CN111880090A - Distribution layered online fault detection method for million-kilowatt ultra-supercritical unit - Google Patents

Distribution layered online fault detection method for million-kilowatt ultra-supercritical unit Download PDF

Info

Publication number
CN111880090A
CN111880090A CN201910579518.8A CN201910579518A CN111880090A CN 111880090 A CN111880090 A CN 111880090A CN 201910579518 A CN201910579518 A CN 201910579518A CN 111880090 A CN111880090 A CN 111880090A
Authority
CN
China
Prior art keywords
variable
block
sub
variables
matrix
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.)
Granted
Application number
CN201910579518.8A
Other languages
Chinese (zh)
Other versions
CN111880090B (en
Inventor
赵春晖
张淑美
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.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
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 Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201910579518.8A priority Critical patent/CN111880090B/en
Publication of CN111880090A publication Critical patent/CN111880090A/en
Application granted granted Critical
Publication of CN111880090B publication Critical patent/CN111880090B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
    • G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
    • G01R31/34—Testing dynamo-electric machines
    • G01R31/343—Testing dynamo-electric machines in operation
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M99/00—Subject matter not provided for in other groups of this subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Probability & Statistics with Applications (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Testing Or Calibration Of Command Recording Devices (AREA)

Abstract

The invention discloses a distribution layering type online fault detection method for a million kilowatt ultra-supercritical unit. Aiming at the problems of numerous process variables and complex changing conditions of a million-kilowatt ultra-supercritical unit, comprehensively considering the correlation among the variables and the distribution condition of the variables in the sample direction, blocking the variables by using a multi-layer information theory decomposition method, and establishing a multi-layer distributed monitoring algorithm facing the million-kilowatt ultra-supercritical unit based on a blocking result and combining a Gaussian mixture model method and a Bayesian theory. The method fully explores relevant information among process variables, is beneficial to understanding the complex process characteristics of the million kilowatt ultra-supercritical unit, can effectively excavate local information of the process, can analyze the relevant relation among different variable subblocks, and greatly improves the fault detection performance of the million kilowatt ultra-supercritical unit in the complex process, thereby ensuring the safe and reliable operation of a large coal-fired generator set.

