CN117851877A - Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium - Google Patents

Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium Download PDF

Info

Publication number
CN117851877A
CN117851877A CN202311540515.6A CN202311540515A CN117851877A CN 117851877 A CN117851877 A CN 117851877A CN 202311540515 A CN202311540515 A CN 202311540515A CN 117851877 A CN117851877 A CN 117851877A
Authority
CN
China
Prior art keywords
data
wind power
power blade
value
blade
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
CN202311540515.6A
Other languages
Chinese (zh)
Inventor
王岁岁
周勃
常丽
孙宁
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.)
Shenyang University of Technology
Original Assignee
Shenyang University of Technology
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 Shenyang University of Technology filed Critical Shenyang University of Technology
Priority to CN202311540515.6A priority Critical patent/CN117851877A/en
Publication of CN117851877A publication Critical patent/CN117851877A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • 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
    • G06F18/23—Clustering techniques

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Wind Motors (AREA)

Abstract

The invention provides a wind power blade fracture early warning method based on SCADA data association analysis. The method comprises the following steps: and (3) reducing the dimension of SCADA data before and after the wind power blade breaks, and screening out each parameter related to the running state of the wind power blade for the first time. And carrying out data cleaning on the original data corresponding to the screened parameters in the SCADA data in the normal running state of the wind power blade, extracting the data corresponding to each screened parameter in the SCADA data after data cleaning, dividing the extracted data into a training set and a verification set, and carrying out NSET modeling by using the training set. And verifying and further screening parameters by utilizing historical data to finally screen out each parameter related to the running state of the wind power blade. And outputting residual errors according to the model by inputting real-time observation data to obtain Euclidean distance curves so as to judge the running condition of the wind power blade at the moment. The invention fully excavates the threshold value of the early warning of the blade breakage on the basis of the existing SCADA data, thereby achieving the purpose of on-line real-time monitoring.

