unsupervised anomaly detection approach based on industrial big data
Qian et al. A weighted kNN fault detection based on multistep index and dynamic neighborhood scale under complex working conditions
CN114021992B (en) Evaluation method for real-time output performance of large wind turbine generator
Liu et al. Wind Turbine Fault Diagnosis Based on Feature Selection and Stacking Model Fusion with Small-scale Data.

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
WW01 Invention patent application withdrawn after publication

Application publication date: 20220128

WW01 Invention patent application withdrawn after publication