Description

Distribution layered online fault detection method for million-kilowatt ultra-supercritical unit
Technical Field
The invention belongs to the field of thermal power process fault detection, and particularly relates to a distributed layered online process monitoring method for a million-kilowatt ultra-supercritical unit with a plurality of heavy-faced variables and frequent fluctuation of working conditions.
Background
The power industry is an important basic industry of national economy and is a key project in the national economic development strategy. With the rapid development of economy, the demand for electricity is also rapidly increasing. Coal resources are the main energy in China, so that the energy structure mainly based on coal is difficult to change fundamentally in a long period in the future. As a main power source in China, the installed capacity of coal-fired power generation is always over 70 percent. According to statistics, under the promotion of vigorous electricity demand, the total electricity consumption of the whole society in 1-8 months in 2018 is up to 45296 hundred million kilowatt hours, and the electricity consumption is increased by 9.0 percent on a same scale. Wherein, the accumulated power generation amount of the thermal power is 33103 hundred million kilowatts, which accounts for 73.1 percent of the total power generation amount of the whole country and is increased by 7.2 percent on a par. In recent years, in order to realize sustainable development of electric power, structural adjustment is actively carried out in the thermal power generation industry, the upper large pressure is small, a high-energy-consumption small thermal power generating unit is replaced by a high-capacity and low-energy-consumption supercritical (supercritical) unit, and an electric power energy structure mainly comprising a large coal-fired generating unit such as a million-kilowatt supercritical unit is basically formed. Therefore, the method has great practical significance and application value for the analysis and research of the million-kilowatt ultra-supercritical unit.
Compared with the traditional generator set, the million kilowatt ultra-supercritical generator set has the advantages of large scale, various devices, numerous parameters and mutual influence, long industrial process, multiple unit devices, wide spatial distribution and high safety requirement in the whole power generation process, and brings difficulty to state monitoring and fault diagnosis of the million kilowatt ultra-supercritical generator set. In addition, due to different reasons such as environmental conditions, fuel characteristics and load, the million kilowatt ultra-supercritical unit can operate under different working conditions. Especially, in recent years, due to the fact that new energy such as wind power, photoelectricity and the like are connected to the power grid, the load fluctuation of the power grid, the peak-valley difference is increased, and the requirement of a user side changes, the unit is in a full-working-condition operation mode with different working conditions switched frequently due to new normality such as frequent deep peak regulation and the like. And the large coal-fired power generation process has complex environment and long industrial process, and a plurality of variables still present different data distribution characteristics even under the same working condition. These all present significant challenges to fault detection and diagnosis for large coal-fired power generating units.
For the problem of fault detection of the thermal generator set, the predecessors have made corresponding research and discussion from different angles, and a corresponding online process monitoring method is provided. However, most of the existing methods are mainly centralized single-working-condition monitoring methods. The centralized single-working-condition monitoring method cannot obtain a good monitoring effect in the face of the characteristics of long flow, numerous variables, complex correlation and dynamic working conditions of the million-kilowatt ultra-supercritical unit. The invention further considers the complex correlation among a plurality of variables of the million kilowatt ultra-supercritical unit and the multi-distribution condition of the variables in the sample direction, and provides a novel multi-layer distributed online fault detection method for the million kilowatt ultra-supercritical unit.
Disclosure of Invention
The invention aims to provide a multi-layer distributed monitoring algorithm for million-kilowatt ultra-supercritical units, aiming at the problem that the existing fault detection method for the million-kilowatt ultra-supercritical units cannot accurately describe local information. The method comprehensively considers the correlation among the variables and the distribution condition of the variables in the sample direction, blocks the variables by using a multi-layer information theory decomposition method, fully explores the process information among the process variables, and is beneficial to understanding the complex process characteristics of the million kilowatt ultra-supercritical unit. The multilayer distributed monitoring method can effectively mine local information of the process, can analyze the correlation among different variable subblocks, and greatly improves the fault detection performance of the million-kilowatt ultra-supercritical unit in the complex process, thereby ensuring the safe and reliable operation of the large coal-fired generator set.
The purpose of the invention is realized by the following technical scheme: a distribution layered online fault detection method for a million kilowatt ultra-supercritical unit comprises the following steps:
(1) acquiring normal data to be analyzed: a million-kilowatt ultra-supercritical unit is provided with J measurement variables and operation variables, a vector of 1 XJ can be obtained by sampling every time, and data acquired after sampling N times is expressed as a two-dimensional matrix X ═ X1,X2,...,XJ]∈RN×JThe measured variables are state parameters which can be measured in the normal operation process of the unit, and comprise flow, voltage, current, temperature, speed and the like; the operation variables comprise air intake, feeding amount, valve opening and the like;
(2) the process variable is divided into different sub-blocks by using a spectral clustering method based on mutual information, the variable in the same sub-block has stronger correlation, and the correlation between different sub-blocks is weaker. This step is realized by the following substeps:
(2.1) obtaining mutual information among variables:
I(Xi,Xj)＝H(Xi)+H(Xj)-H(Xi,Xj) (1)
wherein, Xi(i ═ 1, 2.., J) denotes the ith variable, H (X)i) Is a variable XiThe information entropy of (2):
H(Xi)＝-∫xp(Xi)logp(Xi)dx (2)
H(Xi,Xj) Is a variable XiAnd XjJoint information entropy of (a):
Figure BDA0002112808330000031
p(Xi) And p (X)j) Represents variable XiAnd XjP (X) is a probability density function ofi,Xj) Is a joint probability density function.
(2.2) solving the generalized correlation coefficient between every two variables based on the mutual information solved by the formula (1):
Figure BDA0002112808330000032
wherein r isij∈[0,1]。
(2.3) based on equation (4), a correlation matrix of the variables is found:
Figure BDA0002112808330000033
(2.4) solving a diagonal matrix D based on the correlation matrix R defined by the formula (5):
D＝diag{Dii} (6)
wherein D isiiIs the sum of all elements in row i in equation (5):
Figure BDA0002112808330000034
(2.5) solving Laplace matrix of diagonal matrix D