Description

Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium
Technical Field
The invention relates to the technical field of wind power blade monitoring, in particular to a wind power blade fracture early warning method based on SCADA data association analysis, a computer device and a computer readable storage medium.
Background
The wind power blade is a key component for converting wind energy into mechanical energy and finally driving a generator to generate power, and in the maintenance cost ratio of each component of wind power equipment, the wind power blade accounts for 29.55% at the highest, and besides, the value of the wind power blade accounts for more than two of the whole machine set. Wind blades operate under severe conditions, and despite their design and testing life of 20-25 years, experience has shown that preventive maintenance is required to achieve this goal. With the continuous development of wind power generation industry and the increasing social electricity demand, the large development trend of the wind turbine generator is more and more obvious, and the risk of blade fracture is increased when the length of the blade is increased. The breakage of wind power blades not only can cause the burning of the wind power generation set, but also can influence other sets due to flying blade debris, and the potential safety hazard is increased while huge economic loss is caused. Therefore, the fracture monitoring of the wind power blade is particularly important.
With the development trend of large-scale and large-scale wind turbine generators in China, the damage degree assessment standard of the wind turbine generators is difficult to uniformly quantify due to the large differences of wing shapes, structures, materials and appearance of the wind turbine generators of different types, and the failure fracture threshold value is dynamically changed under different wind field conditions, so that the problem of dynamically setting the fracture threshold value is a difficult problem for real-time fracture monitoring of the large wind turbine generators. In addition, the existing integrated forming technology of the wind power blade and the intelligent sensor is still immature, if various sensors are additionally arranged in the inner cavity or the outer wall skin of the giant wind power blade, the sensors are easy to fall off or lose efficacy, and the cost and the manufacturing cost of the blade monitoring system are also uncontrollable. Therefore, the wind power blade fracture early warning method can ensure universality and wider application range by considering the dynamic change of the fracture alarm threshold value and the cost problem.
At present, existing wind turbine generator monitoring systems at home and abroad mostly adopt artificial intelligent algorithms without models, rely on a large amount of sample data and the accuracy of the models, and the screening of modeling parameters is often not based, and the reliability analysis and verification are also lacking in the calculation process. Particularly, the blade monitoring system collects the existing SCADA system sensor signals, and the problems of large data volume, large dimension, multiple redundant variables and the like of data cleaning can cause low on-line monitoring efficiency, so that the accuracy of a model is reduced, and real-time on-line monitoring is difficult.
Disclosure of Invention
The present invention aims to solve at least one of the technical problems existing in the prior art or related art.
Therefore, a first object of the present invention is to provide a wind power blade fracture early warning method based on the correlation analysis of SCADA data.
A second object of the present invention is to provide a computer apparatus.
A third object of the present invention is to propose a computer readable storage medium.
In order to achieve the above object, according to a first aspect of the present invention, there is provided a wind turbine blade breakage early warning method based on association analysis of SCADA data, where the SCADA data is data collected by a sensor group in a SCADA system of a wind turbine generator, and the SCADA data includes parameters as follows: the power of the wind turbine generator, the rotation speed of the rotor, the angle of wind power blades, the current of the grid side and the wind speed; the early warning method comprises the following steps: acquiring a first set of SCADA data before and after the wind power blade breaks for a certain period of time; screening the first set of SCADA data by using chi-square verification to screen out various parameters related to the running state of the wind power blade by using chi-square verification; adding the wind speed parameter into each parameter which is screened out by using chi-square verification and is related to the running state of the wind power blade, and screening each parameter which is related to the running state of the wind power blade for the first time; acquiring a second set of SCADA data of the wind power blade in a normal operation state for a certain period of time; screening each wind power blade running state related to the first time in the second set of SCADA data Extracting original data corresponding to the parameters; fitting a wind speed-power scatter diagram according to the wind speed data in the second set of SCADA data and the power data of the wind turbine generator after data extraction; performing data cleaning on the wind speed-power scatter diagram by adopting a DBSCAN clustering algorithm; dividing the second group of SCADA data after data cleaning into a first training set and a first verification set; building a first NSET model of the wind power blade in a normal running state according to the first training set and the first verification set; inputting an original first set of SCADA data into the first NSET model, and calculating a residual error E between a predicted value output by the first NSET model and a corresponding observed value n Fitting a first Euclidean distance curve corresponding to the first group of SCADA data according to the predicted value and the observed value so as to observe that the first Euclidean distance curve starts to be in an ascending trend after a certain position of the fracture point delay; from the residual error E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i The method comprises the steps of carrying out a first treatment on the surface of the Removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time; the data corresponding to each parameter related to the running state of the wind power blade is screened out for the second time in the second set of SCADA data after data cleaning, the extracted data is divided into a second training set and a second verification set; establishing a second NSET model of the wind power blade in a normal running state according to the second training set and the second verification set; acquiring a Euclidean distance curve corresponding to the second verification set, and taking the maximum value of the Euclidean distance on the Euclidean distance curve corresponding to the second verification set as a first threshold; inputting the original first set of SCADA data into the second NSET model, and calculating a residual error E between a predicted value output by the second NSET model and a corresponding observed value n ' and fitting the predicted and observed values to the modelA second Euclidean distance curve corresponding to the first set of SCADA data, so as to observe that the position of the second Euclidean distance curve at the breaking point starts to be in an ascending trend; acquiring a maximum value of Euclidean distance on a second Euclidean distance curve before the breaking point as a second threshold value; acquiring the maximum value of the Euclidean distance on a second Euclidean distance curve after the breaking point as a third threshold; SCADA data acquired by a sensor group in an SCADA system of the wind turbine generator is acquired in real time: inputting SCADA data obtained in real time into the second NSET model, and calculating residual error E between predicted value output by the second NSET model and corresponding observed value n Fitting a third Euclidean distance curve corresponding to the SCADA data obtained in real time according to the predicted value and the observed value; and carrying out fracture early warning on the wind power blades of the wind turbine generator set according to the values of the corresponding Euclidean distances on the third Euclidean distance curve, and the first threshold value, the second threshold value and the third threshold value.
Preferably, the performing fracture early warning on the wind turbine blade of the wind turbine generator according to the value of the euclidean distance on the third euclidean distance curve, and the first threshold, the second threshold and the third threshold specifically includes: determining the range of the value of each corresponding Euclidean distance on the third Euclidean distance curve; when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is more than or equal to 0 and less than or equal to the first threshold value, judging that the wind power blade of the wind turbine generator is in a normal running state; when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than the first threshold value and smaller than or equal to the second threshold value, judging that the wind power blade of the wind power generation set is in an abnormal state; and when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than or equal to the third threshold value, judging that the wind power blade of the wind power generation set is in a broken state.
Preferably, the parameters related to the running state of the wind power blade, which are screened by using chi-square verification, are respectively: the power of the wind turbine generator, the rotation speed of the rotor, the wind power blade angle, the first grid side current, the second grid side current and the third grid side current; the first screening of each parameter related to the running state of the wind power blade is as follows: wind speed, power of a wind turbine generator, generator rotation speed, rotor rotation speed, wind power blade angle, first grid side current, second grid side current and third grid side current;
said residual error E n The expression of (2) is:
E n ＝x n -x n *
wherein x is n Observations of an nth parameter in the first set of SCADA data; x is x n * A predicted value corresponding to an nth parameter in the first set of SCADA data; e (E) n The difference between the observed value of the nth parameter and the predicted value corresponding to the nth parameter; x is x 1 To x 7 Verifying each screened parameter related to the running state of the wind power blade for the using chi-square; said determining said residual E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i The method specifically comprises the following steps:
from the residual error E n Calculating the failure times e of the parameters related to the running state of the wind power blade screened by using chi-square verification c ；
The number of faults e c The expression of (2) is:
let a = observation vector group number/observation vector group number per input, and according to the number of failures e c Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i ；
The failure accumulation contribution rate e i The expression of (2) is:
removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The method comprises the steps of adding a parameter smaller than a certain set value into each parameter related to the running state of the wind power blade after removal, and screening out each parameter related to the running state of the wind power blade for the second time, wherein the method specifically comprises the following steps: removing x 1 To x 7 The failure accumulation contribution rate e in (2) i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time; wherein, each parameter related to the running state of the wind power blade is screened out for the second time, and the parameters are respectively as follows: wind speed, power of a wind turbine generator, generator speed, rotor speed and wind turbine blade angle.
Preferably, the establishing a first NSET model of the wind power blade in a normal running state according to the first training set and the first verification set specifically includes: acquiring an original NSET model, and inputting the first training set into the original NSET model to obtain a trained NSET model;
representing the whole observation matrix of the wind turbine generator as M with the size of n multiplied by b n×b The M is n×b The expression of (2) is:
wherein n is a time state, and b is the number of observation variables in each time; matrix M n×b The row vector is X i ＝[x i (t 1 ) x i (t 2 )...x i (t b )]Matrix M n×b The row vector being a given observation parameter X i All observations within a certain observation period; matrix M n×b Is X (t) j )＝[x 1 (t j ) x 2 (t j )...x b (t j )] T Matrix M n×b Is t j Time of dayThe observed value of the observed parameter exists; from said M n×b Selecting a parameter of a period of time to be recorded as a historical observation matrix K, wherein the historical observation matrix K is the health state of each observation parameter, and the expression of the historical observation matrix K is as follows:
selecting a part of state data from the history observation matrix K, and forming a process memory matrix D by using the selected part of state data n . Process matrix D n Can be expressed as:
to observe matrix X obs And the observation matrix X obs Corresponding D n Input to the expression In (1) obtaining a prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the Setting the first validation set to an observation matrix X obs The method comprises the steps of carrying out a first treatment on the surface of the Inputting the first validation set into a trained NSET model; the NSET model after inputting the first verification set outputs a corresponding prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the Calculating an observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Residual errors between the first verification set and the observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the first verification set to establish a first NSET model of the wind power blade in a normal running state; and establishing a second NSET model of the wind power blade in a normal running state according to the second training set and the second verification set, wherein the method specifically comprises the following steps: acquiring an original NSET model, and inputting the second training set into the original NSET model to obtain a trained NSET model; setting the second validation set to an observation matrix X obs The method comprises the steps of carrying out a first treatment on the surface of the Inputting the second validation set into a trained NSET model; after entering the second verification setNSET model outputs corresponding prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the Calculating an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Residual errors between the two, and an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the second verification set to establish a second NSET model of the wind power blade in a normal running state.
Preferably, a part of state data is selected from the historical observation matrix K, and the selected part of state data is used for forming the process memory matrix D n The method specifically comprises the following steps: setting each observation vector of the history observation matrix K to be composed of n variables; for each of the n variables, 0,1 will be]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n In (a) and (b);
the pair of each of the n variables will be [0,1 ]]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n Specifically, the method comprises the following steps: setting i=1; wherein i is a positive integer; perform a=1/h i; wherein h is a positive integer; setting k=1; wherein k is a positive integer; judging whether the absolute value of the (X (k) -A is smaller than delta; wherein δ is a positive number; when |X (k) -A| is less than delta, X (k) is added to the process memory matrix D n In (a) and (b); when |x (k) -a| is equal to or greater than δ, judging whether k is larger than M; wherein M is the number of columns of the history observation matrix K; when k is less than or equal to M, executing k=k+1, and returning to the step of judging whether the |x (k) -a| is less than delta; when k is greater than M, judging whether i is greater than h; when i is less than or equal to h, executing i=i+1, and returning to the step of executing a=1/h×i; when i is greater than h, execution ends.
Preferably, the screening of the first set of SCADA data using chi-square verification to screen out parameters related to the running state of the wind power blade using chi-square verification specifically includes: establishing an original hypothesis H0, which isH0 is independent between the running state of the wind power blade and each parameter in the first set of SCADA data; taking the data of the running state of the wind power blade as a first variable threshold value; taking one parameter in the first set of SCADA data as a second variable threshold at each time; respectively recording the actual value of the data number of the first variable threshold under the condition that the wind power blade is normal as a, the actual value of the data number of the first variable threshold under the condition that the wind power blade is faulty as b and the actual value of the total data number of the wind power blade as a+b; the actual value of the data number of the y-th parameter in the SCADA data under the normal condition of the wind power blade is recorded as c y The actual value of the data number under the fault of the wind power blade is d y And the actual value of the total data number of the wind power blade is c y +d y Wherein y is a positive integer; recording the actual value of the total data number of the first variable threshold value and the second variable threshold value under the normal condition of the wind power blade as a+c respectively y And the actual value of the total number of data in the wind power blade fault is b+d y And the actual value of the total data number of the wind power blades of the two is a+b+c y +d y The method comprises the steps of carrying out a first treatment on the surface of the Respectively calculating to obtain theoretical value of the number of data of the first variable threshold under the condition that the wind power blade is normal as (a+b) × (a+c) y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (a+b) x (b+d) y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is a+b; respectively calculating to obtain the theoretical value of the number of the data of the second variable threshold under the normal condition of the wind power blade as (c) y +d y )×(a+c y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (c y +d y )×(b+d y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is c y +d y The method comprises the steps of carrying out a first treatment on the surface of the Setting the degree of freedom to be 1; calculating a chi-square value according to a chi-square value calculation formula, wherein the chi-square value is used for measuring the difference degree between each actual value and each theoretical value, and the chi-square value calculation formula is as follows: x-shaped articles 2 ＝∑(A-T) 2 /T
Wherein χ is 2 The value is chi square, A is the actual value, and T is the theoretical value; finding a corresponding P value according to the degree of freedom and the chi-square value table, wherein the P value is the probability of making a first class of false rejection errors; and selecting one parameter in the corresponding first set of SCADA data when the P value is smaller than 0.05 as a parameter related to the running state of the wind power blade, so as to sequentially screen out each parameter related to the running state of the wind power blade.
Preferably, the data cleaning of the wind speed-power scatter diagram by using a DBSCAN clustering algorithm specifically includes: taking all points in the wind speed-power scatter plot as sample data S, and marking each data point in the sample data S as an unprocessed state; assigning initial values to Eps and Minpts, wherein Eps is a neighborhood distance threshold value of a certain data point p in the sample data S, and Minpts is the minimum number of data points in the neighborhood with the radius of Eps of the certain data point p in the sample data S; setting the neighborhood with the radius of Eps of a certain data point p as N Eps (p); clustering the high-density areas formed by the Eps and the Minpts; the clustering of the high-density region formed by the Eps and the Minpts specifically comprises the following steps: determining whether the certain data point p has been added to a certain cluster or has been listed as noise; if the certain data point p has been added to a certain cluster or has been listed as noise, the classification ends; if the data point p is not added to a cluster and is not listed as noise, N is determined Eps (p) whether there are at least Minpts objects within it; if N Eps At least Minpts objects exist in (p), a new class cluster U is constructed, and a certain data point p is added in the U; if N Eps And (4) if the number of objects in the data point (p) is less than Minpts, the data point (p) is listed as a boundary point or noise.
Preferably, the Eps is set to 4.5 and the Minpts is set to 18.5.
According to a second aspect of the present invention, there is further provided a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the steps of the wind power blade breakage early warning method based on the correlation analysis of SCADA data in any of the above-mentioned aspects are implemented when the processor executes the computer program.
According to a third aspect of the present invention, there is further provided a computer readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the steps of the wind turbine blade breakage early warning method based on correlation analysis of SCADA data in any of the above-mentioned aspects.
The invention has the beneficial effects that:
1. according to the wind power blade fracture early warning method based on the correlation analysis of the SCADA data, a plurality of technical means are adopted aiming at the blank of the prior art, so that the technical problem that the wind power blade cannot utilize the prior SCADA data to monitor blade fracture in real time is overcome. Specifically, the data collected by the sensor group in the SCADA system of the wind turbine generator set contains a plurality of redundant variables, which can cause the problem that the modeling time is increased, the efficiency is reduced and the accuracy of the model is reduced. Aiming at the problems of large SCADA data quantity and large dimension, the invention provides a method for preliminarily screening the SCADA data related to the running state of the blade by using a chi-square verification method, thereby realizing the dimension reduction processing of the SCADA data and improving the calculation efficiency of a monitoring system.
2. And clustering the wind speed-power scatter diagram by adopting a DBSCAN algorithm aiming at null values, singular points, power limiting data and noise data in the SCADA data, and removing interference data such as shutdown data, power limiting data, noise and the like to clean the data.
3. In the prior art, most of monitoring systems depend on sample data and model accuracy, modeling parameters are often not screened without basis, and reliability analysis and verification are also lacking in a calculation process.
4. Aiming at the problem that the damage of different degrees of the blade is difficult to quantify, the invention provides a semi-supervised learning model established by using SCADA data of a normal running state of the wind power blade, and adopts a data fusion method to carry out data fusion on wind speed, power, generator rotating speed, rotor rotating speed and blade angle, wherein the whole data fusion result in the normal state of the blade is used as a comparison parameter for judging the fracture of the blade, and Euclidean distance between real-time monitoring data of the blade and the normal state is used as a quantitative index of the fracture of the blade.
5. Aiming at the problem of dynamic setting of a blade fracture early warning threshold, the invention inputs SCADA data before and after the wind power blade fracture for a certain period of time into a model based on the existing SCADA system history data of the wind turbine without adding other sensors, and provides a method for comparing and analyzing Euclidean distance curves before the wind power blade fracture by using the SCADA data before and after the wind power blade fracture to obtain the maximum Euclidean distance before the wind power blade fracture early warning threshold of the model, thereby updating the fracture threshold of the running service state of the blade in real time.
In summary, the invention provides the method for carrying out association quantification preliminary screening on the blade and the sensor signal in the SCADA system through chi-square verification, then carrying out further verification screening on the input parameters of the model by utilizing fault data, finally establishing an NSET model of the screened SCADA parameters under the normal state of the wind power blade as a semi-supervision model to monitor the running state of the blade, utilizing the verification modeling result of the known wind power blade fracture dataset and setting an early warning threshold, and fully excavating the early warning threshold of the blade fracture on the basis of the existing SCADA data, thereby achieving the purpose of online real-time and efficient monitoring. The wind power blade fracture early warning method based on the association analysis of the SCADA data can realize online real-time fracture early warning of various wind power blades, has universality, can effectively avoid huge economic loss and potential safety hazard, and has wide popularization and application prospects.
Additional aspects and advantages of the invention will become apparent in the following description or may be learned by practice of the invention.
Drawings
