CN109538311B - Real-time monitoring method for control performance of steam turbine in high-end power generation equipment - Google Patents

Real-time monitoring method for control performance of steam turbine in high-end power generation equipment Download PDF

Info

Publication number
CN109538311B
CN109538311B CN201811110373.9A CN201811110373A CN109538311B CN 109538311 B CN109538311 B CN 109538311B CN 201811110373 A CN201811110373 A CN 201811110373A CN 109538311 B CN109538311 B CN 109538311B
Authority
CN
China
Prior art keywords
matrix
new
slow
monitoring
steam turbine
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.)
Active
Application number
CN201811110373.9A
Other languages
Chinese (zh)
Other versions
CN109538311A (en
Inventor
赵春晖
李明超
范海东
陈积明
孙优贤
李清毅
沙万里
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 CN201811110373.9A priority Critical patent/CN109538311B/en
Publication of CN109538311A publication Critical patent/CN109538311A/en
Application granted granted Critical
Publication of CN109538311B publication Critical patent/CN109538311B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
    • F01—MACHINES OR ENGINES IN GENERAL; ENGINE PLANTS IN GENERAL; STEAM ENGINES
    • F01D—NON-POSITIVE DISPLACEMENT MACHINES OR ENGINES, e.g. STEAM TURBINES
    • F01D21/00—Shutting-down of machines or engines, e.g. in emergency; Regulating, controlling, or safety means not otherwise provided for
    • F01D21/003—Arrangements for testing or measuring

Abstract

The invention discloses a real-time monitoring method for control performance of a steam turbine in high-end power generation equipment. The invention discloses a method for extracting relevant information among variables of a steam turbine control system by using typical variable analysis, and then extracting dynamic information in the relevant information by using a slow characteristic analysis algorithm, aiming at the problem that the control performance monitoring of a steam turbine in high-end power generation equipment is difficult due to numerous parameters and variable working conditions. And finally, combining the correlation and the change speed information of the variables to construct an online monitoring model of the control performance of the steam turbine. The method solves the problem that the control performance of a large-scale steam turbine is difficult to monitor due to numerous variables and working condition changes, greatly improves the accuracy of the on-line monitoring of the dynamic process control performance, is beneficial to effectively and timely monitoring a steam turbine control system by a thermal power plant, and has great significance for ensuring the safe and reliable operation of high-end power generation equipment.