L＝D-1/2RD-1/2(8)
(2.6) spectral decomposition of Laplace matrix
L＝PΛPT(9)
Wherein, P ═ P1,P2,...,PJ]Are orthogonal eigenvectors.
(2.7) selecting the eigenvectors corresponding to the k maximum eigenvalues to form a matrix E ═ P1,P2,...,Pk]∈RJ×kNormalizing each row in the matrix E to obtain a matrix Y
Figure BDA0002112808330000041
(2.8) clustering Y by using a kmeans clustering algorithm, and if the ith row belongs to the b-th class, carrying out variable XiDivision into sub-block b Xb. Thus, a plurality of operation variables of the million-kilowatt ultra-supercritical unit are divided into B variable blocks according to the degree of correlation.
X＝[X1X2… XB](11)
Wherein,
Figure BDA0002112808330000042
is the (B ═ 1, 2.., B) th variable block, JbRepresents XbThe number of variables contained in (1).
(3) Further decomposing the variables in the variable block according to the distribution situation in the sample direction by using an information theory decomposition method based on a Gaussian mixture model, wherein the step is realized by the following sub-steps:
(3.1) randomly dividing the variable block into WbIndividual block:
Figure BDA0002112808330000043
(3.2) using gaussian mixture model method to obtain W (W1, 2.., W)b) Probability density of individual variable subblocks:
Figure BDA0002112808330000044
wherein,
Figure BDA0002112808330000045
is the number of sub-gaussian components;
Figure BDA0002112808330000046
is the prior probability of the mth sub-Gaussian component, satisfies
Figure BDA0002112808330000047
And
Figure BDA0002112808330000048
is a mean value containing sub-Gaussian components
Figure BDA0002112808330000049
Sum covariance matrix
Figure BDA00021128083300000410
Of the parameter set (c).
Figure BDA00021128083300000411
Is a multivariate gaussian probability density:
Figure BDA00021128083300000412
wherein, Jb,wIs composed of
Figure BDA0002112808330000051
The number of medium variables.
(3.3) solving probability density distribution functions of all variables in the sub-blocks:
Figure BDA0002112808330000052
wherein the variable
Figure BDA0002112808330000053
Jb,wIs a variable block
Figure BDA0002112808330000054
The number of the medium variables is equal to or greater than the total number of the medium variables,
Figure BDA0002112808330000055
represents Xb,iBelonging to sub-blocks
Figure BDA0002112808330000056
When Xb,iThe conditional probability density of (2).
(3.4) solving variable subblocks
Figure BDA0002112808330000057
And
Figure BDA0002112808330000058
KL divergence of (W, v ∈ [1, W ]b])：
Figure BDA0002112808330000059
Wherein,
Figure BDA00021128083300000510
and
Figure BDA00021128083300000511
are respectively
Figure BDA00021128083300000512
And
Figure BDA00021128083300000513
the probability density function of (2) can be calculated by using the formula (13).
(3.5) optimizing the random partitions of step (3.1) using an ant colony algorithm such that the following objective function is maximized:
Figure BDA00021128083300000514
(3.6) each variable block (B ═ 1, 2.., B) is further divided into sub-blocks by repeating steps (3.2) - (3.5). The original data set X is divided into different sub-blocks:
Figure BDA00021128083300000515
wherein, the variable block
Figure BDA00021128083300000516
All variables in the system have strong correlation, and variable sub-blocks
Figure BDA00021128083300000517
The variables in (a) have both strong correlations and similar data distributions.
(4) Based on the variable block result obtained in the steps (2) and (3), firstly, describing the variable sub-block by using a Principal Component Analysis (PCA)
Figure BDA00021128083300000518
Correlation of each variable in the
Figure BDA00021128083300000519
Wherein, Pb,wIs a load matrix, Tb,wIs a principal component matrix.
(5) Principal component matrix T established by Gaussian Mixture Model (GMM) methodb,wThe distribution of (c):
Figure BDA0002112808330000061
wherein,
Figure BDA0002112808330000062
is the number of gaussian components;
Figure BDA0002112808330000063
the weight of the m-th component is represented,
Figure BDA0002112808330000064
is a mean value containing sub-Gaussian components
Figure BDA0002112808330000065
Sum covariance matrix
Figure BDA0002112808330000066
Of the parameter set (c).
(6) For each variable sub-block
Figure BDA0002112808330000067
Establishing BIP statistics
Figure BDA0002112808330000068
Wherein,
Figure BDA0002112808330000069
to represent
Figure BDA00021128083300000610
Belonging to the m-th component
Figure BDA00021128083300000611
The probability of (a) of (b) being,
Figure BDA00021128083300000612
as a principal component matrix Tb,wAn nth (N ═ 1, 2.., N) row vector.
Figure BDA00021128083300000613
Is based on the probability of local Mahalanobis distance, which is defined as
Figure BDA00021128083300000614
Wherein,
Figure BDA00021128083300000615
is composed of
Figure BDA00021128083300000616
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(7) The relationship between each sub-block in each variable block is monitored by a Gaussian Mixture Model (GMM) method, which is implemented by the following sub-steps.
(7.1) blocking each variable block XbThe first columns of the pivot matrices of the respective sub-blocks are combined together:
Figure BDA00021128083300000617
wherein, tb,w(w＝1,2,...,Wb) As a principal component matrix Tb,wThe 1 st column vector.
(7.2) describing principal metadata with GMM
Figure BDA00021128083300000618
The distribution of (c):
Figure BDA00021128083300000619
wherein,
Figure BDA00021128083300000620
the number of Gaussian components in the b variable sub-block is shown;
Figure BDA00021128083300000621
the weight of the m-th component is represented,
Figure BDA00021128083300000622
is a mean value containing sub-Gaussian components
Figure BDA00021128083300000623
Sum covariance matrix
Figure BDA00021128083300000624
Of the parameter set (c).
(7.3) Main metadata for each variable Block
Figure BDA00021128083300000625
Establishing BIP statistic:
Figure BDA0002112808330000071
wherein,
Figure BDA0002112808330000072
as a principal component matrix
Figure BDA0002112808330000073
An nth (N ═ 1, 2.., N) row vector.
Figure BDA0002112808330000074
Is based on the probability of the local mahalanobis distance.
Figure BDA0002112808330000075
Is based on the probability of the local mahalanobis distance,
Figure BDA0002112808330000076
is composed of
Figure BDA0002112808330000077
Mahalanobis distance to the mth gaussian component,
Figure BDA0002112808330000078
is composed of
Figure BDA0002112808330000079
Any one of the rows.
(8) During online fault detection, the process is monitored from three levels of variable subblocks, variable blocks and the whole unit. This step is realized by the following substeps.
(8.1) acquiring new data: and (4) collecting the variable values of the measuring points according to the step (1) and recording as z (1 multiplied by J).
(8.2) according to the variable blocking results obtained in the step (2) and the step (3), carrying out sub-block decomposition on the new data:
z＝[z1z2… zb… zB](26)
Figure BDA00021128083300000710
wherein z isb(B ═ 1, 2.., B) is the B variable sub-block.
(8.3) at the bottom layer, i.e., variable sub-block layer, each sub-block
Figure BDA00021128083300000711
Figure BDA00021128083300000712
The data of (2) are projected to the principal element direction of the corresponding sub-block:
Figure BDA00021128083300000713
wherein
Figure BDA00021128083300000714
Is a load matrix.
(8.4) obtaining each sub-block
Figure BDA00021128083300000715
The online statistic index of (1):
Figure BDA00021128083300000716
wherein, the meaning of each parameter in the above formula is similar to that in the formula (22).
Figure BDA00021128083300000717
To represent
Figure BDA00021128083300000718
Belonging to the m-th component
Figure BDA00021128083300000719
The probability of (c).
Figure BDA00021128083300000720
Is based on the probability of the local mahalanobis distance,
Figure BDA00021128083300000721
is composed of
Figure BDA00021128083300000722
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(8.5) at the variable block level, z is first putbThe main elements of each variable sub-block are combined together:
Figure BDA0002112808330000081
(8.6) obtaining each variable block zbThe online statistic index of (1):
Figure BDA0002112808330000082
wherein, the meaning of each parameter of the above formula is similar to that in the formula (25).
Figure BDA0002112808330000083
To represent
Figure BDA0002112808330000084
Belonging to the m-th component
Figure BDA0002112808330000085
The probability of (c).
Figure BDA0002112808330000086
Is based on the probability of the local mahalanobis distance,
Figure BDA0002112808330000087
is composed of
Figure BDA0002112808330000088
Mahalanobis distance to the mth gaussian component,
Figure BDA0002112808330000089
is composed of
Figure BDA00021128083300000810
Any one of the rows.
(8.7) in order to analyze the relation between different variable subblocks, the operation condition of the million kilowatt ultra-supercritical unit is monitored from the whole unit level, and the BIP index of each variable block is firstly converted into the probability of normal (marked as 'N') and fault (marked as 'F'):
Figure BDA00021128083300000811
Figure BDA00021128083300000812
wherein BIPb,lmtA control limit for the statistical BIP indicator;
Figure BDA00021128083300000813
representing the normal conditional probability of the b variable block;
Figure BDA00021128083300000814
indicating the conditional probability of the failure of the b-th variable block.
(8.8) calculating the posterior probability of the b variable block failing by the Bayes rule
Figure BDA00021128083300000815
Wherein, Pb(F)＝α；Pb(N) ═ 1-alpha, respectively, for significanceThe level is the prior probability of a process failing or being normal at alpha.
(8.9) comprehensively considering the fault probability of all variable blocks and calculating the global monitoring statistic
Figure BDA00021128083300000816
(9) Judging the running state of the process: and analyzing the process state from three levels of the variable subblocks, the variable blocks and the whole unit. Three levels of statistics are compared with the control limits in real time:
(a) at each variable sub-block
Figure BDA0002112808330000091
In, if BIPb,w1-alpha, in sub-blocks
Figure BDA0002112808330000092
The variable in (b) fails, otherwise the variable in the sub-block is considered to be operating within the normal range.
(b) At the variable block level, if BIPbIf the value is more than 1-alpha, the related relation of each variable sub-block in the variable block is abnormal, otherwise, all the variables in the b-th sub-block are normally operated.
(c) At the unit level, if PFzIf the measured value is more than alpha, the abnormal condition or the fault occurs in the running process of the million kilowatt ultra-supercritical unit, otherwise, the whole unit runs normally.
Compared with the prior art, the invention has the beneficial effects that: the invention aims to provide a multi-layer distributed monitoring algorithm for million-kilowatt ultra-supercritical units, aiming at the characteristics that million-kilowatt ultra-supercritical units are large in scale, various in equipment, numerous in parameters and mutually influenced, and the whole power generation process is long in industrial process, multiple in unit devices, wide in spatial distribution and frequent in working condition switching. The method comprehensively considers the correlation among the variables and the distribution condition of the variables in the sample direction, blocks the variables by using a multi-layer information theory decomposition method, fully explores the process information among the process variables, and is beneficial to understanding the complex process characteristics of the million kilowatt ultra-supercritical unit. The multilayer distributed monitoring method can effectively mine local information of the process, can analyze the correlation among different variable subblocks, and greatly improves the fault detection performance of the million-kilowatt ultra-supercritical unit in the complex process, thereby ensuring the safe and reliable operation of the large coal-fired generator set.
Description of the drawings:
FIG. 1 is an explanatory diagram of a distributed hierarchical online fault detection method for a million kilowatt ultra-supercritical unit according to the invention;
fig. 2 shows the monitoring results of the variable sub-blocks in the specific embodiment of the method of the present invention, (a) shows the monitoring results of two variable sub-blocks in the 5 th variable block, (b) shows the monitoring results of three variable sub-blocks in the 7 th variable block, and (c) shows the monitoring results of three variable sub-blocks in the 9 th variable block.
Fig. 3 shows the monitoring results in the variable layer of the method of the present invention in a specific embodiment, (a) is the monitoring result in the 5 th variable block, (b) is the monitoring result in the 7 th variable block, and (c) is the monitoring result in the 9 th variable block.
Fig. 4 shows the unit level monitoring result of the method according to the present invention in an embodiment.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific examples.
The invention takes the unit of Zhe energy group subordinate Jiahua power plant No. 3 as an example, the unit is a million kilowatt ultra-supercritical unit, the power of the unit is 600MW, the unit totally comprises 154 process variables, and the variables relate to pressure, temperature, flow rate and the like.
As shown in figure 1, the invention discloses an online monitoring method for the dynamic and static characteristic collaborative analysis of a million kilowatt ultra-supercritical unit, which comprises the following steps:
(1) acquiring normal data to be analyzed: setting a million kilowatt ultra-supercritical unit to have J measurement variables and operation variables, obtaining a 1 XJ vector by sampling every time, and expressing data obtained after sampling N times into twoDimension matrix X ═ X1,X2,...,XJ]∈RN×J. In this example, the sampling period is 1 minute, 2940 sample data in the normal operation process of the thermal power generating unit are collected for variable blocking and offline modeling, and 154 process variables, namely, the modeling data is X (2940 × 154). The measurement variables are state parameters which can be measured in the normal operation process of the unit, and comprise flow, voltage, current, temperature, speed and the like; the operation variables comprise air intake, feeding amount, valve opening and the like;
(2) the process variable is divided into different sub-blocks by using a spectral clustering method based on mutual information, the variable in the same sub-block has stronger correlation, and the correlation between different sub-blocks is weaker. This step is realized by the following substeps:
(2.1) obtaining mutual information among variables:
I(Xi,Xj)＝H(Xi)+H(Xj)-H(Xi,Xj) (1)
wherein, Xi(i ═ 1, 2.., J) denotes the ith variable, H (X)i) Is a variable XiThe information entropy of (2):
H(Xi)＝-∫xp(Xi)logp(Xi)dx (2)
H(Xi,Xj) Is a variable XiAnd XjJoint information entropy of (a):
Figure BDA0002112808330000101
p(Xi) And p (X)j) Represents variable XiAnd XjP (X) is a probability density function ofi,Xj) Is a joint probability density function.
(2.2) solving the generalized correlation coefficient between every two variables based on the mutual information solved by the formula (1):
Figure BDA0002112808330000111
wherein r isij∈[0,1]。
(2.3) based on the formula (4), obtaining a correlation matrix of the variables:
Figure BDA0002112808330000112
(2.4) solving a diagonal matrix D based on the correlation matrix R defined by the formula (5):
D＝diag{Dii} (6)
wherein D isiiIs the sum of all elements in the ith row in formula (5)
Figure BDA0002112808330000113
(2.5) solving Laplace matrix of diagonal matrix D
L＝D-1/2RD-1/2(8)
(2.6) spectral decomposition of Laplace matrix
L＝PΛPT(9)
Wherein, P ═ P1,P2,...,PJ]Are orthogonal eigenvectors.
(2.7) selecting the eigenvectors corresponding to the k maximum eigenvalues to form a matrix E ═ P1,P2,...,Pk]∈RJ×kNormalizing each row in the matrix E to obtain a matrix Y
Figure BDA0002112808330000114
(2.8) clustering Y by using a kmeans clustering algorithm, and if the ith row belongs to the b-th class, carrying out variable XiDivision into sub-block b Xb. Thus, a plurality of operation variables of the million-kilowatt ultra-supercritical unit are divided into B variable blocks according to the degree of correlation.
X＝[X1X2... XB](11)
Wherein,
Figure BDA0002112808330000121
is the (B ═ 1, 2.., B) th variable block, JbRepresents XbThe number of variables contained in (1).
In this example, 154 process variables are divided into 11 sub-blocks according to the correlation, as shown in table 1, the variables in each sub-block have a strong correlation, and the correlation between different sub-blocks is weak.
TABLE 1 variable blocking in megawatt ultra supercritical units
Figure BDA0002112808330000122
(3) Further decomposing the variables in the 11 variable blocks according to the distribution situation in the sample direction by using an information theory decomposition method based on a Gaussian mixture model, wherein the step is realized by the following sub-steps
(3.1) randomly dividing the variable block into WbIndividual block:
Figure BDA0002112808330000123
(3.2) using gaussian mixture model method to obtain W (W1, 2.., W)b) Probability density of individual variable subblocks:
Figure BDA0002112808330000124
wherein,
Figure BDA0002112808330000131
is the number of sub-gaussian components;
Figure BDA0002112808330000132
is the prior probability of the mth sub-Gaussian component, satisfies
Figure BDA0002112808330000133
And
Figure BDA0002112808330000134
is composed of sub-Gaussian componentsMean value of
Figure BDA0002112808330000135
Sum covariance matrix
Figure BDA0002112808330000136
Of the parameter set (c).
Figure BDA0002112808330000137
Is a multivariate gaussian probability density:
Figure BDA0002112808330000138
wherein, Jb,wIs composed of
Figure BDA0002112808330000139
The number of medium variables.
(3.3) solving probability density distribution functions of all variables in the sub-blocks:
Figure BDA00021128083300001310
wherein the variable
Figure BDA00021128083300001311
Jb,wIs a variable block
Figure BDA00021128083300001312
The number of the medium variables is equal to or greater than the total number of the medium variables,
Figure BDA00021128083300001313
represents Xb,iBelonging to sub-blocks
Figure BDA00021128083300001314
When XbiThe conditional probability density of (2).
(3.4) solving variable subblocks
Figure BDA00021128083300001315
And
Figure BDA00021128083300001316
KL divergence of (W, v ∈ [1, W ]b])：
Figure BDA00021128083300001317
Wherein,
Figure BDA00021128083300001318
and
Figure BDA00021128083300001319
are respectively
Figure BDA00021128083300001320
And
Figure BDA00021128083300001321
the probability density function of (2) can be calculated by using the formula (13).
(3.5) optimizing the random partitions of step (3.1) using an ant colony algorithm such that the following objective function is maximized:
Figure BDA00021128083300001322
(3.6) each variable block (B ═ 1, 2.., B) is further divided into sub-blocks by repeating steps (3.2) - (3.5). The original data set X is divided into different sub-blocks:
Figure BDA00021128083300001323
wherein, the variable block
Figure BDA00021128083300001324
All variables in the system have strong correlation, and variable sub-blocks
Figure BDA00021128083300001325
The variables in (a) have both strong correlations and similar data distributions.
In this example, the 11 variable blocks obtained in step (2) are further divided into 27 variable sub-blocks according to the distribution of the variables, and the variables in each sub-block have both strong correlation and the same distribution.
TABLE 2 partitioning of variable sub-blocks in megawatt-hour ultra-supercritical units
Figure BDA0002112808330000141
(4) Based on the variable block result obtained in the steps (2) and (3), firstly, describing the variable sub-block by using a Principal Component Analysis (PCA)
Figure BDA0002112808330000142
Correlation of each variable in the
Figure BDA0002112808330000143
Wherein, Pb,wIs a load matrix, Tb,wIs a principal component matrix.
(5) Principal component matrix T established by Gaussian Mixture Model (GMM) methodb,wThe distribution of (c):
Figure BDA0002112808330000144
wherein,
Figure BDA0002112808330000145
is the number of gaussian components;
Figure BDA0002112808330000146
the weight of the m-th component is represented,
Figure BDA0002112808330000147
is a mean value containing sub-Gaussian components
Figure BDA0002112808330000151
Sum covarianceMatrix array
Figure BDA0002112808330000152
Of the parameter set (c).
(6) For each variable sub-block
Figure BDA0002112808330000153
Establishing BIP statistics
Figure BDA0002112808330000154
Wherein,
Figure BDA0002112808330000155
to represent
Figure BDA0002112808330000156
Belonging to the m-th component
Figure BDA0002112808330000157
The probability of (a) of (b) being,
Figure BDA0002112808330000158
as a principal component matrix Tb,wAn nth (N ═ 1, 2.., N) row vector.
Figure BDA0002112808330000159
Is based on the probability of local Mahalanobis distance, which is defined as
Figure BDA00021128083300001510
Wherein,
Figure BDA00021128083300001511
is composed of
Figure BDA00021128083300001512
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(7) The relationship between each sub-block in each variable block is monitored by a Gaussian Mixture Model (GMM) method, which is implemented by the following sub-steps.
(7.1) blocking each variable block XbThe first columns of the pivot matrices of the respective sub-blocks are combined together:
Figure BDA00021128083300001513
wherein, tb,w(w＝1,2,...,Wb) As a principal component matrix Tb,wThe 1 st column vector.
(7.2) describing principal metadata with GMM
Figure BDA00021128083300001514
The distribution of (c):
Figure BDA00021128083300001515
wherein,
Figure BDA00021128083300001516
the number of Gaussian components in the b variable sub-block is shown;
Figure BDA00021128083300001517
the weight of the m-th component is represented,
Figure BDA00021128083300001518
is a mean value containing sub-Gaussian components
Figure BDA00021128083300001519
Sum covariance matrix
Figure BDA00021128083300001520
Of the parameter set (c).
(7.3) Main metadata for each variable Block
Figure BDA00021128083300001521
Establishing BIP statistic:
Figure BDA00021128083300001522
wherein,
Figure BDA00021128083300001523
as a principal component matrix
Figure BDA00021128083300001524
An nth (N ═ 1, 2.., N) row vector.
Figure BDA00021128083300001525
Is based on the probability of the local mahalanobis distance.
Figure BDA0002112808330000161
Is based on the probability of the local mahalanobis distance,
Figure BDA0002112808330000162
is composed of
Figure BDA0002112808330000163
Mahalanobis distance to the mth gaussian component,
Figure BDA0002112808330000164
is composed of
Figure BDA0002112808330000165
Any one of the rows.
(8.1) fault data preparation: here, the collected fault data contains 460 samples in total, and the data is recorded as Z (460 × 154), the fault is the increase of the circulating water pump outlet pressure, and the fault occurs at the 121 th sampling point.
(8.2) according to the variable blocking results obtained in the steps (2) and (3), new data is recorded as z (1 × 154), and sub-block decomposition is performed, wherein in the present example, B is 11:
z＝[z1z2... zb... zB](26)
Figure BDA0002112808330000166
wherein z isb(B ═ 1, 2.., B) is the B variable sub-block.
(8.3) at the bottom layer, i.e., variable sub-block layer, each sub-block
Figure BDA0002112808330000167
Figure BDA0002112808330000168
The data of (2) are projected to the principal element direction of the corresponding sub-block:
Figure BDA0002112808330000169
(8.4) obtaining each sub-block
Figure BDA00021128083300001610
The online statistic index of (1):
Figure BDA00021128083300001611
wherein, the meaning of each parameter in the above formula is similar to that in the formula (22).
Figure BDA00021128083300001612
To represent
Figure BDA00021128083300001613
Belonging to the m-th component
Figure BDA00021128083300001614
The probability of (c).
Figure BDA00021128083300001615
Is based on the probability of the local mahalanobis distance,
Figure BDA00021128083300001616
is composed of
Figure BDA00021128083300001617
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(8.5) at the variable block level, z is first putbThe main elements of each variable sub-block are combined together:
Figure BDA00021128083300001618
(8.6) obtaining each variable block zbThe online statistic index of (1):
Figure BDA00021128083300001619
wherein, the meaning of each parameter of the above formula is similar to that in the formula (25).
Figure BDA00021128083300001620
To represent
Figure BDA00021128083300001621
Belonging to the m-th component
Figure BDA0002112808330000171
The probability of (c).
Figure BDA0002112808330000172
Is based on the probability of the local mahalanobis distance,
Figure BDA0002112808330000173
is composed of
Figure BDA0002112808330000174
Mahalanobis distance to the mth gaussian component,
Figure BDA0002112808330000175
is composed of
Figure BDA0002112808330000176
Any one of the rows.
(8.7) in order to analyze the relation between different variable subblocks, the operation condition of the million kilowatt ultra-supercritical unit is monitored from the whole unit level, and the BIP index of each variable block is firstly converted into the probability of normal (marked as 'N') and fault (marked as 'F'):
Figure BDA0002112808330000177
Figure BDA0002112808330000178
wherein BIPb,lmtFor the control limit of the statistic BIP, 0.5 is taken in the specific embodiment;
Figure BDA0002112808330000179
representing the normal conditional probability of the b variable block;
Figure BDA00021128083300001710
indicating the conditional probability of the failure of the b-th variable block.
(8.8) calculating the posterior probability of the b variable block failing by the Bayes rule
Figure BDA00021128083300001711
Wherein, Pb(F)＝α；Pb1- α represents the prior probability of the process failing or being normal, respectively, at a significance level α, which in this example is 0.5.
(8.9) comprehensively considering the fault probability of all variable blocks and calculating the global monitoring statistic
Figure BDA00021128083300001712
(9) Judging the running state of the process: and analyzing the process state from three levels of the variable subblocks, the variable blocks and the whole unit. Three levels of statistics are compared with the control limits in real time:
(a) at each variable sub-block
Figure BDA00021128083300001713
In, if BIPb,w1-alpha, in sub-blocks
Figure BDA00021128083300001714
The variable in (b) fails, otherwise the variable in the sub-block is considered to be operating within the normal range.
(b) At the variable block level, if BIPbIf the value is more than 1-alpha, the related relation of each variable sub-block in the variable block is abnormal, otherwise, all the variables in the b-th sub-block are normally operated.
(c) At the unit level, if PFzIf the measured value is more than alpha, the abnormal condition or the fault occurs in the running process of the million kilowatt ultra-supercritical unit, otherwise, the whole unit runs normally.
The thermal power process is monitored on line by using the monitoring method disclosed by the invention, and the results are shown in fig. 2-4. FIG. 2 shows the monitoring results of the method of the present invention in 8 variable sub-blocks. As can be seen from fig. 2(a), the statistics of the two variable sub-blocks of the fifth variable block are basically below the control limit, which indicates that the current fault does not affect the variables in the variable block #5, and these variables are all operating normally. As can be seen from fig. 2(b), in the first 120 samples, the statistics of the first two sub-blocks of variable block #7 are basically below the control limit, indicating that the process is operating normally. Starting from the 121 th sample, the variable sub-block statistic BIPsub7,1And BIPsub7,2The control limit is quickly exceeded and a fault occurrence is detected, indicating that the fault significantly affects the variables in both sub-blocks. Also, analyzing the monitoring results of fig. 2(c), it can be found that the variables in the first two sub-blocks of variable block #9 operate substantially normally, but BIPsub9,3The occurrence of the failure is effectively detected.
Fig. 3 shows the partial monitoring results of the method of the invention in the variable layer. As can be seen in fig. 3(a), the fifth variable block is substantially below the control limit, indicating that the correlation between the two variable sub-blocks in variable block #5 is not affected by a fault and the process variable in variable block #5 is operating normally. The monitoring results of FIG. 3(b) and FIG. 3(c) are shown separatelyThe correlation between the variable subblocks in the variable block #7 and the variable block #9 is affected by a failure, and an abnormality occurs. Fig. 4 shows the monitoring results of the inventive method at the stack level. It can be seen that the statistics of the first 120 samples are basically operated below the control limit, which indicates that the megawatt ultra-supercritical unit is operated under the normal working condition. Starting from the 121 th sample, the statistic PFzAnd immediately exceeding the control limit, and effectively detecting the occurrence of the fault. Generally speaking, the layered distributed fault detection method has excellent fault detection performance when monitoring a typical large-scale multi-working-condition process of million kilowatt ultra-supercritical, and the blocking result effectively analyzes complex correlation among a plurality of variables, so that the understanding of an operator to the process can be deepened, high-precision online process monitoring results can be provided for a technical management department on an actual industrial field of a thermal power plant, a reliable basis is provided for judging the process running state in real time and identifying whether a fault occurs, and the safety, reliability and effectiveness of running of the million kilowatt ultra-supercritical unit are further improved.

Claims (1)

1. A distribution layered online fault detection method for a million kilowatt ultra-supercritical unit is characterized by comprising the following steps:
(1) acquiring normal data to be analyzed: a million-kilowatt ultra-supercritical unit is provided with J measurement variables and operation variables, a vector of 1 XJ can be obtained by sampling every time, and data acquired after sampling N times is expressed as a two-dimensional matrix X ═ X1,X2,...,XJ]∈RN×JThe measured variables are state parameters which can be measured in the normal operation process of the unit, and comprise flow, voltage, current, temperature, speed and the like; the operation variables comprise air intake, feeding amount, valve opening and the like;
(2) the process variable is divided into different sub-blocks by using a spectral clustering method based on mutual information, the variable in the same sub-block has stronger correlation, and the correlation between different sub-blocks is weaker. This step is realized by the following substeps:
(2.1) obtaining mutual information among variables:
I(Xi,Xj)＝H(Xi)+H(Xj)-H(Xi,Xj) (1)
wherein, Xi(i ═ 1, 2.., J) denotes the ith variable, H (X)i) Is a variable XiThe information entropy of (2):
H(Xi)＝-∫xp(Xi)logp(Xi)dx (2)
H(Xi,Xj) Is a variable XiAnd XjJoint information entropy of (a):
Figure FDA0002112808320000011
p(Xi) And p (X)j) Represents variable XiAnd XjP (X) is a probability density function ofi,Xj) Is a joint probability density function.
(2.2) solving the generalized correlation coefficient between every two variables based on the mutual information solved by the formula (1):
Figure FDA0002112808320000012
wherein r isij∈[0,1]。
(2.3) based on equation (4), a correlation matrix of the variables is found:
Figure FDA0002112808320000021
(2.4) solving a diagonal matrix D based on the correlation matrix R defined by the formula (5):
D＝diag{Dii} (6)
wherein D isiiIs the sum of all elements in row i in equation (5):
Figure FDA0002112808320000022
(2.5) solving a Laplace matrix of the diagonal matrix D:
L＝D-1/2RD-1/2(8)
(2.6) performing spectral decomposition on the Laplace matrix:
L＝PΛPT(9)
wherein, P ═ P1,P2,...,PJ]Are orthogonal eigenvectors.
(2.7) selecting the eigenvectors corresponding to the k maximum eigenvalues to form a matrix E ═ P1,P2,...,Pk]∈RJ×kNormalizing each row in the matrix E to obtain a matrix Y:
Figure FDA0002112808320000023
(2.8) clustering Y by using a kmeans clustering algorithm, and if the ith row belongs to the b-th class, carrying out variable XiDivision into sub-block b Xb. Thus, a plurality of operation variables of the million-kilowatt ultra-supercritical unit are divided into B variable blocks according to the correlation degree:
X＝[X1X2… XB](11)
wherein,
Figure FDA0002112808320000024
is the (B ═ 1, 2.., B) th variable block, JbRepresents XbThe number of variables contained in (1).
(3) The variables in the variable block are further decomposed according to the distribution situation in the sample direction by using an information theory decomposition method based on a Gaussian mixture model, and the step is realized by the following substeps.
(3.1) randomly dividing the variable block into WbIndividual block:
Figure FDA0002112808320000031
(3.2) using gaussian mixture model method to obtain W (W1, 2.., W)b) Probability density of individual variable subblocks:
Figure FDA0002112808320000032
wherein,
Figure FDA0002112808320000033
is the number of sub-gaussian components;
Figure FDA0002112808320000034
is the prior probability of the mth sub-Gaussian component, satisfies
Figure FDA0002112808320000035
And
Figure FDA0002112808320000036
is a mean value containing sub-Gaussian components
Figure FDA0002112808320000037
Sum covariance matrix
Figure FDA0002112808320000038
Of the parameter set (c).
Figure FDA0002112808320000039
Is a multivariate gaussian probability density:
Figure FDA00021128083200000310
wherein, Jb,wIs composed of
Figure FDA00021128083200000311
The number of medium variables.
(3.3) solving probability density distribution functions of all variables in the sub-blocks:
Figure FDA00021128083200000312
wherein the variable
Figure FDA00021128083200000313
Jb,wIs a variable block
Figure FDA00021128083200000314
The number of the medium variables is equal to or greater than the total number of the medium variables,
Figure FDA00021128083200000315
represents Xb,iBelonging to sub-blocks
Figure FDA00021128083200000316
When Xb,iThe conditional probability density of (2).
(3.4) solving variable subblocks
Figure FDA00021128083200000317
And
Figure FDA00021128083200000318
KL divergence of (W, v ∈ [1, W ]b])：
Figure FDA00021128083200000319
Wherein,
Figure FDA00021128083200000320
and
Figure FDA00021128083200000321
are respectively
Figure FDA00021128083200000322
And
Figure FDA00021128083200000323
the probability density function of (2) can be calculated by using the formula (13).
(3.5) optimizing the random partitions of step (3.1) using an ant colony algorithm such that the following objective function is maximized:
Figure FDA00021128083200000324
(3.6) each variable block (B ═ 1, 2.., B) is further divided into sub-blocks by repeating steps (3.2) - (3.5). The original data set X is divided into different sub-blocks:
Figure FDA0002112808320000041
wherein, the variable block
Figure FDA0002112808320000042
(4) Based on the variable block result obtained in the steps (2) and (3), firstly, describing the variable sub-block by using a Principal Component Analysis (PCA)
Figure FDA0002112808320000043
The correlation relationship of each variable in (1):
Figure FDA0002112808320000044
wherein, Pb,wIs a load matrix, Tb,wIs a principal component matrix.
(5) Principal component matrix T established by Gaussian Mixture Model (GMM) methodb,wThe distribution of (c):
Figure FDA0002112808320000045
wherein,
Figure FDA0002112808320000046
is the number of gaussian components;
Figure FDA0002112808320000047
the weight of the m-th component is represented,
Figure FDA0002112808320000048
is a mean value containing sub-Gaussian components
Figure FDA0002112808320000049
Sum covariance matrix
Figure FDA00021128083200000410
Of the parameter set (c).
(6) For each variable sub-block
Figure FDA00021128083200000411
Establishing BIP statistics
Figure FDA00021128083200000412
Wherein,
Figure FDA00021128083200000413
to represent
Figure FDA00021128083200000414
Belonging to the m-th component
Figure FDA00021128083200000415
The probability of (a) of (b) being,
Figure FDA00021128083200000416
as a principal component matrix Tb,wAn nth (N ═ 1, 2.., N) row vector.
Figure FDA00021128083200000417
Is based on the probability of local Mahalanobis distance, which is defined as
Figure FDA00021128083200000418
Wherein,
Figure FDA00021128083200000419
is composed of
Figure FDA00021128083200000420
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(7) The relationship between each sub-block in each variable block is monitored by a Gaussian Mixture Model (GMM) method, which is implemented by the following sub-steps.
(7.1) blocking each variable block XbThe first columns of the pivot matrices of the respective sub-blocks are combined together:
Figure FDA0002112808320000051
wherein, tb,w(w＝1,2,...,Wb) As a principal component matrix Tb,wThe 1 st column vector.
(7.2) describing principal metadata with GMM
Figure FDA0002112808320000052
The distribution of (c):
Figure FDA0002112808320000053
wherein,
Figure FDA0002112808320000054
the number of Gaussian components in the b variable sub-block is shown;
Figure FDA0002112808320000055
the weight of the m-th component is represented,
Figure FDA0002112808320000056
is composed of sub-Gaussian componentsMean value of
Figure FDA0002112808320000057
Sum covariance matrix
Figure FDA0002112808320000058
Of the parameter set (c).
(7.3) Main metadata for each variable Block
Figure FDA0002112808320000059
Establishing BIP statistic:
Figure FDA00021128083200000510
wherein,
Figure FDA00021128083200000511
as a principal component matrix
Figure FDA00021128083200000512
An nth (N ═ 1, 2.., N) row vector.
Figure FDA00021128083200000513
To represent
Figure FDA00021128083200000514
Belonging to the m-th component
Figure FDA00021128083200000515
The probability of (c).
Figure FDA00021128083200000516
Is based on the probability of the local mahalanobis distance,
Figure FDA00021128083200000517
is composed of
Figure FDA00021128083200000518
Mahalanobis distance to the mth gaussian component,
Figure FDA00021128083200000519
is composed of
Figure FDA00021128083200000520
Any one of the rows.
(8) During online fault detection, the process is monitored from three levels of variable subblocks, variable blocks and the whole unit. This step is realized by the following substeps.
(8.1) acquiring new data: and (4) collecting the variable values of the measuring points according to the step (1) and recording as z (1 multiplied by J).
(8.2) according to the variable blocking results obtained in the step (2) and the step (3), carrying out sub-block decomposition on the new data:
z＝[z1z2... zb... zB](26)
Figure FDA00021128083200000521
wherein z isb(B ═ 1, 2.., B) is the B variable sub-block.
(8.3) at the bottom layer, i.e., variable sub-block layer, each sub-block
Figure FDA00021128083200000522
Figure FDA00021128083200000523
The data of (2) are projected to the principal element direction of the corresponding sub-block:
Figure FDA0002112808320000061
wherein
Figure FDA0002112808320000062
Is a load matrix.
(8.4) obtaining each sub-block
Figure FDA0002112808320000063
The online statistic index of (1):
Figure FDA0002112808320000064
wherein, the meaning of each parameter in the above formula is similar to that in the formula (22).
Figure FDA0002112808320000065
To represent
Figure FDA0002112808320000066
Belonging to the m-th component
Figure FDA0002112808320000067
The probability of (c).
Figure FDA0002112808320000068
Is based on the probability of the local mahalanobis distance,
Figure FDA0002112808320000069
is composed of
Figure FDA00021128083200000610
Mahalanobis distance to mth gaussian component, T being Tb,wAny one of the rows.
(8.5) at the variable block level, z is first putbThe main elements of each variable sub-block are combined together:
Figure FDA00021128083200000611
(8.6) obtaining each variable block zbThe online statistic index of (1):
Figure FDA00021128083200000612
wherein, the meaning of each parameter of the above formula is similar to that in the formula (25).
Figure FDA00021128083200000613
To represent
Figure FDA00021128083200000614
Belonging to the m-th component
Figure FDA00021128083200000615
The probability of (c).
Figure FDA00021128083200000616
Is based on the probability of the local mahalanobis distance,
Figure FDA00021128083200000617
is composed of
Figure FDA00021128083200000618
Mahalanobis distance to the mth gaussian component,
Figure FDA00021128083200000619
is composed of
Figure FDA00021128083200000620
Any one of the rows.
(8.7) converting the BIP index of each variable block into the probability of normal (labeled 'N') and fault (labeled 'F'):
Figure FDA00021128083200000621
Figure FDA00021128083200000622
wherein BIPb,lmtA control limit for the statistical BIP indicator;
Figure FDA00021128083200000623
representing the normal conditional probability of the b variable block;
Figure FDA00021128083200000624
indicating the conditional probability of the failure of the b-th variable block.
(8.8) calculating the posterior probability of the b variable block failing by the Bayes rule
Figure FDA0002112808320000071
Wherein, Pb(F)＝α；Pb(N) ═ 1- α represents the prior probability of the process failing or being normal, respectively, at a level of significance α.
(8.9) comprehensively considering the fault probability of all variable blocks and calculating the global monitoring statistic
Figure FDA0002112808320000072
(9) Judging the running state of the process: and analyzing the process state from three levels of the variable subblocks, the variable blocks and the whole unit. Three levels of statistics are compared with the control limits in real time:
(a) at each variable sub-block
Figure FDA0002112808320000073
In, if BIPb,w1-alpha, in sub-blocks
Figure FDA0002112808320000074
The variable in (b) fails, otherwise the variable in the sub-block is considered to be operating within the normal range.
(b) At the variable block level, if BIPbIf the value is more than 1-alpha, the related relation of each variable sub-block in the variable block is abnormal, otherwise, all the variables in the b-th sub-block are normally operated.
(c) At the unit level, if PFzIf the measured value is more than alpha, the abnormal condition or the fault occurs in the running process of the million kilowatt ultra-supercritical unit, otherwise, the whole unit runs normally.
CN201910579518.8A 2019-06-28 2019-06-28 Online fault detection method for million-kilowatt ultra-supercritical unit Active CN111880090B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910579518.8A CN111880090B (en) 2019-06-28 2019-06-28 Online fault detection method for million-kilowatt ultra-supercritical unit

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910579518.8A CN111880090B (en) 2019-06-28 2019-06-28 Online fault detection method for million-kilowatt ultra-supercritical unit

Publications (2)

Publication Number Publication Date
CN111880090A true CN111880090A (en) 2020-11-03
CN111880090B CN111880090B (en) 2021-07-06

Family

ID=73153874

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910579518.8A Active CN111880090B (en) 2019-06-28 2019-06-28 Online fault detection method for million-kilowatt ultra-supercritical unit

Country Status (1)

Country Link
CN (1) CN111880090B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113092907A (en) * 2021-04-02 2021-07-09 长春工业大学 System fault detection method based on block slow characteristic analysis
CN113158769A (en) * 2021-03-03 2021-07-23 安徽大学 CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5461329A (en) * 1992-01-21 1995-10-24 Martin Marietta Energy Systems, Inc. Method and apparatus for generating motor current spectra to enhance motor system fault detection
CN102086784A (en) * 2010-12-16 2011-06-08 浙江大学 Distributed remote vibration monitoring and fault diagnosis system of large steam turbine-generator
CN105425779A (en) * 2015-12-24 2016-03-23 江南大学 ICA-PCA multi-working condition fault diagnosis method based on local neighborhood standardization and Bayesian inference
CN108490908A (en) * 2018-02-11 2018-09-04 浙江大学 A kind of dynamic distributed monitoring method towards gigawatt extra-supercritical unit variable parameter operation
CN108508866A (en) * 2018-03-21 2018-09-07 浙江大学 A kind of gigawatt extra-supercritical unit failure identification variables method based on sparse opposite discriminant analysis
CN109491338A (en) * 2018-11-09 2019-03-19 南通大学 A kind of relevant method for diagnosing faults of multimode procedure quality based on sparse GMM

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5461329A (en) * 1992-01-21 1995-10-24 Martin Marietta Energy Systems, Inc. Method and apparatus for generating motor current spectra to enhance motor system fault detection
CN102086784A (en) * 2010-12-16 2011-06-08 浙江大学 Distributed remote vibration monitoring and fault diagnosis system of large steam turbine-generator
CN105425779A (en) * 2015-12-24 2016-03-23 江南大学 ICA-PCA multi-working condition fault diagnosis method based on local neighborhood standardization and Bayesian inference
CN108490908A (en) * 2018-02-11 2018-09-04 浙江大学 A kind of dynamic distributed monitoring method towards gigawatt extra-supercritical unit variable parameter operation
CN108508866A (en) * 2018-03-21 2018-09-07 浙江大学 A kind of gigawatt extra-supercritical unit failure identification variables method based on sparse opposite discriminant analysis
CN109491338A (en) * 2018-11-09 2019-03-19 南通大学 A kind of relevant method for diagnosing faults of multimode procedure quality based on sparse GMM

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
DONG WANG ET AL.: "A simple and fast guideline for generating enhanced/squared envelope spectra from spectral coherence for bearing fault diagnosis", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING》 *
范海东等: "基于稀疏故障演化判别分析的故障根源追溯", 《控制工程》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113158769A (en) * 2021-03-03 2021-07-23 安徽大学 CEEMDAN and FastICA-based electromechanical device bearing vibration signal denoising method
CN113092907A (en) * 2021-04-02 2021-07-09 长春工业大学 System fault detection method based on block slow characteristic analysis

Also Published As

Publication number Publication date
CN111880090B (en) 2021-07-06

Similar Documents

Publication Publication Date Title
Wang et al. Recent advances and summarization of fault diagnosis techniques for proton exchange membrane fuel cell systems: A critical overview
US11740619B2 (en) Malfunction early-warning method for production logistics delivery equipment
CN106709662B (en) Power equipment operation condition division method
CN111237134B (en) Offshore double-fed wind driven generator fault diagnosis method based on GRA-LSTM-stacking model
CN111537219B (en) Fan gearbox performance detection and health assessment method based on temperature parameters
CN104809658B (en) A kind of rapid analysis method of low-voltage distribution network taiwan area line loss
CN110458230A (en) A kind of distribution transforming based on the fusion of more criterions is with adopting data exception discriminating method
CN110362045B (en) Marine doubly-fed wind turbine generator fault discrimination method considering marine meteorological factors
CN112036089A (en) Coal mill fault early warning method based on DPC-MND and multivariate state estimation
CN109359662B (en) Non-stationary analysis and causal diagnosis method for million-kilowatt ultra-supercritical unit
CN108490908B (en) A kind of dynamic distributed monitoring method towards gigawatt extra-supercritical unit variable parameter operation
CN112836941B (en) Online health condition assessment method for high-pressure system of steam turbine of coal motor unit
CN111880090B (en) Online fault detection method for million-kilowatt ultra-supercritical unit
CN111652479A (en) Data driving method for dynamic security assessment of power system
CN110320893A (en) A kind of fired power generating unit device maintenance method based on RCM
CN111585277B (en) Power system dynamic security assessment method based on hybrid integration model
Luo et al. Extreme scenario extraction of a grid with large scale wind power integration by combined entropy-weighted clustering method
Wang et al. A SCADA-data-driven condition monitoring method of wind turbine generators
Bao et al. Wind turbine condition monitoring based on improved active learning strategy and KNN algorithm
CN111461565A (en) Power supply side power generation performance evaluation method under power regulation
Zhu et al. Robust representation learning for power system short-term voltage stability assessment under diverse data loss conditions
CN112380763A (en) System and method for analyzing reliability of in-pile component based on data mining
Hempelmann et al. Evaluation of unsupervised anomaly detection approaches on photovoltaic monitoring data
Liu et al. Discriminative Signal Recognition for Transient Stability Assessment via Discrete Mutual Information Approximation and Eigen Decomposition of Laplacian Matrix
Zhou et al. Abnormal data processing of wind turbine based on combined algorithm and class center imputation

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
GR01 Patent grant
GR01 Patent grant