FIG. 1 shows a schematic flow chart of a wind blade breakage pre-warning method based on correlation analysis of SCADA data according to one embodiment of the invention;
FIG. 2 shows a schematic flow chart of parameter screening and modeling of SCADA data of one embodiment of the present invention;
FIG. 3 illustrates a Euclidean distance plot of training model outputs corresponding to unfiltered parameters for one embodiment of the invention;
FIG. 4 shows a Euclidean distance graph of training model output corresponding to parameters screened using chi-square verification and DBSCAN clustering algorithms according to one embodiment of the present invention;
FIG. 5 shows a graph of Euclidean distance of the final model output for one embodiment of the invention;
FIG. 6 shows a schematic flow chart of a process memory matrix construction procedure of one embodiment of the present invention;
FIG. 7 shows a schematic flow chart of wind blade breakage real-time monitoring of an embodiment of the invention;
FIG. 8 is a graph showing the Euclidean distance output when the verification set in the data corresponding to the finally screened parameters is input into the NSET model under the normal working condition of the finally established blade according to one embodiment of the invention;
FIG. 9 illustrates a wind blade early warning threshold split profile for one embodiment of the present invention;
FIG. 10 illustrates a graph of Euclidean distance output when SCADA data from 1 month of original wind turbine blade breakage before and after the input into a NSET model of a final established blade under normal conditions, in accordance with an embodiment of the present invention;
FIG. 11 illustrates a graph of Euclidean distance output when SCADA data from 2 months of an original wind blade before and after breaking is input into a NSET model of a finally established blade under normal conditions, in accordance with an embodiment of the present invention;
FIG. 12 illustrates a graph of Euclidean distance output when SCADA data from 3 months 1 day to 3 months 10 days before and after breaking an original wind blade is input into a NSET model under normal operating conditions of the finally established blade, in accordance with an embodiment of the present invention;
FIG. 13 illustrates a graph of Euclidean distance output when SCADA data from 3 months 10 to 3 months 14 days before and after breaking an original wind blade is input into a NSET model under normal operating conditions of the finally established blade, in accordance with an embodiment of the present invention;
FIG. 14 shows a schematic block diagram of a computer apparatus of one embodiment of the invention.
Detailed Description
In order that the above-recited objects, features and advantages of the present invention will be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description. It should be noted that, without conflict, the embodiments of the present application and features of the embodiments may be combined with each other.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those described herein, and the scope of the invention is therefore not limited to the specific embodiments disclosed below.
FIG. 1 shows a schematic flow chart of a wind blade breakage pre-warning method based on correlation analysis of SCADA data according to one embodiment of the invention. As shown in FIG. 1, the dimension of the historical SCADA data before and after the wind power blade fracture is reduced to screen out each sensor parameter related to the running state of the wind power blade for the first time. And (3) carrying out data cleaning on original data corresponding to the screened parameters in the SCADA data in the normal operation state of the wind power blade for a certain period of time, extracting data corresponding to the screened sensor parameters in the SCADA data in the normal operation state of the wind power blade after data cleaning, dividing the extracted data into a training set and a verification set, and carrying out NSET modeling (NSET modeling) by using the training set to establish a NSET model (establish a normal state model) in the normal operation state of the wind power blade. And verifying and further screening parameters by utilizing the historical data to finally screen out each parameter related to the running state of the wind power blade. Further, the input real-time related observation data can output residual errors according to the model, and an overall Euclidean distance curve and a fitting curve of each parameter predicted value and the observation value are obtained, so that the running condition of the wind power blade at the moment is judged, whether the blade is at risk of fracture is further judged, and if the risk exceeds an early warning threshold value, wind field staff is notified.
Specifically, the SCADA data are data collected by a sensor group in an SCADA system of the wind turbine, and the SCADA data comprise the following parameters: the power of the wind turbine generator, the rotation speed of the rotor, the angle of wind power blades, the current of the grid side and the wind speed; the wind power blade fracture early warning method based on the association analysis of SCADA data comprises the following steps: acquiring a first set of SCADA data before and after the wind power blade breaks for a certain period of time; screening the first set of SCADA data by using chi-square verification to screen out various parameters related to the running state of the wind power blade by using chi-square verification; adding the wind speed parameter into each parameter which is screened out by using chi-square verification and is related to the running state of the wind power blade, and screening each parameter which is related to the running state of the wind power blade for the first time; acquiring a second set of SCADA data of the wind power blade in a normal operation state for a certain period of time; extracting original data corresponding to each parameter related to the running state of the wind power blade, wherein the original data are screened out for the first time from the second set of SCADA data; fitting a wind speed-power scatter diagram according to the wind speed data in the second set of SCADA data and the power data of the wind turbine generator after data extraction; performing data cleaning on the wind speed-power scatter diagram by adopting a DBSCAN clustering algorithm; dividing the second group of SCADA data after data cleaning into a first training set and a first verification set; building a first NSET model of the wind power blade in a normal running state according to the first training set and the first verification set; inputting an original first set of SCADA data into the first NSET model, and calculating a residual error E between a predicted value output by the first NSET model and a corresponding observed value n Fitting a first Euclidean distance curve corresponding to the first set of SCADA data according to the predicted value and the observed value to observe the first Euclidean distance curveThe line starts to be in an ascending trend after a certain position of the breaking point delay; from the residual error E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i The method comprises the steps of carrying out a first treatment on the surface of the Removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time; the data corresponding to each parameter related to the running state of the wind power blade is screened out for the second time in the second set of SCADA data after data cleaning, the extracted data is divided into a second training set and a second verification set; establishing a second NSET model of the wind power blade in a normal running state according to the second training set and the second verification set; acquiring a Euclidean distance curve corresponding to the second verification set, and taking the maximum value of the Euclidean distance on the Euclidean distance curve corresponding to the second verification set as a first threshold; inputting the original first set of SCADA data into the second NSET model, and calculating a residual error E between a predicted value output by the second NSET model and a corresponding observed value n ' fitting a second Euclidean distance curve corresponding to the first set of SCADA data according to the predicted value and the observed value so as to observe that the position of the second Euclidean distance curve at the breaking point starts to be in an ascending trend; acquiring a maximum value of Euclidean distance on a second Euclidean distance curve before the breaking point as a second threshold value; acquiring the maximum value of the Euclidean distance on a second Euclidean distance curve after the breaking point as a third threshold; SCADA data acquired by a sensor group in an SCADA system of the wind turbine generator is acquired in real time: inputting SCADA data obtained in real time into the second NSET model, and calculating residual error E between predicted value output by the second NSET model and corresponding observed value n Fitting a third Euclidean distance curve corresponding to the SCADA data obtained in real time according to the predicted value and the observed value; according to the third Euclidean distance curveAnd carrying out fracture early warning on the wind power blades of the wind turbine generator set by the values of the Euclidean distances corresponding to the lines, and the first threshold value, the second threshold value and the third threshold value.
In this embodiment, the data collected by the sensor group in the SCADA system of the wind turbine generator set at the beginning includes: and the data corresponding to the parameters such as the power of the wind turbine generator, the rotation speed of the rotor, the angle of wind power blades, the current of the grid side, the wind speed and the like.
In a specific embodiment, a certain period of time, which may be a period of one minute, and a time stamp of 0.02s, is obtained from a first set of SCADA data before and after breaking the wind turbine blade for a certain period of time, and the total number of data is 3000. The second NSET model is the NSET model under the normal working condition of the finally established blade.
In an embodiment of the present invention, the performing fracture early warning on the wind turbine blade of the wind turbine generator according to the value of the euclidean distance on the third euclidean distance curve and the first threshold, the second threshold, and the third threshold specifically includes: determining the range of the value of each corresponding Euclidean distance on the third Euclidean distance curve; when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is more than or equal to 0 and less than or equal to the first threshold value, judging that the wind power blade of the wind turbine generator is in a normal running state; when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than the first threshold value and smaller than or equal to the second threshold value, judging that the wind power blade of the wind power generation set is in an abnormal state; and when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than or equal to the third threshold value, judging that the wind power blade of the wind power generation set is in a broken state.
In one embodiment of the invention, the parameters related to the running state of the wind power blade, which are screened by using chi-square verification, are respectively: the power of the wind turbine generator, the rotation speed of the rotor, the wind power blade angle, the first grid side current, the second grid side current and the third grid side current; the first screening of each parameter related to the running state of the wind power blade is as follows: wind speed, power of a wind turbine generator, generator rotation speed, rotor rotation speed, wind power blade angle, first grid side current, second grid side current and third grid side current;
said residual error E n The expression of (2) is:
E n ＝x n -x n *
wherein x is n Observations of an nth parameter in the first set of SCADA data; x is x n * A predicted value corresponding to an nth parameter in the first set of SCADA data; e (E) n The difference between the observed value of the nth parameter and the predicted value corresponding to the nth parameter; x is x 1 To x 7 Verifying each screened parameter related to the running state of the wind power blade for the using chi-square; said determining said residual E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i The method specifically comprises the following steps:
From the residual error E n Calculating the failure times e of the parameters related to the running state of the wind power blade screened by using chi-square verification c ；
The number of faults e c The expression of (2) is:
let a = observation vector group number/observation vector group number per input, and according to the number of failures e c Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i ；
The failure accumulation contribution rate e i The expression of (2) is:
removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The method comprises the steps of adding a parameter smaller than a certain set value into each parameter related to the running state of the wind power blade after removal, and screening out each parameter related to the running state of the wind power blade for the second time, wherein the method specifically comprises the following steps: removing x 1 To x 7 The failure accumulation contribution rate e in (2) i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time; wherein, each parameter related to the running state of the wind power blade is screened out for the second time, and the parameters are respectively as follows: wind speed, power of a wind turbine generator, generator speed, rotor speed and wind turbine blade angle.
In one embodiment of the present invention, the building a first NSET model under the normal running state of the wind turbine blade according to the first training set and the first verification set specifically includes: acquiring an original NSET model, and inputting the first training set into the original NSET model to obtain a trained NSET model;
representing the whole observation matrix of the wind turbine generator as M with the size of n multiplied by b n×b The M is n×b The expression of (2) is:
wherein n is a time state, and b is the number of observation variables in each time; matrix M n×b The row vector is X i ＝[x i (t 1 ) x i (t 2 )...x i (t b )]Matrix M n×b The row vector being a given observation parameter X i All observations within a certain observation period; matrix M n×b Is X (t) j )＝[x 1 (t j ) x 2 (t j )...x b (t j )] T Matrix M n×b Is t j Observing values of all observing parameters at any moment; from said M n×b Selecting a parameter of a period of time as a history observation matrix K, and recording the historyThe observation matrix K is the health state of each observation parameter, and the expression of the history observation matrix K is:
selecting a part of state data from the history observation matrix K, and forming a process memory matrix D by using the selected part of state data n . Process matrix D n Can be expressed as:
to observe matrix X obs And the observation matrix X obs Corresponding D n Input to the expressionIn (1) obtaining a prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the Setting the first validation set to an observation matrix X obs The method comprises the steps of carrying out a first treatment on the surface of the Inputting the first validation set into a trained NSET model; the NSET model after inputting the first verification set outputs a corresponding prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the Calculating an observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Residual errors between the first verification set and the observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the first verification set to establish a first NSET model of the wind power blade in a normal running state; and establishing a second NSET model of the wind power blade in a normal running state according to the second training set and the second verification set, wherein the method specifically comprises the following steps: acquiring an original NSET model, and inputting the second training set into the original NSET model to obtain a trained NSET model; setting the second validation set to an observation matrix X obs The method comprises the steps of carrying out a first treatment on the surface of the Inputting the second validation set into a trained NSET model; outputting a corresponding prediction output matrix X by using the NSET model after the second verification set is input est The method comprises the steps of carrying out a first treatment on the surface of the Calculating an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Residual errors between the two, and an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the second verification set to establish a second NSET model of the wind power blade in a normal running state.
In one embodiment of the present invention, the screening the first set of SCADA data using chi-square verification to screen out parameters related to the running state of the wind turbine blade using chi-square verification specifically includes: establishing an original assumption H0, wherein the original assumption H0 is independent between the running state of the wind power blade and each parameter in the first set of SCADA data; taking the data of the running state of the wind power blade as a first variable threshold value; taking one parameter in the first set of SCADA data as a second variable threshold at each time; respectively recording the actual value of the data number of the first variable threshold under the condition that the wind power blade is normal as a, the actual value of the data number of the first variable threshold under the condition that the wind power blade is faulty as b and the actual value of the total data number of the wind power blade as a+b; the actual value of the data number of the y-th parameter in the SCADA data under the normal condition of the wind power blade is recorded as c y The actual value of the data number under the fault of the wind power blade is d y And the actual value of the total data number of the wind power blade is c y +d y Wherein y is a positive integer; recording the actual value of the total data number of the first variable threshold value and the second variable threshold value under the normal condition of the wind power blade as a+c respectively y And the actual value of the total number of data in the wind power blade fault is b+d y And the actual value of the total data number of the wind power blades of the two is a+b+c y +d y The method comprises the steps of carrying out a first treatment on the surface of the Respectively calculating to obtain theoretical value of the number of data of the first variable threshold under the condition that the wind power blade is normal as (a+b) × (a+c) y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (a+b) x (b+d) y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is a+b; respectively calculating the second variable threshold value in windThe theoretical value of the number of data under the normal condition of the electric blade is (c) y +d y )×(a+c y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (c y +d y )×(b+d y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is c y +d y The method comprises the steps of carrying out a first treatment on the surface of the Setting the degree of freedom to be 1; calculating a chi-square value according to a chi-square value calculation formula, wherein the chi-square value is used for measuring the difference degree between each actual value and each theoretical value, and the chi-square value calculation formula is as follows: x-shaped articles 2 ＝∑(A-T) 2 /T
Wherein χ is 2 The value is chi square, A is the actual value, and T is the theoretical value; finding a corresponding P value according to the degree of freedom and the chi-square value table, wherein the P value is the probability of making a first class of false rejection errors; and selecting one parameter in the corresponding first set of SCADA data when the P value is smaller than 0.05 as a parameter related to the running state of the wind power blade, so as to sequentially screen out each parameter related to the running state of the wind power blade.
In one embodiment of the present invention, the data cleaning of the wind speed-power scatter diagram by using a DBSCAN clustering algorithm specifically includes: taking all points in the wind speed-power scatter plot as sample data S, and marking each data point in the sample data S as an unprocessed state; assigning initial values to Eps and Minpts, wherein Eps is a neighborhood distance threshold value of a certain data point p in the sample data S, and Minpts is the minimum number of data points in the neighborhood with the radius of Eps of the certain data point p in the sample data S; setting the neighborhood with the radius of Eps of a certain data point p as N Eps (p); clustering the high-density areas formed by the Eps and the Minpts; the clustering of the high-density region formed by the Eps and the Minpts specifically comprises the following steps: determining whether the certain data point p has been added to a certain cluster or has been listed as noise; if the certain data point p has been added to a certain cluster or has been listed as noise, the classification ends; if the data point p is not added to a cluster and is not listed as noise, N is determined Eps Whether or not in (p)At least Minpts objects; if N Eps At least Minpts objects exist in (p), a new class cluster U is constructed, and a certain data point p is added in the U; if N Eps And (4) if the number of objects in the data point (p) is less than Minpts, the data point (p) is listed as a boundary point or noise.
In this embodiment, p points are taken as the center, eps is taken as a circle with a radius, whether the number of all points in the circle is greater than the Minpts, if yes, the circle is reserved, and if not, the circle is regarded as noise. N (N) Eps (p) means traversing N points, i.e., all points. The parameter setting of the DBSCAN clustering algorithm is complex, and joint repeated adjustment is needed. In a specific embodiment, after normalizing the data, setting Eps to 4.5, setting mps to 18.5, and removing noise.
FIG. 2 shows a schematic flow chart of parameter screening and modeling of SCADA data of one embodiment of the present invention. As shown in fig. 2, first, the SCADA data before and after the wind power blade breaks for a certain period of time is primarily screened by using chi-square verification, the total fault data number of the blade and the abnormal number of each parameter data in the SCADA when the wind power blade breaks are recorded, and the total fault data number and the abnormal number of each parameter data in the SCADA are recorded in the following table:
calculating theoretical values according to actual values, and recording the theoretical values into a table:
Let H0: the running state of the wind power blade is independent from other parameters; calculating test statistics to measure the difference degree between an actual value and a theoretical value, and obtaining a chi-square value by a chi-square value calculation formula; searching a table according to the degree of freedom and the chi-square value, finding a corresponding P value, determining rejection or acceptance of the original assumption H0, and primarily screening SCADA parameters related to the running state of the blade; the parameter with the P value smaller than 0.05 is selected as the parameter related to the operation of the wind power blade, namely the original assumption is refused, and the operation state of the wind power blade is related to other parameters.
In a specific embodiment, the number of normal and abnormal SCADA parameters before and after the blade fracture is counted, and parameters related to the running state of the blade obtained through screening by the algorithm are shown in the following table, and after the original dozens of parameters are screened, namely, each parameter related to the running state of the wind power blade screened by using chi-square verification is respectively: the power, the rotation speed of the generator, the rotation speed of the rotor, the angle of the blades, the network side current 1, the network side current 2 and the network side current 3 are obtained, and the dimension reduction processing of data is realized.
Further, as shown in fig. 2, because the chi-square verifies that the relationship between wind speed and blade state is not obtained, but in order to classify the conditions of the SCADA data, a parameter of wind speed is added. And (3) performing data cleaning on the wind speed-power scatter diagram by adopting a DBSCAN clustering algorithm to remove null values, singular points, power limiting data and the like. As shown in FIG. 3, the Euclidean distance of the model established before data screening after blade fracture is 0.42, and as shown in FIG. 4, the Euclidean distance of the model established after data screening after blade fracture is 8.24, the effect is better and more remarkable than that before data processing, and the result shows that the accuracy of the model is improved by chi-square verification and DBSCAN data cleaning on SCADA data processing, and the problems of long model modeling time and low accuracy caused by SCADA data redundancy and noise are solved.
Further, as shown in fig. 2, the SCADA data of the preliminarily screened wind power blade in the normal running state is calculated according to 7:3, dividing the training set and the verification set into a NSET model; solving a process memory matrix Dn of each scheme, wherein n is a positive integer; the core of whether NSET modeling is successful is whether the construction of the process memory matrix is successful.
Representing the whole observation matrix of the wind turbine generator as M with the size of n multiplied by b n×b The M is n×b The expression of (2) is:
wherein n is a time state, and b is the number of observation variables in each time; matrix M n×b The row vector is X i ＝[x i (t 1 ) x i (t 2 )...x i (t b )]Matrix M n×b The row vector being a given observation parameter X i All observations within a certain observation period; matrix M n×b Is X (t) j )＝[x 1 (t j ) x 2 (t j )...x b (t j )] T Matrix M n×b Is t j Observing values of all observing parameters at any moment; from said M n×b Selecting a parameter of a period of time to be recorded as a historical observation matrix K, wherein the historical observation matrix K is the health state of each observation parameter, and the expression of the historical observation matrix K is as follows:
selecting a part of state data from the history observation matrix K, and forming a process memory matrix D by using the selected part of state data n . Process matrix D n Can be expressed as:
to observe matrix X obs And the observation matrix X obs Corresponding D n Input to the expressionIn (1) obtaining a prediction output matrix X est ；
Setting 7 training sets in SCADA data of the preliminarily screened wind power blade in a normal operation state as an observation matrix X obs The method comprises the steps of carrying out a first treatment on the surface of the Inputting 3 verification sets into a trained NSET model; the NSET model after inputting the verification set outputs a corresponding prediction output matrix X est The method comprises the steps of carrying out a first treatment on the surface of the According to the observation matrix X obs And predicting outputGo out matrix X est Fitting a corresponding Euclidean distance curve.
Further, the validity of the model is verified by using the Euclidean distance curve, and the Euclidean distance can represent the absolute value of the individual difference, for example, the two-point coordinate is (x i ，y i )、(x j ，y j ) Euclidean distanceThe model validity is verified by inputting fracture data (namely SCADA data before and after the wind power blade breaks for one minute, and the time stamp is 0.02s, and 3000 data are all) for the wind power blade. As shown in fig. 4, 3000 pieces of sample data before and after the blade fracture are input into a normal state model of the blade established by NSET, the blade fracture is at the 1500 th sample point, the euclidean distance curve is greatly increased after the data are input into the normal model, and the model can be used for diagnosing and monitoring the blade fracture.
Further, since the euclidean distance curve is subject to change at the breaking point of 1500 samples due to the limitation of chi-square validation, the breaking point is delayed to 1664 samples, and thus further screening is required. Continuing with residual E between observation matrix and prediction output matrix n ＝x n -x n * Analysis may determine the anomaly parameters. Wherein x is n The method comprises the steps of (1) observing an nth parameter in SCADA data before and after wind power blade fracture; x is x n * The predicted value corresponding to the nth parameter in the SCADA data before and after the wind power blade fracture; e (E) n The difference between the observed value of the nth parameter and the predicted value corresponding to the nth parameter; x is x 1 To x 7 Verifying each screened parameter related to the running state of the wind power blade for the using chi-square;
from the residual error E n Calculating the failure times e of the parameters related to the running state of the wind power blade screened by using chi-square verification c ；
Number of failures e c The expression of (2) is:
let a = observation vector group number/observation vector group number per input, and according to the number of failures e c Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i ；
The failure accumulation contribution rate ei is expressed as:
removing x 1 To x 7 The failure accumulation contribution rate e in (2) i The wind speed parameter is added into all parameters related to the running state of the wind power blade after removal, namely, the parameters with low fault contribution rate are removed, and the parameters with higher sensitivity to blade fracture are left as final modeling parameters so as to finally screen all the parameters related to the running state of the wind power blade; and finally screening out each parameter related to the running state of the wind power blade, wherein the parameters are respectively as follows: wind speed, power of a wind turbine generator, generator speed, rotor speed and wind turbine blade angle.
Blade breakage data (namely SCADA data before and after wind power blade breakage in one minute time, with a time stamp of 0.02s and 3000 data total) are input into the model, and the overall Euclidean distance result is shown in FIG. 5. As can be seen from comparison between fig. 4 and fig. 5, the euclidean distance curve of fig. 5 has larger fluctuation of euclidean distance before fracture, and changes at 1500 sample points, thereby being more in line with actual working conditions and being more accurate. The reliability analysis is utilized to screen the parameters in a basis, so that the reliability problem of modeling parameters is solved.
FIG. 6 shows a schematic flow chart of a process memory matrix construction procedure of one embodiment of the present invention. Selecting a part of state data from the history observation matrix K, and forming a process memory matrix D by using the selected part of state data n . The construction of the process memory matrix D requires that the k observation vectors X (1) X (2) X (k) within it be as large as possible to cover the deviceNormal working space.
The part of state data is selected from the historical observation matrix K, and the selected part of state data is used for forming a process memory matrix D n The method specifically comprises the following steps: setting each observation vector of the history observation matrix K to be composed of n variables; for each of the n variables, 0,1 will be ]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n In (a) and (b);
as shown in FIG. 6, for each of the n variables, the method will [0,1]]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n Specifically, the method comprises the following steps: setting i=1; wherein i is a positive integer; perform a=1/h i; wherein h is a positive integer; setting k=1; wherein k is a positive integer; judging whether the absolute value of the (X (k) -A is smaller than delta; wherein δ is a positive number; when |X (k) -A| is less than delta, X (k) is added to the process memory matrix D n In (a) and (b); when |x (k) -a| is equal to or greater than δ, judging whether k is larger than M; wherein M is the number of columns of the history observation matrix K; when k is less than or equal to M, executing k=k+1, and returning to the step of judging whether the |x (k) -a| is less than delta; when k is greater than M, judging whether i is greater than h; when i is less than or equal to h, executing i=i+1, and returning to the step of executing a=1/h×i; when i is greater than h, execution ends.
In a specific embodiment, δ is 0.001.
Each observation vector of the normal working space of the device consists of n variables, and its observations have been normalized. For each variable, equally dividing the space between 0 and 1 into h parts, searching a plurality of observation vectors from the set K by taking 1/h as step distance, and adding the observation vectors into a matrix D, wherein delta is a small positive number in the figure. For the remaining n-1 variables, the same flow as the illustration is adopted to select the observation vector from the set K and add the observation vector into the matrix D at the step distance of 1/h, and the method is adopted to construct a process memory matrix, so that the histories corresponding to different measured values of the n variables forming the observation vector can be selected into the matrix D, and the normal working space of the equipment can be better covered.
FIG. 7 shows a schematic flow chart of wind blade breakage real-time monitoring of an embodiment of the invention. As shown in fig. 7, the cleaned data corresponding to the finally screened parameters are processed according to the following formula 7:3, dividing the training set and the verification set into a training set and a verification set for carrying out NSET modeling for the second time, carrying out data fusion on the screened parameters by taking the training set and the verification set as an optimal model, inputting the verification set to obtain and record a maximum normal Euclidean distance threshold; the breaking data is input to obtain and record the Euclidean distance of the breaking of the blade. Inputting real-time SCADA parameters, comparing real-time data with Euclidean distance curves of the historical model, judging whether the real-time data exceeds a blade fracture early warning threshold value, and carrying out blade fracture grading early warning.
In this embodiment, as shown in fig. 8, the maximum normal euclidean distance threshold of the model is 0.13. As shown in FIG. 5, the Euclidean distance curve at the broken position of the blade is greatly increased, the Euclidean distance value after the blade is broken is 8.24, the damage of the blade is quantified, and the problem of quantification of damage of different degrees of the blade is solved.
FIG. 5 illustrates a blade early warning threshold split profile for one embodiment of the present invention. As shown in FIG. 5, data of several months before and after the blade fracture is used as input, the data is input into a NSET model under the normal working condition of the finally established blade, the fracture threshold is divided through the change of Euclidean distance, and the optimal early warning threshold is obtained by comparing Euclidean distance curves before the target monitoring wind power blade fracture because early warning thresholds obtained by different model units and blades with different lengths are different. Specifically, the euclidean distance between the euclidean distance 0 and the maximum value of the verification set is used as the normal variation range of the running state of the blade, the value between the maximum value of the verification set and the maximum value of the euclidean distance before the breaking of the blade is used as the variation range for checking the condition of the blade, the maximum value of the euclidean distance before the breaking of the blade is used as the emergency checking basis of the blade, the euclidean distance value after the breaking of the blade is used as the broken signal of the blade, and the early warning threshold values of different models possibly have differences. In a specific example, 1-3 months 12 days data were entered into the best model, as 3 months 12 days the leaves broke, placing the main analysis to the time before break. As shown in fig. 5, 8, 9, and 10 to 13, the maximum euclidean distance threshold value of the normal operation of the blade is 0.13, the maximum euclidean distance before the blade breaks is 2.21, and the euclidean distance after the blade breaks is 8.24. Therefore, the Euclidean distance 2.21 can be used as a blade fracture early warning threshold, when the Euclidean distance is between 0.13 and 2.21, the blade is checked at intervals, and when the Euclidean distance reaches 2.21, the machine is stopped in time for checking, so that the problem of setting the blade fracture early warning threshold is solved.
As shown in fig. 14, a computer apparatus 1400 includes: memory 1402, processor 1404, and a computer program stored on memory 1402 and executable on processor 1404, the processor 1404 implementing the steps of a method as in any of the embodiments described above when the computer program is executed by processor 1404.
When the processor 1404 executes the computer program, the correlation quantization preliminary screening is carried out on the blades and the sensor signals in the SCADA system through chi-square verification, the input parameters of the model are further verified and screened by utilizing fault data, finally, the screened SCADA parameters are established to be an NSET model in the normal state of the wind power blade as a semi-supervision model to monitor the running state of the blade, the known wind power blade fracture dataset is utilized to verify the modeling result and set an early warning threshold, and the threshold of the blade fracture early warning can be fully mined on the basis of the existing SCADA data, so that the purpose of online real-time and efficient monitoring is achieved.
The invention also proposes a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of the frequency closed loop control method for an oil beam pumping unit as in any of the embodiments described above.
When the computer program is executed by the processor, the correlation quantification preliminary screening is carried out on the blades and the sensor signals in the SCADA system through chi-square verification, the input parameters of the model are further verified and screened by utilizing fault data, finally, the screened SCADA parameters are established to be an NSET model in the normal state of the wind power blade as a semi-supervision model to monitor the running state of the blade, the known wind power blade fracture dataset is utilized to verify the modeling result and set an early warning threshold value, and the early warning threshold value of the blade fracture can be fully mined on the basis of the existing SCADA data, so that the purpose of online real-time and efficient monitoring is achieved.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (10)