Description

Real-time monitoring method for control performance of steam turbine in high-end power generation equipment
Technical Field
The invention belongs to the field of performance monitoring of thermal power process control systems, and particularly relates to an online performance monitoring method for correlation information and dynamic information of steam turbine operation in high-end power generation equipment.
Background
The control system occupies a very important position in the modern industrial process, and the indexes influencing the economic benefit, such as production quality, operation safety, physical energy consumption and the like, are directly or indirectly related to the performance of the control system. In the actual production process, the performance of the control system is good at the initial stage of putting into use, but after the control system runs for a period of time, the performance of the control system can be reduced due to the reasons of abrasion of equipment, untimely maintenance and untimely maintenance, the control performance is poor, the production quality can be directly influenced, the economic benefit is lost, and if production faults are caused, the life safety of people and even the property safety of social enterprises can be involved, so that great threat is brought. Torres et al examined more than 700 control loops in 12 brazil factories (petrifaction, paper making, cement, steel, mining, etc.) in 2005, and the results showed that 14% of loops had excessive valve wear, 15% of the valves had hysteresis problems, 16% of the loops had severe setting problems, 24% of the controller outputs had saturation, and 41% of the loops had oscillation phenomena due to the setting problems, coupling, disturbance and actuator problems.
In addition, in actual production, thousands of control loops may be combined in a production process, and 14000 control loops are available for two distillation production facilities in Eastman chemical company, and the number of control loops can reach even one hundred thousand in the HVAC production process. The high-end power generation equipment has higher complexity, and is embodied in the aspects of large scale, numerous equipment, diversified parameters, mutual influence and the like. In addition, the large-scale generator set has the characteristics of high temperature, high pressure, high noise and the like on site.
The control performance evaluation and monitoring technology is an important technology emerging in the field of process control, and can monitor the change of the control performance of a monitoring system in real time by utilizing the daily operation data of equipment to perform early identification and optimization on the problems of the control system. For the generator set, because the power load in the power system is constantly changed, in order to maintain the active power balance and keep the system frequency stable, the output of the generator needs to be correspondingly changed by the power generation department to adapt to the change of the power load, that is, the working condition of the generator set is not stable and constant. However, the existing control performance evaluation and monitoring methods such as principal component analysis, partial least square method and fisher discriminant analysis are all performed based on the ideal assumption that the working conditions are stable, so that the performance monitoring method cannot achieve a good monitoring effect when the method is applied to the performance monitoring of the steam turbine control system of the large-scale thermal power generator unit.
Disclosure of Invention
The steam turbine is a rotary heat engine which takes continuously flowing steam as a working medium and converts the heat energy of the steam into mechanical energy. Generally, the turbine engine is composed of a nozzle (or a stationary vane), a moving blade, an impeller, a shaft bearing, a cylinder and the like. The steam turbine has the advantages of high rotating speed, stable and reliable operation, large unit power (over 1000MW) and convenient direct connection with the generator, but the whole structure is complex, has strict requirements on design, manufacture, installation, operation and maintenance technologies, and is necessary to be provided with boilers with corresponding steam parameters and capacity. Besides being widely used as main engines of modern medium and large thermal power plants, nuclear power stations and large ships, the engine also can be used in large chemical industry or steel and other combined enterprises.
The invention aims to solve the problem that a steam turbine of a large thermal generator set is difficult to monitor control performance due to numerous parameters, complex structure and variable working conditions, extract relevant information and change speed information among variables of a steam turbine control system by using a typical variable analysis and slow characteristic analysis fusion algorithm, and overcome the problem that the large steam turbine is difficult to monitor control performance due to numerous variables and working condition changes.
The purpose of the invention is realized by the following technical scheme: the method for monitoring the control performance of the steam turbine in the high-end power generation equipment in real time comprises the following steps:
(1) acquiring training data, wherein a control system of the steam turbine has J measurement variables and operation variables, and an observation vector y of J × 1 can be obtained by sampling each timekWherein the subscript k is a time index, and data obtained after N times of sampling is expressed as a two-dimensional observation matrix
Figure GDA0001941354690000021
The measured variable is steamThe state parameters which can be measured in the running process of the turbine comprise bearing metal temperature, bearing axial vibration, generator active power, steam turbine rotating speed, excitation three-phase temperature and the like; the operation variables comprise condensate flow (average of three minutes), feed water pressure, feed water flow, coal feed amount of a coal feeder and the like; the training data should be selected from sampled data of the steam turbine under normal operating conditions.
(2) Extracting time sequence related information of data by using a CVA algorithm, wherein the step is realized by the following substeps:
(2.1) constructing a past matrix and a future matrix by the time sequence expansion: at a particular sampling instant k, the observation vector y iskExpanding p steps to k times to generate past observation vector
Figure GDA0001941354690000031
Expanding f steps after k to generate future observation vectors
Figure GDA0001941354690000032
Then to yp,k，yf,kCarrying out equalization treatment:
Figure GDA0001941354690000033
wherein: mean (y)p,k) To represent
Figure GDA0001941354690000034
Mean of (y)f,k) To represent
Figure GDA0001941354690000035
Is measured.
The past observation matrix Y is constructed by using all past observation vectors and future observation vectors respectivelypAnd future observation matrix Yf：
Figure GDA0001941354690000036
Where M is N-f-p +1, p, f are two types of time lag parameters, and let p be f, whose value can be determined by the sample autocorrelation function:
Figure GDA0001941354690000037
wherein: autocorr (Y)jP) represents the matrix YpThe autocorrelation coefficient of the jth column vector and its time lag p;
(2.2) construction of Hankel matrix calculation of covariance matrix of past and future matrices ∑pp，∑ffAnd their cross-covariance matrices ∑fpAnd then constructing a Hankel matrix H by utilizing the covariance matrix and the cross covariance matrix:
Figure GDA0001941354690000041
Figure GDA0001941354690000042
(2.3) singular value decomposition: performing singular value decomposition on the Hankel matrix to obtain Jp group typical variable pairs, using (a)i TYp,bi TYf) Representing the i-th set of pairs of typical variables, ai T、bi TRepresents the correlation coefficient between the i-th group of typical variable pairs:
H＝UDVT(6)
Figure GDA0001941354690000043
D＝diag(γ1,γ2,…,γJp)
u and V are singular vectors Ui，viForming an orthogonal matrix, D is a singular value matrix, singular vectors in U, V are only in pairwise correlation, and the correlation value is determined by the corresponding ith singular value gamma in DiAnd (5) characterizing. The larger the singular value (gamma)1>γ2>…>γJp) The greater the correlation between the typical variables.
(2.4) calculating transformation matrix and extracting representativenessVariables and residual variables: truncation matrix
Figure GDA0001941354690000044
The first r column of (2), generate the matrix after dimensionality reduction
Figure GDA0001941354690000045
VrMost of the timing related information is still retained. Wherein, the magnitude of the r value can be determined by the following criteria:
Figure GDA0001941354690000051
cr denotes a criterion value, β denotes a determination threshold, β ═ 0.5.
From VrCalculating the representative variable transformation matrix C and the residual variable transformation matrix L:
Figure GDA0001941354690000052
and obtaining a typical variable space Z and a residual error space E by using the transformation matrix:
Figure GDA0001941354690000053
z, column vector Z in Εk∈r×1，k∈ Jp × 1 denotes the typical and residual variables at the sampling instant k, respectively, Z, the line vector Z in Et，tTime sequence information of the same variable at different time is contained.
(3) Slow Feature Analysis (SFA) is utilized to respectively extract Slow features s in typical variable space Z and residual space EeZ，sE. To extract slow features s in the typical variable space ZZFor example, the method mainly comprises the following steps:
(3.1) data normalization: the typical variable space Z is normalized according to variables, and the calculation formula is as follows:
Figure GDA0001941354690000054
zttable time sequence vectors, mean (z), of the same variable at different timest) Denotes ztMean value of (a), std (z)t) Denotes ztStandard deviation of (2).
(3.2) the output signal of Z after projection is sZj，sZjDenotes sZThe jth slow signature sequence. In consideration of the linear condition, the method is,
Figure GDA0001941354690000055
representing a coefficient vector, which is equivalent to finding a slow feature signal s extracted from the normalized input signal ZZ＝[sZ1 T，sZ2 T，...，sZr T]TIs converted into a matrix
Figure GDA0001941354690000061
I.e. sZ＝WZAnd Z. Slow characteristic signal sZjThe objective function and the constraint condition to be satisfied are as follows:
an objective function:
Figure GDA0001941354690000062
the constraint conditions are as follows:
Figure GDA0001941354690000063
wherein:
Figure GDA0001941354690000064
signal s representing a slow characteristicZIs calculated by time-sequence difference<·>Is shown as
Figure GDA0001941354690000065
t1，t0Respectively representing the upper and lower time limits.
(3.3) whitening: covariance matrix of input data using singular value decomposition<ZZT>The whitening treatment can remove the correlation in the data, so that the extracted slow characteristic value carries different information:
Figure GDA0001941354690000066
Figure GDA0001941354690000067
wherein ΛZ -1/2BTBeing a whitening matrix, OZIs the corresponding whitened input signal.
(3.4) calculating the transformation matrix WZ: to the input matrix OZDifferential processing is carried out to obtain a time sequence differential signal
Figure GDA0001941354690000068
Can prove that
Figure GDA0001941354690000069
Covariance matrix of
Figure GDA00019413546900000610
After singular value decomposition, a series of singular values omega are obtainedZjThe objective function value described in equation (12)
Figure GDA00019413546900000611
Figure GDA00019413546900000612
WZ＝PΛZ -1/2BT(17)
The slow features s in the residual space ΕEAnd the slow characteristic s in the above typical variable space ZZThe extraction method is the same.
(4) Partitioning slow features sZ: the slowest feature corresponds to the minimum feature value, the feature values are arranged from small to large, and the first features are divided into s according to the size of the feature valuesZOf slower changing features, by sZ,dRepresents; dividing the last (r-l) features into sZFeature of faster change by sZ,eAnd (4) showing. The determination method of the division basis l is that firstly, a slow characteristic value s is utilizedZIs used to represent the process variable
Figure GDA0001941354690000071
The speed of change of (2):
Figure GDA0001941354690000072
wherein: r isjiIs a matrix RZElement of row j and column i, sZiRepresents the ith slow signature sequence, and Δ (-) represents an operation to calculate how slowly the sequence changes:
Figure GDA0001941354690000073
dividing the extracted features with slowness larger than the slowness of the input data in the slow feature values into fast features, wherein the fast features have M in totaleOne such fast feature:
Figure GDA0001941354690000074
here card {. is } represents the number of elements in the set {. is }. M determined according to equation (19)eValue, corresponding to the matrix omegaZIt is also divided into two parts:
Figure GDA0001941354690000075
Figure GDA0001941354690000076
(5) calculating dynamic monitoring indexes: starting from the first sample point of the typical variable space, each sample point can obtain a set of dynamic monitoring indexes (S)Z,d 2，SZ,e 2)。
Figure GDA0001941354690000077
(6) Determining a control limit based on the dynamic monitoring indicator: by using the method of nuclear density estimation, a dynamic monitoring index S is estimated firstlyZ,d 2For a given significance level α, SZ,d 2Control limit of
Figure GDA0001941354690000083
The calculation method is as follows:
Figure GDA0001941354690000081
s can be calculated in the same wayZ,e 2Control limit of
Figure GDA0001941354690000084
(7) According to the method from the step (3) to the step (6), extracting the slow features s of the residual space EEEAnd will sEDivided into two parts sE,d，sE,eEstablishing a monitoring index SE,d 2，SE,e 2And calculating a control limit
Figure GDA0001941354690000085
Figure GDA0001941354690000086
The processing method is the same as that of the typical variable space Z, and the description is omitted.
(8) And (3) online monitoring and controlling performance: on-line monitoring the performance state of the steam turbine control system based on the CVA-SFA model established in the steps (2) to (4) and the four monitoring statistics obtained in the steps (5) to (7), wherein the steps are realized by the following sub-steps:
(8.1) acquiring new online data and preprocessing the new data: collecting a new section of observation data
Figure GDA0001941354690000082
Thereafter, where the following Table new represents the new observed data, Y is first added according to step (2)newDeveloping into a past matrix, and normalizing the past matrix according to the mean value and the standard deviation obtained in the step (2) to obtain Ypnew。
(8.2) extracting typical variables and residual variables of the new observed data: after the standardization processing, the conversion matrix V determined in the step (2) is utilizedrAnd L calculating the typical variable space Z of the new observed datanewAnd residual space Enew。
(8.3) extracting the typical variable space Z of the new observed datanewSlow characteristics of (1): firstly, according to the mean value and variance pair Z determined in the step (3.1)newPerforming normalization processing, and then using the slow feature transformation matrix W determined in step (3.4)ZExtracting the normalized ZnewSlow characteristic s ofZnewAnd dividing s according to the previous division parameterZnewIs divided intoZ,d newAnd sZ,e newAlso according to WECan obtain EnewFurther obtaining sE,d newAnd sE,e new。
(8.4) calculating a new monitoring statistical index: calculating a monitoring statistical index in a typical variable space according to the established model and the calculation method determined in the steps (5) and (7)
Figure GDA0001941354690000092
Sum residual space monitoring index
Figure GDA0001941354690000093
Figure GDA0001941354690000091
(8.5) online judging the control performance state of the steam turbine: comparing the four monitoring indexes with respective statistical control limits in real time, and if the four monitoring indexes are all located within the statistical control limits, indicating that the control system works normally; if one or more monitoring indexes exceed the normal control limit, the abnormal condition of the control system is indicated.
The invention has the beneficial effects that: the invention provides a real-time monitoring method for the control performance of a steam turbine in high-end power generation equipment, aiming at the problem of difficult monitoring of the control performance of the steam turbine of a large thermal power generating unit caused by numerous parameters, complex structure and variable working conditions. And finally, combining the correlation and the change speed information of the variables to construct an online monitoring model of the control performance of the steam turbine. The method solves the problem that the control performance of the large steam turbine is difficult to monitor due to numerous variables and working condition changes, greatly improves the accuracy of the on-line monitoring of the dynamic process control performance, is beneficial to effectively and timely monitoring a steam turbine control system by a thermal power plant, is beneficial to ensuring the safe and reliable operation of a large generating set, and meets the production requirement of improving the production benefit.
Drawings
FIG. 1 is a flow chart of a method for monitoring the control performance of a steam turbine in high-end power generation equipment in real time according to the present invention, (a) is a flow chart of an off-line modeling process, and (b) is a flow chart of an on-line monitoring process;
FIG. 2 is a graph showing the results of the CVA-SFA method of the present invention in statistical process monitoring, where (a) is a graph showing the results of normal monitoring, and (b) is a graph showing the results of abnormal monitoring.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific examples.
The invention takes a steam turbine of a No. 5 unit of a Jianghua power plant belonging to Zhe energy group as an example, the power of the steam turbine is 60 ten thousand kilowatts, the steam turbine is a large-scale thermal power generator set, and the steam turbine comprises 60 process variables, including bearing metal temperature, bearing axial vibration, generator active power, steam turbine rotating speed, excitation three-phase temperature, condensate water flow (average three minutes), water supply pressure, water supply flow, coal supply amount of a coal feeder and the like, and valve opening degrees.
It should be understood that the present invention is not limited to the thermal power generation process of the above example, and that equivalent modifications or substitutions can be made by those skilled in the art without departing from the present invention, and the equivalents or substitutions are included in the scope of the claims of the present application.
As shown in fig. 1, the invention is a real-time monitoring method for control performance of a steam turbine in high-end power generation equipment, comprising the following steps:
(1) acquiring training data, wherein a control system of the steam turbine has J measurement variables and operation variables, and an observation vector y of J × 1 can be obtained by sampling each timekWherein the subscript k is a time index, and data obtained after N times of sampling is expressed as a two-dimensional observation matrix
Figure GDA0001941354690000101
In the example, the sampling period is 10 minutes, 4566 samples are used totally, 60 process variables are used, and the measured variables are flow, vibration, temperature, pressure, valve opening and the like in the operation process of the steam turbine;
(2) extracting time sequence related information of data by using a CVA algorithm, wherein the step is realized by the following substeps:
(2.1) constructing a past matrix and a future matrix by the time sequence expansion: at a particular sampling instant k, the observation vector y iskExpanding p steps to k times to generate past observation vector
Figure GDA0001941354690000102
Expanding f steps after k to generate future observation vectors
Figure GDA0001941354690000103
Then to yp,k，yf,kCarrying out equalization treatment:
Figure GDA0001941354690000111
wherein: mean (y)p,k) To represent
Figure GDA0001941354690000112
Mean of (y)f,k) To represent
Figure GDA0001941354690000113
Is measured.
The past observation matrix Y is constructed by using all past observation vectors and future observation vectors respectivelypAnd future observation matrix Yf：
Figure GDA0001941354690000114
Where M is N-f-p +1, p, f are two types of time lag parameters, and let p be f, whose value can be determined by the sample autocorrelation function:
Figure GDA0001941354690000115
wherein: autocorr (Y)jP) represents the matrix YpThe autocorrelation coefficient of the jth column vector and its time lag p;
(2.2) construction of Hankel matrix calculation of covariance matrix of past and future matrices ∑pp，∑ffAnd their cross-covariance matrices ∑fpAnd then constructing a Hankel matrix H by utilizing the covariance matrix and the cross covariance matrix:
Figure GDA0001941354690000116
Figure GDA0001941354690000117
(2.3) singular value decomposition: performing singular value decomposition on the Hankel matrix to obtain Jp group typical variable pairs, using (a)i TYp,bi TYf) Representing the i-th set of pairs of typical variables, ai T、bi TRepresents the correlation coefficient between the i-th group of typical variable pairs:
H＝UDVT(6)
Figure GDA0001941354690000121
D＝diag(γ1,γ2,…,γJp)
u and V are singular vectors Ui，viForming an orthogonal matrix, D is a singular value matrix, singular vectors in U, V are only in pairwise correlation, and the correlation value is determined by the corresponding ith singular value gamma in DiAnd (5) characterizing. The larger the singular value (gamma)1>γ2>…>γJp) The greater the correlation between the typical variables.
(2.4) calculating a transformation matrix and extracting a typical variable and a residual variable: truncation matrix
Figure GDA0001941354690000122
The first r column of (2), generate the matrix after dimensionality reduction
Figure GDA0001941354690000123
VrMost of the timing related information is still retained. Wherein, the magnitude of the r value can be determined by the following criteria:
Figure GDA0001941354690000124
cr denotes a criterion value, β denotes a determination threshold, β ═ 0.5.
From VrCalculating the representative variable transformation matrix C and the residual variable transformation matrix L:
Figure GDA0001941354690000125
and obtaining a typical variable space Z and a residual error space E by using the transformation matrix:
Figure GDA0001941354690000126
z, column vector Z in Εk∈r×1，k∈ Jp × 1 each show a typical sampling instant kVariables and residual variables; z, line vector Z in Εt，tTime sequence information of the same variable at different time is contained.
(3) Slow Feature Analysis (SFA) is utilized to respectively extract Slow features s in typical variable space Z and residual space EeZ，sE. To extract slow features s in the typical variable space ZZFor example, the method mainly comprises the following steps:
(3.1) data normalization: the typical variable space Z is normalized according to variables, and the calculation formula is as follows:
Figure GDA0001941354690000131
zttable time sequence vectors, mean (z), of the same variable at different timest) Denotes ztMean value of (a), std (z)t) Denotes ztStandard deviation of (2).
(3.2) the output signal of Z after projection is sZj，sZjDenotes sZThe jth slow signature sequence. In consideration of the linear condition, the method is,
Figure GDA0001941354690000132
representing a coefficient vector, which is equivalent to finding a slow feature signal s extracted from the normalized input signal ZZ＝[sZ1 T，sZ2 T，…，sZr T]TIs converted into a matrix
Figure GDA0001941354690000133
I.e. sZ＝WZAnd Z. Slow characteristic signal sZjThe objective function and the constraint condition to be satisfied are as follows:
an objective function:
Figure GDA0001941354690000134
the constraint conditions are as follows:
Figure GDA0001941354690000135
wherein:
Figure GDA0001941354690000136
signal s representing a slow characteristicZIs calculated by time-sequence difference<·>Is shown as
Figure GDA0001941354690000137
t1，t0Respectively representing the upper and lower time limits.
(3.3) whitening: covariance matrix of input data using singular value decomposition<ZZT>The whitening treatment can remove the correlation in the data, so that the extracted slow characteristic value carries different information:
Figure GDA0001941354690000141
Figure GDA0001941354690000142
wherein ΛZ -1/2BTBeing a whitening matrix, OZIs the corresponding whitened input signal.
(3.4) calculating the transformation matrix WZ: to the input matrix OZDifferential processing is carried out to obtain a time sequence differential signal
Figure GDA0001941354690000143
Can prove that
Figure GDA0001941354690000144
Covariance matrix of
Figure GDA0001941354690000145
After singular value decomposition, a series of singular values omega are obtainedZjThe objective function value described in equation (12)
Figure GDA0001941354690000146
Figure GDA0001941354690000147
WZ＝PΛZ -1/2BT(17)
The slow features s in the residual space ΕEAnd the slow characteristic s in the above typical variable space ZZThe extraction method is the same.
(4) Partitioning slow features sZ: the slowest feature corresponds to the minimum feature value, the feature values are arranged from small to large, and the first features are divided into s according to the size of the feature valuesZOf slower changing features, by sZ,dRepresents; dividing the last (r-l) features into sZFeature of faster change by sZ,eAnd (4) showing. The determination method of the division basis l is that firstly, a slow characteristic value s is utilizedZIs used to represent the process variable
Figure GDA0001941354690000148
The speed of change of (2):
Figure GDA0001941354690000149
wherein: r isjiIs a matrix RZElement of row j and column i, sZiRepresents the ith slow signature sequence, and Δ (-) represents an operation to calculate how slowly the sequence changes:
Figure GDA00019413546900001410
dividing the extracted features with slowness larger than the slowness of the input data in the slow feature values into fast features, wherein the fast features have M in totaleOne such fast feature:
Figure GDA0001941354690000151
here card {. is } represents the number of elements in the set {. is }. M determined according to equation (19)eValue, corresponding to the matrix omegaZIt is also divided into two parts:
Figure GDA0001941354690000152
Figure GDA0001941354690000153
(5) calculating dynamic monitoring indexes: starting from the first sample point of the typical variable space, each sample point can obtain a set of dynamic monitoring indexes (S)Z,d 2，SZ,e 2)。
Figure GDA0001941354690000154
(6) Determining a control limit based on the dynamic monitoring indicator: by using the method of nuclear density estimation, a dynamic monitoring index S is estimated firstlyZ,d 2For a given significance level α, SZ,d 2Control limit of
Figure GDA0001941354690000159
The calculation method is as follows:
Figure GDA0001941354690000155
s can be calculated in the same wayZ,e 2Control limit of
Figure GDA0001941354690000156
(7) According to the method from the step (3) to the step (6), extracting the slow features s of the residual space EEEAnd will sEDivided into two parts sE,d，sE,eEstablishing a monitoring index SE,d 2，SE,e 2And calculating a control limit
Figure GDA0001941354690000157
Figure GDA0001941354690000158
The processing method is the same as that of the typical variable space Z, and the description is omitted.
(8) And (3) online monitoring and controlling performance: and (4) on the basis of the CVA-SFA model established in the steps (2) to (4) and the four monitoring statistics obtained in the steps (5) to (7), the performance state of the steam turbine control system can be monitored on line. This step is realized by the following substeps:
(8.1) acquiring new online data and preprocessing the new data: collecting a new section of observation data
Figure GDA0001941354690000161
Thereafter, where the following Table new represents the new observed data, Y is first added according to step (2)newDeveloping into a past matrix, and normalizing the past matrix according to the mean value and the standard deviation obtained in the step (2) to obtain Ypnew. In the example, the new data are divided into two parts, wherein the first data are data collected under normal working conditions, the sampling period is 10 minutes, 2271 samples are totally obtained, 60 process variables are obtained, the second data are data recorded under abnormal working conditions, the sampling period is 10 minutes, 2563 samples are totally obtained, 60 process variables are obtained, and the measured variables are flow, vibration, temperature, rotating speed, current and the like in the operation process of the steam turbine;
(8.2) extracting typical variables and residual variables of the new observed data: after the standardization processing, the conversion matrix V determined in the step (2) is utilizedrAnd L calculating the typical variable space Z of the new observed datanewAnd residual space Enew。
(8.3) extracting the typical variable space Z of the new observed datanewSlow characteristics of (1): firstly, according to the mean value and variance pair Z determined in the step (3.1)newPerforming normalization processing, and then using the slow feature transformation matrix W determined in step (3.4)ZExtracting the normalized ZnewSlow characteristic s ofZnewAnd dividing s according to the previous division parameterZnewIs divided intoZ,d newAnd sZ,e newAlso according to WECan obtain EnewFurther obtaining sE,d newAnd sE,e new。
(8.4) calculating a new monitoring statistical index: calculating a monitoring statistical index in a typical variable space according to the established model and the calculation method determined in the steps (5) and (7)
Figure GDA0001941354690000162
Sum residual space monitoring index
Figure GDA0001941354690000163
Figure GDA0001941354690000171
(8.5) judging the control performance state of the steam turbine on line, namely comparing four monitoring indexes with respective statistical control limits in real time, if the four monitoring indexes are all positioned within the statistical control limits, indicating that the control system works normally, if one or more monitoring indexes exceed the normal control limits, indicating that an abnormal condition occurs in the control system, in the step (a) of figure 2, only the statistics of individual points in four groups of statistics and corresponding control lines exceed the control lines, and under the condition that the confidence level α is 0.05, the new working condition data can be considered to be normal, namely the control system is shown to be normal, in the step (b) of figure 2, four groups of statistics SZ,d 2，SZ,e 2，SE,d 2，SE,e 2Are obviously beyond the threshold line in the four sections 1330, 1430, 1808, 1825, 1943, 1972 and 2093, 2128, the statistic QkAfter the first obvious overrun at about 326 th sampling point, the overrun state is maintained, so that the control system is judged to be abnormal in the periods, and a proper fault diagnosis method, such as a contribution diagram method, can be adopted to analyze and isolate possible fault variables.
The invention extracts the relevant information among the variables of the steam turbine control system by using typical variable analysis, and extracts the dynamic characteristics in the relevant information by using a slow characteristic analysis algorithm, and the characteristics extracted by the method can reflect the regulation function of the controller. Finally, a steam turbine control performance online monitoring model is constructed by combining the correlation of variables and the information of the speed of change, the method solves the problem that a large steam turbine is difficult to monitor control performance due to numerous variables and working condition changes, the accuracy of dynamic process control performance online monitoring is greatly improved, a steam turbine system is effectively and timely monitored by a thermal power plant, the safe and reliable operation of a large thermal generator set is guaranteed, and meanwhile the production requirement for improving the production benefit is met.

Claims (1)

1. A real-time monitoring method for control performance of a steam turbine in high-end power generation equipment is characterized by comprising the following steps:
(1) acquiring training data, wherein a control system of the steam turbine is provided with J measurement variables and operation variables, and an observation vector y of J × 1 is obtained by sampling each timekWherein the subscript k is a time index, and data obtained after N times of sampling is expressed as a two-dimensional observation matrix
Figure FDA0002503688500000011
The measurement variables are state parameters which can be measured in the operation process of the steam turbine and comprise bearing metal temperature, bearing axial vibration, generator active power, steam turbine rotating speed and excitation three-phase temperature; the operation variables comprise condensate flow, water supply pressure, water supply flow and coal supply quantity of a coal feeder; the training data should select the sampling data of the steam turbine in the normal operation state;
(2) extracting time sequence related information of data by using a CVA algorithm, wherein the step is realized by the following substeps:
(2.1) constructing a past matrix and a future matrix by the time sequence expansion: at a particular sampling instant k, the observation vector y iskExpanding p steps to k times to generate past observation vector
Figure FDA0002503688500000012
Expanding f steps after k to generate future observation vectors
Figure FDA0002503688500000013
Then to yp,k，yf,kCarrying out equalization treatment:
Figure FDA0002503688500000014
wherein: mean (y)p,k) To represent
Figure FDA0002503688500000015
Mean of (y)f,k) To represent
Figure FDA0002503688500000016
The mean value of (a);
the past observation matrix Y is constructed by using all past observation vectors and future observation vectors respectivelypAnd future observation matrix Yf：
Figure FDA0002503688500000017
Where M is N-f-p +1, p, f are two types of time lag parameters, let p be f, whose value passes through the sample autocorrelation function ap+1To determine:
Figure FDA0002503688500000021
wherein: autocorr (Y)jP) represents the matrix YpThe autocorrelation coefficient of the jth column vector and its time lag p;
(2.2) construction of Hankel matrix calculation of covariance matrix of past and future matrices ∑pp，∑ffAnd their cross-covariance matrices ∑fpAnd then constructing a Hankel matrix H by utilizing the covariance matrix and the cross covariance matrix:
Figure FDA0002503688500000022
Figure FDA0002503688500000023
Figure FDA0002503688500000024
(2.3) singular value decomposition: performing singular value decomposition on the Hankel matrix to obtain Jp group typical variable pairs, and using ai TYp,bi TYfRepresenting the i-th set of pairs of typical variables, ai T、bi TRepresents the correlation coefficient between the i-th group of typical variable pairs:
H＝UDVT(6)
Figure FDA0002503688500000025
u and V are singular vectors Ui，viForming an orthogonal matrix, D is a singular value matrix, singular vectors in U, V are only in pairwise correlation, and the correlation value is determined by the corresponding ith singular value gamma in DiCharacterizing; the larger the singular value, the greater the correlation between the typical variables;
(2.4) calculating a transformation matrix and extracting a typical variable and a residual variable: truncation matrix
Figure FDA0002503688500000031
The first r column of (2), generate the matrix after dimensionality reduction
Figure FDA0002503688500000032
VrMost of the timing related information is still kept; wherein the magnitude of the r value is determined by the following criteria:
Figure FDA0002503688500000033
cr represents a criterion value, β is a judgment threshold value, β is 0.5;
from VrCalculating the representative variable transformation matrix C and the residual variable transformation matrix L:
Figure FDA0002503688500000034
and then obtaining a typical variable space Z and a residual error space E by using a transformation matrix:
Figure FDA0002503688500000035
z, column vector Z in Εk∈r×1，k∈ Jp × 1 denotes the typical and residual variables at the sampling instant k, respectively, Z, the line vector Z in Et，tThe time sequence information of the same variable at different moments is contained;
(3) slow features s in typical variable space Z and residual error space Ee are respectively extracted by using a slow feature analysis algorithmZ，sESlow features s in the canonical variable space ZZThe extraction method comprises the following steps:
(3.1) data normalization: the typical variable space Z is normalized according to variables, and the calculation formula is as follows:
Figure FDA0002503688500000036
zttable time sequence vectors, mean (z), of the same variable at different timest) Denotes ztMean value of (a), std (z)t) Denotes ztStandard deviation of (d);
(3.2) the output signal of Z after projection is sZj，sZjDenotes sZThe jth slow signature sequence; in consideration of the linear condition, the method is,
Figure FDA0002503688500000041
representing a coefficient vector, which is equivalent to finding a slow feature signal s extracted from the normalized input signal ZZ＝psZ1 T，sZ2 T，…，sZr T]TIs converted into a matrix
Figure FDA0002503688500000042
I.e. sZ＝WZZ; slow characteristic signal sZjThe objective function and the constraint condition to be satisfied are as follows:
an objective function:
Figure FDA0002503688500000043
the constraint conditions are as follows:
Figure FDA0002503688500000044
wherein:
Figure FDA0002503688500000045
signal s representing a slow characteristicZIs calculated and expressed as
Figure FDA0002503688500000046
t1，t0Respectively representing the upper and lower time limits;
(3.3) whitening: covariance matrix of input data using singular value decomposition<ZZT>Whitening processing is carried out to remove the correlation in the data, so that the extracted slow characteristic value carries different information:
Figure FDA0002503688500000047
Figure FDA0002503688500000048
wherein ΛZ -1/2BTBeing a whitening matrix, OZIs the corresponding whitened input signal;
(3.4) calculating the transformation matrix WZ: to the input matrix OZDifferential processing is carried out to obtain a time sequence differential signal
Figure FDA0002503688500000051
Prove that is to
Figure FDA0002503688500000052
Covariance matrix of
Figure FDA0002503688500000053
After singular value decomposition, a series of singular values omega are obtainedZjThe objective function value described in equation (12)
Figure FDA0002503688500000054
Figure FDA0002503688500000055
WZ＝PΛZ -1/2BT(17)
The slow features s in the residual space ΕEAnd the slow characteristic s in the above typical variable space ZZThe extraction method is the same;
(4) partitioning slow features sZ: the slowest feature corresponds to the minimum feature value, the feature values are arranged from small to large, and the first features are divided into s according to the size of the feature valuesZOf slower changing features, by sZ,dRepresents; dividing the last (r-l) features into sZFeature of faster change by sZ,eRepresents; the determination method of the division basis l is that firstly, a slow characteristic value s is utilizedZIs used to represent the process variable
Figure FDA0002503688500000056
The speed of change of (2):
Figure FDA0002503688500000057
wherein: r isjiIs a matrix RZElement of row j and column i, sZiIndicating the ith slow signature sequenceColumn, Δ (-) represents an operation to calculate how slowly the sequence changes:
Figure FDA0002503688500000058
dividing the extracted features with slowness larger than the slowness of the input data in the slow feature values into fast features, wherein the fast features have M in totaleOne such fast feature:
Figure FDA0002503688500000059
here card {. } represents the number of elements in the set {. }; m determined according to equation (19)eValue, corresponding to the matrix omegaZIt is also divided into two parts:
Figure FDA0002503688500000061
Figure FDA0002503688500000062
(5) calculating dynamic monitoring indexes: starting from the first sample point of the typical variable space, each sample point obtains a set of dynamic monitoring indexes SZ,d 2，SZ,e 2；
Figure FDA0002503688500000063
(6) Determining a control limit based on the dynamic monitoring indicator: by using the method of nuclear density estimation, a dynamic monitoring index S is estimated firstlyZ,d 2For a given significance level α, SZ,d 2Control limit of
Figure FDA0002503688500000064
The calculation method is as follows:
Figure FDA0002503688500000065
s is calculated in the same mannerZ,e 2Control limit of
Figure FDA0002503688500000066
(7) According to the method from the step (3) to the step (6), extracting the slow features s of the residual space EEEAnd will sEDivided into two parts sE,d，sE,eEstablishing a monitoring index SE,d 2，SE,e 2And calculating a control limit
Figure FDA0002503688500000067
Figure FDA0002503688500000068
(8) And (3) online monitoring and controlling performance: on-line monitoring the performance state of the steam turbine control system based on the formulas established in the steps (2) to (4) and the four monitoring statistics obtained in the steps (5) to (7), wherein the step is realized by the following substeps:
(8.1) acquiring new online data and preprocessing the new data: collecting a new section of observation data
Figure FDA0002503688500000069
Thereafter, where the subscript new denotes the new observed data, Y is first placed according to step (2)newDeveloping into a past matrix, and normalizing the past matrix according to the mean value and the standard deviation obtained in the step (2) to obtain Ypnew；
(8.2) extracting typical variables and residual variables of the new observed data: after the standardization processing, the conversion matrix V determined in the step (2) is utilizedrAnd L calculating the typical variable space Z of the new observed datanewAnd residual space Enew；
(8.3) extracting the typical variable space Z of the new observed datanewSlow characteristics of (1): firstly according to the mean value and the square determined in the step (3.1)Difference pair ZnewPerforming normalization processing, and then using the slow feature transformation matrix W determined in step (3.4)ZExtracting the normalized ZnewSlow characteristic s ofZnewAnd dividing s according to the previous division parameterZnewIs divided intoZ,d newAnd sZ,e newAlso according to WETo obtain EnewFurther obtaining sE,d newAnd sE,e new；
(8.4) calculating a new monitoring statistical index: calculating a monitoring statistical index in a typical variable space according to the established model and the calculation method determined in the steps (5) and (7)
Figure FDA0002503688500000071
Sum residual space monitoring index
Figure FDA0002503688500000072
Figure FDA0002503688500000073
(8.5) online judging the control performance state of the steam turbine: comparing the four monitoring indexes with respective statistical control limits in real time, and if the four monitoring indexes are all located within the statistical control limits, indicating that the control system works normally; if one or more monitoring indexes exceed the normal control limit, the abnormal condition of the control system is indicated.
CN201811110373.9A 2018-09-21 2018-09-21 Real-time monitoring method for control performance of steam turbine in high-end power generation equipment Active CN109538311B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811110373.9A CN109538311B (en) 2018-09-21 2018-09-21 Real-time monitoring method for control performance of steam turbine in high-end power generation equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811110373.9A CN109538311B (en) 2018-09-21 2018-09-21 Real-time monitoring method for control performance of steam turbine in high-end power generation equipment

Publications (2)

Publication Number Publication Date
CN109538311A CN109538311A (en) 2019-03-29
CN109538311B true CN109538311B (en) 2020-08-04

Family

ID=65843543

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811110373.9A Active CN109538311B (en) 2018-09-21 2018-09-21 Real-time monitoring method for control performance of steam turbine in high-end power generation equipment

Country Status (1)

Country Link
CN (1) CN109538311B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111188761A (en) * 2019-12-31 2020-05-22 杭州哲达科技股份有限公司 Monitoring method for pump equipment based on Fourier-CVA model

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111474920B (en) * 2020-04-29 2021-04-06 电子科技大学 ICA and ACVA-based fault detection method
CN113065583B (en) * 2021-03-13 2023-11-14 宁波大学科学技术学院 Rectification process abnormity monitoring method based on online nonlinear discriminant feature analysis
CN114396317B (en) * 2021-12-01 2022-12-16 上海发电设备成套设计研究院有限责任公司 Multi-target multi-dimensional online combined monitoring method and system for nuclear turbine

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1494020A1 (en) * 2003-06-30 2005-01-05 Siemens Westinghouse Power Corporation Method and apparatus for measuring on line failure of turbine thermal barrier coatings
US20160033580A1 (en) * 2012-05-29 2016-02-04 Board Of Regents Of The University Of Nebraska Detecting Faults in Turbine Generators
CN106097151A (en) * 2016-06-24 2016-11-09 清华大学 A kind of method reducing power plant's data uncertainty based on data harmonization
CN106680012A (en) * 2017-01-25 2017-05-17 浙江大学 Fault detection method and fault diagnosis method for the non-stationary process of large coal-fired generator set
CN107766457A (en) * 2017-09-27 2018-03-06 华能国际电力股份有限公司上海石洞口第二电厂 A kind of thermal power plant's operation management system and its task executing method
US20180171774A1 (en) * 2016-12-21 2018-06-21 Schlumberger Technology Corporation Drillstring sticking management framework
CN108334674A (en) * 2018-01-17 2018-07-27 浙江大学 A kind of steam turbine high-pressure cylinder method for monitoring operation states based on parameter association intellectual analysis

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1494020A1 (en) * 2003-06-30 2005-01-05 Siemens Westinghouse Power Corporation Method and apparatus for measuring on line failure of turbine thermal barrier coatings
US20160033580A1 (en) * 2012-05-29 2016-02-04 Board Of Regents Of The University Of Nebraska Detecting Faults in Turbine Generators
CN106097151A (en) * 2016-06-24 2016-11-09 清华大学 A kind of method reducing power plant's data uncertainty based on data harmonization
US20180171774A1 (en) * 2016-12-21 2018-06-21 Schlumberger Technology Corporation Drillstring sticking management framework
CN106680012A (en) * 2017-01-25 2017-05-17 浙江大学 Fault detection method and fault diagnosis method for the non-stationary process of large coal-fired generator set
CN107766457A (en) * 2017-09-27 2018-03-06 华能国际电力股份有限公司上海石洞口第二电厂 A kind of thermal power plant's operation management system and its task executing method
CN108334674A (en) * 2018-01-17 2018-07-27 浙江大学 A kind of steam turbine high-pressure cylinder method for monitoring operation states based on parameter association intellectual analysis

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111188761A (en) * 2019-12-31 2020-05-22 杭州哲达科技股份有限公司 Monitoring method for pump equipment based on Fourier-CVA model

Also Published As

Publication number Publication date
CN109538311A (en) 2019-03-29

Similar Documents

Publication Publication Date Title
CN109491358B (en) Control performance monitoring method for boiler dynamic information of million-kilowatt ultra-supercritical unit
CN109538311B (en) Real-time monitoring method for control performance of steam turbine in high-end power generation equipment
CN109471420B (en) CVA-SFA-based method for monitoring control performance of air preheater of large coal-fired power generator set of intelligent power plant
CN110262450B (en) Fault prediction method for cooperative analysis of multiple fault characteristics of steam turbine
CN108446529B (en) Organic Rankine cycle system fault detection method based on generalized mutual entropy-DPCA algorithm
CN112179655B (en) Turbo generator fault early warning method based on threshold classification
US20040215353A1 (en) Process and device for evaluating the performance of a process control system
CN111159844B (en) Abnormity detection method for exhaust temperature of gas turbine of power station
CN112836941B (en) Online health condition assessment method for high-pressure system of steam turbine of coal motor unit
CN115730191A (en) Attention mechanism-based coal mill fault early warning method
CN114116832A (en) Power distribution network abnormity identification method based on data driving
CN109270917B (en) Intelligent power plant steam turbine bearing-oriented closed-loop control system fault degradation state prediction method
CN102136038A (en) State space-based multi-featured device state evaluation method and application
CN109184821B (en) Online monitoring method for closed-loop information analysis of steam turbine of large-scale generator set
CN109283912B (en) Distributed dynamic and static cooperative comprehensive monitoring method for pulverizing system of large coal-fired generator set of intelligent power plant
CN114326486B (en) Process monitoring method based on probability slow feature analysis and elastic weight consolidation
CN111949003B (en) Closed-loop control loop performance evaluation method based on SFA and Hellinger distance
Xinyi et al. Fault diagnosis in chemical process based on self-organizing map integrated with fisher discriminant analysis
CN105279553A (en) Method for identifying fault degree of high-pressure heater water supply system
CN115130564A (en) High-pressure heater online monitoring method based on parallel GMM-LCKSVD
Wang et al. Power plant data filtering based on Gaussian naive Bayesian classification and prediction error method
CN111695581B (en) Nonlinear dynamic process monitoring method based on autoregressive extreme learning machine
CN113298133A (en) Supercritical unit boiler tube burst fault diagnosis method
Ma et al. Process monitoring of the pneumatic control valve using canonical variate analysis
Li et al. Physics-Guided Deep Learning Method for Tool Condition Monitoring in Smart Machining 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
GR01 Patent grant
GR01 Patent grant