1. A wind power blade fracture early warning method based on association analysis of SCADA data, wherein the SCADA data are data collected by a sensor group in an SCADA system of a wind turbine, and the SCADA data comprise the following parameters: the power of the wind turbine generator, the rotation speed of the rotor, the angle of wind power blades, the current of the grid side and the wind speed; the early warning method is characterized by comprising the following steps:
Acquiring a first set of SCADA data before and after the wind power blade breaks for a certain period of time;
screening the first set of SCADA data by using chi-square verification to screen out various parameters related to the running state of the wind power blade by using chi-square verification;
adding the wind speed parameter into each parameter which is screened out by using chi-square verification and is related to the running state of the wind power blade, and screening each parameter which is related to the running state of the wind power blade for the first time;
acquiring a second set of SCADA data of the wind power blade in a normal operation state for a certain period of time;
extracting original data corresponding to each parameter related to the running state of the wind power blade, wherein the original data are screened out for the first time from the second set of SCADA data;
fitting a wind speed-power scatter diagram according to the wind speed data in the second set of SCADA data and the power data of the wind turbine generator after data extraction;
performing data cleaning on the wind speed-power scatter diagram by adopting a DBSCAN clustering algorithm;
dividing the second group of SCADA data after data cleaning into a first training set and a first verification set;
building a first NSET model of the wind power blade in a normal running state according to the first training set and the first verification set;
Inputting an original first set of SCADA data into the first NSET model, and calculating a residual error E between a predicted value output by the first NSET model and a corresponding observed value n Fitting a first Euclidean distance curve corresponding to the first group of SCADA data according to the predicted value and the observed value so as to observe that the first Euclidean distance curve starts to be in an ascending trend after a certain position of the fracture point delay;
from the residual error E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i ；
Removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time;
the data corresponding to each parameter related to the running state of the wind power blade is screened out for the second time in the second set of SCADA data after data cleaning, the extracted data is divided into a second training set and a second verification set;
Establishing a second NSET model of the wind power blade in a normal running state according to the second training set and the second verification set;
acquiring a Euclidean distance curve corresponding to the second verification set, and taking the maximum value of the Euclidean distance on the Euclidean distance curve corresponding to the second verification set as a first threshold;
inputting the original first set of SCADA data into the second NSET model, and calculating a residual error E between a predicted value output by the second NSET model and a corresponding observed value n ' fitting a second Euclidean distance curve corresponding to the first set of SCADA data according to the predicted value and the observed value so as to observe that the position of the second Euclidean distance curve at the breaking point starts to be in an ascending trend;
acquiring a maximum value of Euclidean distance on a second Euclidean distance curve before the breaking point as a second threshold value;
acquiring the maximum value of the Euclidean distance on a second Euclidean distance curve after the breaking point as a third threshold;
SCADA data acquired by a sensor group in an SCADA system of the wind turbine generator is acquired in real time:
inputting SCADA data obtained in real time into the second NSET model, and calculating residual error E between predicted value output by the second NSET model and corresponding observed value n Fitting a third Euclidean distance curve corresponding to the SCADA data obtained in real time according to the predicted value and the observed value;
and carrying out fracture early warning on the wind power blades of the wind turbine generator set according to the values of the corresponding Euclidean distances on the third Euclidean distance curve, and the first threshold value, the second threshold value and the third threshold value.
2. The wind turbine blade breakage early warning method based on the correlation analysis of the SCADA data according to claim 1, wherein the method specifically comprises the steps of:
determining the range of the value of each corresponding Euclidean distance on the third Euclidean distance curve;
when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is more than or equal to 0 and less than or equal to the first threshold value, judging that the wind power blade of the wind turbine generator is in a normal running state;
when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than the first threshold value and smaller than or equal to the second threshold value, judging that the wind power blade of the wind power generation set is in an abnormal state;
And when the value of a certain Euclidean distance corresponding to the third Euclidean distance curve is larger than or equal to the third threshold value, judging that the wind power blade of the wind power generation set is in a broken state.
3. The wind power blade breakage early warning method based on the association analysis of SCADA data according to claim 1, wherein,
the parameters which are screened out by using the chi-square verification and are related to the running state of the wind power blade are respectively as follows: the power of the wind turbine generator, the rotation speed of the rotor, the wind power blade angle, the first grid side current, the second grid side current and the third grid side current;
the first screening of each parameter related to the running state of the wind power blade is as follows: wind speed, power of a wind turbine generator, generator rotation speed, rotor rotation speed, wind power blade angle, first grid side current, second grid side current and third grid side current;
said residual error E n The expression of (2) is:
E n ＝x n -x n *
wherein x is n Observations of an nth parameter in the first set of SCADA data; x is x n * A predicted value corresponding to an nth parameter in the first set of SCADA data; e (E) n The difference between the observed value of the nth parameter and the predicted value corresponding to the nth parameter; x is x 1 To x 7 Verifying each screened parameter related to the running state of the wind power blade for the using chi-square; and
said residual error E n Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i The method specifically comprises the following steps:
from the residual error E n Calculating the failure times e of the parameters related to the running state of the wind power blade screened by using chi-square verification c ；
The number of faults e c The expression of (2) is:
let a = observation vector group number/observation vector group number per input, and according to the number of failures e c Calculating the failure accumulation contribution rate e of each parameter which is screened out by using chi-square verification and related to the running state of the wind power blade i ；
The failure accumulation contribution rate e i The expression of (2) is:
removing the failure accumulation contribution rate e in each parameter related to the running state of the wind power blade screened by using chi-square verification i The method comprises the steps of adding a parameter smaller than a certain set value into each parameter related to the running state of the wind power blade after removal, and screening out each parameter related to the running state of the wind power blade for the second time, wherein the method specifically comprises the following steps:
Removing x 1 To x 7 The failure accumulation contribution rate e in (2) i The parameter which is smaller than a certain set value is added into each parameter which is related to the running state of the wind power blade after removal, and each parameter which is related to the running state of the wind power blade is screened out for the second time; wherein, each parameter related to the running state of the wind power blade is screened out for the second time, and the parameters are respectively as follows: wind speed, power of a wind turbine generator, generator speed, rotor speed and wind turbine blade angle.
4. A wind turbine blade breakage early warning method based on correlation analysis of SCADA data according to any one of claims 1 to 3, wherein the building a first NSET model of a wind turbine blade in a normal operation state according to the first training set and the first verification set specifically comprises:
acquiring an original NSET model, and inputting the first training set into the original NSET model to obtain a trained NSET model;
representing the whole observation matrix of the wind turbine generator as M with the size of n multiplied by b n×b The M is n×b The expression of (2) is:
wherein n is a time state, and b is the number of observation variables in each time; matrix M n×b The row vector is X i ＝[x i (t 1 ) x i (t 2 ) ... x i (t b )]Matrix M n×b The row vector being a given observation parameter X i All observations within a certain observation period; matrix M n×b Is X (t) j )＝[x 1 (t j ) x 2 (t j ) ... x b (t j )] T Matrix M n×b Is t j Observing values of all observing parameters at any moment;
from said M n×b Selecting a parameter of a period of time to be recorded as a historical observation matrix K, wherein the historical observation matrix K is the health state of each observation parameter, and the expression of the historical observation matrix K is as follows:
selecting a part of state data from the history observation matrix K, and forming a process memory matrix D by using the selected part of state data n . Process matrix D n Can be expressed as:
to observe matrix X obs And said observationMatrix X obs Corresponding D n Input to the expressionIn (1) obtaining a prediction output matrix X est ；
Setting the first validation set to an observation matrix X obs ；
Inputting the first validation set into a trained NSET model;
the NSET model after inputting the first verification set outputs a corresponding prediction output matrix X est ；
Calculating an observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Residual errors between the first verification set and the observation matrix X corresponding to the first verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the first verification set to establish a first NSET model of the wind power blade in a normal running state; and
According to the second training set and the second verification set, a second NSET model of the wind power blade in a normal running state is established, and the method specifically comprises the following steps:
acquiring an original NSET model, and inputting the second training set into the original NSET model to obtain a trained NSET model;
setting the second validation set to an observation matrix X obs ；
Inputting the second validation set into a trained NSET model;
outputting a corresponding prediction output matrix X by using the NSET model after the second verification set is input est ；
Calculating an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Residual errors between the two, and an observation matrix X corresponding to the second verification set obs And corresponding prediction output matrix X est Fitting the Euclidean distance curve corresponding to the second verification set to establish a second NSET model of the wind power blade in a normal running state.
5. The method for early warning of wind turbine blade breakage based on correlation analysis of SCADA data according to claim 4, wherein a part of the state data is selected from the history observation matrix K, and the process memory matrix D is constructed using the selected part of the state data n The method specifically comprises the following steps:
setting each observation vector of the history observation matrix K to be composed of n variables;
For each of the n variables, 0,1 will be]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n In (a) and (b);
the pair of each of the n variables will be [0,1 ]]Equally dividing the historical observation matrix K into h parts, searching a plurality of observation vectors X (1) X (2) according to 1/h as step distance, and adding the observation vectors X (K) into the process memory matrix D n Specifically, the method comprises the following steps:
setting i=1; wherein i is a positive integer;
perform a=1/h i; wherein h is a positive integer;
setting k=1; wherein k is a positive integer;
judging whether the absolute value of the (X (k) -A is smaller than delta; wherein δ is a positive number;
adding X (k) to the process memory matrix Dn when |X (k) -A| is less than δ;
when |x (k) -a| is equal to or greater than δ, judging whether k is larger than M; wherein M is the number of columns of the history observation matrix K;
when k is less than or equal to M, executing k=k+1, and returning to the step of judging whether the |x (k) -a| is less than delta;
when k is greater than M, judging whether i is greater than h;
when i is less than or equal to h, executing i=i+1, and returning to the step of executing a=1/h×i;
when i is greater than h, execution ends.
6. The wind power blade breakage early warning method based on the correlation analysis of the SCADA data according to claim 1, wherein the screening of the first set of SCADA data by using chi-square verification to screen out each parameter related to the running state of the wind power blade by using chi-square verification specifically comprises:
establishing an original assumption H0, wherein the original assumption H0 is independent between the running state of the wind power blade and each parameter in the first set of SCADA data;
taking the data of the running state of the wind power blade as a first variable threshold value;
taking one parameter in the first set of SCADA data as a second variable threshold at each time;
respectively recording the actual value of the data number of the first variable threshold under the condition that the wind power blade is normal as a, the actual value of the data number of the first variable threshold under the condition that the wind power blade is faulty as b and the actual value of the total data number of the wind power blade as a+b;
the actual value of the data number of the y-th parameter in the SCADA data under the normal condition of the wind power blade is recorded as c y The actual value of the data number under the fault of the wind power blade is d y And the actual value of the total data number of the wind power blade is c y +d y Wherein y is a positive integer;
Recording the actual value of the total data number of the first variable threshold value and the second variable threshold value under the normal condition of the wind power blade as a+c respectively y And the actual value of the total number of data in the wind power blade fault is b+d y And the actual value of the total data number of the wind power blades of the two is a+b+c y +d y ；
Respectively calculating to obtain theoretical value of the number of data of the first variable threshold under the condition that the wind power blade is normal as (a+b) × (a+c) y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (a+b) x (b+d) y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is a+b;
respectively calculating to obtain the theoretical value of the number of the data of the second variable threshold under the normal condition of the wind power blade as (c) y +d y )×(a+c y )/(a+b+c y +d y ) The theoretical value of the number of data under the fault of the wind power blade is (c y +d y )×(b+d y )/(a+b+c y +d y ) And the theoretical value of the total data number of the wind power blade is c y +d y ；
Setting the degree of freedom to be 1;
calculating chi-square value according to chi-square value calculation formula, wherein the chi-square value is used for measuring the difference degree between each actual value and each theoretical value,
the chi-square value calculation formula is:
χ 2 ＝∑(A-T) 2 /T
wherein χ is 2 The value is chi square, A is the actual value, and T is the theoretical value;
finding a corresponding P value according to the degree of freedom and the chi-square value table, wherein the P value is the probability of making a first class of false rejection errors;
And selecting one parameter in the corresponding first set of SCADA data when the P value is smaller than 0.05 as a parameter related to the running state of the wind power blade, so as to sequentially screen out each parameter related to the running state of the wind power blade.
7. The wind power blade fracture early warning method based on the association analysis of SCADA data according to claim 1, wherein the data cleaning of the wind speed-power scatter diagram by adopting a DBSCAN clustering algorithm specifically comprises the following steps:
taking all points in the wind speed-power scatter plot as sample data S, and marking each data point in the sample data S as an unprocessed state;
assigning initial values to Eps and Minpts, wherein Eps is a neighborhood distance threshold value of a certain data point p in the sample data S, and Minpts is the minimum number of data points in the neighborhood with the radius of Eps of the certain data point p in the sample data S;
setting the neighborhood with the radius of Eps of a certain data point p as N Eps (p)；
Clustering the high-density areas formed by the Eps and the Minpts;
the clustering of the high-density region formed by the Eps and the Minpts specifically comprises the following steps:
determining whether the certain data point p has been added to a certain cluster or has been listed as noise;
If the certain data point p has been added to a certain cluster or has been listed as noise, the classification ends;
if the data point p is not added to a cluster and is not listed as noise, N is determined Eps (p) whether there are at least Minpts objects within it;
if N Eps At least Minpts objects exist in (p), a new class cluster U is constructed, and a certain data point p is added in the U;
if N Eps And (4) if the number of objects in the data point (p) is less than Minpts, the data point (p) is listed as a boundary point or noise.
8. The method for early warning of wind turbine blade breakage based on correlation analysis of SCADA data of claim 7, wherein said Eps is set to 4.5 and said mints is set to 18.5.
9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the computer program, carries out the steps of a wind power blade breakage warning method based on a correlation analysis of SCADA data as claimed in any one of claims 1 to 8.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of a wind blade breakage pre-warning method based on a correlation analysis of SCADA data as claimed in any one of claims 1 to 8.
CN202311540515.6A 2023-11-17 2023-11-17 Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium Pending CN117851877A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311540515.6A CN117851877A (en) 2023-11-17 2023-11-17 Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311540515.6A CN117851877A (en) 2023-11-17 2023-11-17 Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium

Publications (1)

Publication Number Publication Date
CN117851877A true CN117851877A (en) 2024-04-09

Family

ID=90527923

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311540515.6A Pending CN117851877A (en) 2023-11-17 2023-11-17 Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium

Country Status (1)

Country Link
CN (1) CN117851877A (en)

Similar Documents

Publication Publication Date Title
CN110414155B (en) Fan component temperature abnormity detection and alarm method with single measuring point
CN109524139B (en) Real-time equipment performance monitoring method based on equipment working condition change
CN111539553B (en) Wind turbine generator fault early warning method based on SVR algorithm and off-peak degree
CN108627720B (en) Power equipment state monitoring method based on Bayesian algorithm
CN110414154B (en) Fan component temperature abnormity detection and alarm method with double measuring points
CN112746934B (en) Method for diagnosing fan fault through self-association neural network
CN105930629B (en) A kind of on-line fault diagnosis method based on magnanimity service data
CN112487910A (en) Fault early warning method and system for nuclear turbine system
CN105550943A (en) Method for identifying abnormity of state parameters of wind turbine generator based on fuzzy comprehensive evaluation
CN111581597A (en) Wind turbine generator gearbox bearing temperature state monitoring method based on self-organizing kernel regression model
CN109492790A (en) Wind turbines health control method based on neural network and data mining
CN113738595B (en) Wind generating set blade state monitoring method and system
CN113562562A (en) Elevator safety early warning monitoring system and judgment method thereof
CN111103137A (en) Wind turbine gearbox fault diagnosis method based on deep neural network
CN114061743A (en) Vibration monitoring method, device, equipment and medium for wind generating set
CN114004059B (en) Health portrait method for hydroelectric generating set
CN114412726A (en) Wind turbine generator system variable pitch bearing detection method and device and wind turbine generator system
CN117993562A (en) Wind turbine generator system fault prediction method and system based on artificial intelligent big data analysis
CN117670086A (en) Compressor fault prediction method and system based on big data and machine learning
CN111934903A (en) Docker container fault intelligent prediction method based on time sequence evolution genes
CN117851877A (en) Wind power blade fracture early warning method and device based on SCADA data association analysis and readable storage medium
CN117471308A (en) Fan fault diagnosis method based on big data multi-feature technology mining
Tingting et al. Early warning method for power station auxiliary failure considering large-scale operating conditions
CN112508350A (en) Vibration early warning method and system for steam turbine generator unit
CN118096117B (en) Digital twinning-based power equipment fault detection method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination