CN104865944B - Gas separation unit control system performance estimating method based on PCA LSSVM - Google Patents
Gas separation unit control system performance estimating method based on PCA LSSVM Download PDFInfo
- Publication number
- CN104865944B CN104865944B CN201410342442.4A CN201410342442A CN104865944B CN 104865944 B CN104865944 B CN 104865944B CN 201410342442 A CN201410342442 A CN 201410342442A CN 104865944 B CN104865944 B CN 104865944B
- Authority
- CN
- China
- Prior art keywords
- mrow
- control system
- msub
- performance
- data
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 64
- 238000000926 separation method Methods 0.000 title claims abstract description 19
- 238000011156 evaluation Methods 0.000 claims abstract description 33
- 239000011159 matrix material Substances 0.000 claims abstract description 29
- 238000004458 analytical method Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 9
- 238000004519 manufacturing process Methods 0.000 claims description 6
- 238000010845 search algorithm Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 4
- 238000009825 accumulation Methods 0.000 claims description 3
- 238000004886 process control Methods 0.000 claims description 2
- 230000009897 systematic effect Effects 0.000 claims description 2
- 239000007789 gas Substances 0.000 description 22
- 238000005194 fractionation Methods 0.000 description 6
- 238000005457 optimization Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000001186 cumulative effect Effects 0.000 description 3
- 239000003915 liquefied petroleum gas Substances 0.000 description 3
- 239000002994 raw material Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- -1 A-1It is existing Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 241001269238 Data Species 0.000 description 1
- 241000406668 Loxodonta cyclotis Species 0.000 description 1
- 239000004743 Polypropylene Substances 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000009835 boiling Methods 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 239000007788 liquid Substances 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- NJPPVKZQTLUDBO-UHFFFAOYSA-N novaluron Chemical compound C1=C(Cl)C(OC(F)(F)C(OC(F)(F)F)F)=CC=C1NC(=O)NC(=O)C1=C(F)C=CC=C1F NJPPVKZQTLUDBO-UHFFFAOYSA-N 0.000 description 1
- 229920001155 polypropylene Polymers 0.000 description 1
- 238000001556 precipitation Methods 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000010992 reflux Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- URWAJWIAIPFPJE-YFMIWBNJSA-N sisomycin Chemical compound O1C[C@@](O)(C)[C@H](NC)[C@@H](O)[C@H]1O[C@@H]1[C@@H](O)[C@H](O[C@@H]2[C@@H](CC=C(CN)O2)N)[C@@H](N)C[C@H]1N URWAJWIAIPFPJE-YFMIWBNJSA-N 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 210000002700 urine Anatomy 0.000 description 1
- 238000005406 washing Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B23/00—Testing or monitoring of control systems or parts thereof
- G05B23/02—Electric testing or monitoring
- G05B23/0205—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults
- G05B23/0218—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults
- G05B23/0243—Electric testing or monitoring by means of a monitoring system capable of detecting and responding to faults characterised by the fault detection method dealing with either existing or incipient faults model based detection method, e.g. first-principles knowledge model
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Automation & Control Theory (AREA)
- Feedback Control In General (AREA)
Abstract
The present invention relates to the gas separation unit control system performance estimating method based on PCA LSSVM, comprise the following steps:Dimension-reduction treatment is carried out to data using PCA methods;The Principal component of acquisition is trained and modeled by least square method supporting vector machine;The value that performance indications are calculated by the model of gained is used as assessment result.The present invention establishes system model using PCA LSSVM methods, and the Performance Evaluation of system is controlled using the model, so as to the incidence matrix without asking for control system, reduces computational complexity, improves estimating velocity.
Description
Technical field
It is more more particularly to gas fractionation unit the present invention relates to refinery gas's fractionating device control system Performance Evaluation
Variable control system performance estimating method, belong to control system Performance Evaluation field.
Background technology
The main task of refinery gas's fractionating device is:The various components in liquefied petroleum gas are isolated, are follow-up gas
Body deep processing device provides raw material.The system of rectifying tower that gas fractionation unit is made up of multiple rectifying columns, make use of liquefied petroleum gas
The boiling point of middle each component is different with saturated vapor pressure, and the separation of each component is realized using polynary rectificating method, so as to be follow-up
Device provides raw material.
Gas fractionation unit includes many control loops, and it is an arduous times to ensure that these loops are run long-term effectively
Business, and it is to ensure the important means of its normal operation that Performance Evaluation and monitoring are carried out to system.Gas separation unit control loop is most
It is multivariable, therefore the performance estimating method of Study of Multivariable control system has long-term interest.But currently for more
In terms of the theoretical research of variable control system Performance Evaluation and practical application, or the task of a challenge.
Control system performance evaluation is to maintain an important technology of control loop Effec-tive Function in industrial process.It can borrow
A kind of software of on-line operation is helped, performance evaluation is carried out by the analysis to service data.And then whether analyzer-controller is " strong
Health ", judges whether the performance of control loop has much room for improvement, and its performance evaluation should include following several Main Stages:
It is determined that operation in control system ability:Including current quantization performance.Analysis dynamic, which can survey data and calculate, to be worked as
The performance indications of preceding control system.
One suitable performance evaluation benchmark of selection and design:The stage specifies evaluation base for the control loop of operation
Standard, can be historical data benchmark or some other user-defined benchmark etc..
The bad control loop monitoring and evaluation of performance:This step need to examine current control loop performance with it is selected
Assess the deviation between benchmark.More importantly to find the possibility whether current control loop performance has much room for improvement.Work as control
When the potential in loop processed is not excavated sufficiently, subsequent diagnostic phases are considered.
The diagnosis of potential cause:If the evaluation indicate that the performance of control loop and benchmark deviation are very big, then with regard to needing to look for
Be out of order Producing reason.
Improve the suggestion of performance measures:After being diagnosed to be the reason for causing control loop degradation, it is necessary to take corresponding
Measure makes control loop performance recovery to initial kilter.
Existing many scholars propose the performance estimating method of a variety of multivariable control systems in succession at present, but apply most
More or minimum variance (MVC) method of evaluating performance.Because time delay is had expanded into as incidence matrix in multi-variable system,
, it is necessary to the completely information of process model and incidence matrix, and to association when being controlled system performance evaluation using MVC methods
Substantial amounts of calculating is needed when matrix is solved, it can thus be appreciated that this method amount of calculation is very big.It is proposed that one practical
Suboptimum multivariable MVC controls benchmark, and it is the natural extending of SISO System Performance Analysis technologies, it is only necessary to the rank of incidence matrix
It is secondary, it is more simple and practical than general MVC pedestal methods without constructing incidence matrix, but accurate performance can not be provided and referred to
Scale value.
The content of the invention
For above-mentioned technical deficiency, the present invention is proposed based on pivot analysis (Principle Component
Analysis, abbreviation PCA) and least square method supporting vector machine (Least Squares Support Vector Machine, letter
Claim LSSVM) Performance Evaluation Algorithm.Initial data is first carried out dimension-reduction treatment by the method by PCA methods, then the new of acquisition
Principal component carries out least square method supporting vector machine training and modeling, then calculates control system performance by its model again and refers to
Mark.Calculation error is this approach reduce, more practical Performance Evaluating Indexes can be obtained and be not related to incidence matrix in the calculation
Problem, greatly reduce the complexity of calculating.
The technical solution adopted for the present invention to solve the technical problems is:Gas separation unit control system based on PCA-LSSVM
System performance estimating method, comprises the following steps:
Dimension-reduction treatment is carried out to data using PCA methods;
The Principal component of acquisition is trained and modeled by least square method supporting vector machine;
Performance index value is obtained by the model of gained and is used as assessment result.
It is described that data progress dimension-reduction treatment is comprised the following steps using PCA methods:
The real time data that (2-1) chooses gas separation unit control system forms raw sample data Xn×m, n is data length, m
For input variable number, by Xn×mIt is standardized using following formula:
Wherein, xijRepresent Xn×mData,For the average value of jth item variable,
For the standard deviation of jth item variable;
(2-2) asks for the covariance matrix of the matrix, draws output data and input by obtaining new matrix after standardization
Correlation matrix R between data, Σ=diag (λ are obtained using singular value decomposition method to correlation matrix R1,λ2,...,λn), enter
And obtain the variance of each pivotWith total varianceWherein δiFor variance, λiIt is characterized
Value, n is data length;
(2-3) calculates accumulation contribution rate by following formula:
The pivot number k that CPV (k) is more than set-point is found out, obtains the Principal component of k dimensions.
The Principal component acquisition, which is trained and modeled by least square method supporting vector machine, to be comprised the following steps:
(3-1) selects RBF as kernel function, and finds regularization parameter γ and kernel function width ε by three step search algorithm;
(3-2) establishes Lagrangian, obtains Lagrange multiplier α and amount of bias β, and substituted into control system
Mathematical modeling:
Wherein K (xi,xj) it is kernel function, α ∈ αi;
(3-3) obtains the variance of system according to mathematical modeling:
Wherein, n is data length, and x and y are respectively input sample and output sample;
(3-4) is according to the systematic variance obtained, calculation of performance indicatorsWhereinFor the expectation of historical data
Output variance.
The three step search algorithm comprises the following steps:
(4-1) arbitrarily interception γ and ε one section of section, structure parameter is to [γ, ε];
(4-2) finds central point and its 8 points of surrounding on the two dimensional surface that γ and ε is formed;According to mean square error most
Small criterion selects a point as new central point;
(4-3) judges whether step-length is less than a unit;If it is not, step-length is then reduced into half, return to step (4-
2);If it is less, parameter γ, ε corresponding to the central point is parameters obtained.
The invention has the advantages that and advantage:
1. the present invention establishes gas fractionation unit multivariable control system model using PCA-LSSVM methods, the mould is used
Type is controlled the Performance Evaluation of system, so as to the incidence matrix without asking for control system, reduces computational complexity, is lifted
Estimating velocity.
2. the performance indications error that the present invention is drawn is small, close to true a reference value.The performance indications calculated can be anti-
The actual performance of gas separation unit control system is reflected, the evaluation result drawn is more accurate, reliable.It can be the safety in production at scene
Provide safeguard, and effectively utilize raw material, produce more qualified product, and then considerable economic effect is brought for enterprise
Benefit.
Brief description of the drawings
Fig. 1 is control system Performance Evaluation principle framework figure;
Fig. 2 is the gas separation unit control system overall framework figure of the present invention;
Fig. 3 is flow chart of the method for the present invention;
Fig. 4 is the method schematic of the present invention;
White noise acoustic jamming schematic diagram when Fig. 5 is the operation of system application MATLAB software emulations;
Fig. 6 exports Y when being the operation of system application MATLAB software emulations1(temperature) signal schematic representation;
Fig. 7 exports Y when being the operation of system application MATLAB software emulations2(temperature) signal schematic representation;
Fig. 8 is output Y1(temperature) performance evaluation result curve;
Fig. 9 is output Y2(temperature) performance evaluation result curve.
Embodiment
With reference to embodiment, the present invention is described in further detail.
The invention provides a kind of gas separation unit multivariable control system performance estimating method of PCA-LSSVM technologies, such as
Shown in Fig. 1-2:
Understand gas separation unit industrial process, the main control loop of searching system, carried out for these process control loops
Analysis, analyzer-controller structure and characteristic, understand the history optimum state in loop, and need to ensure these loops in current work
Steady under condition, safety operation, powerful guarantee is provided to calculate optimal performance.
Live real time data caused by gas separation unit industrial processes operation passes through Distributed Control System
(Distributed Control System, abbreviation DCS) collects central control room, and selected loop uses PID control
Device, the input of PID controller is using loop settings value r and live real time data y deviation, and PID output u is as control pair
The input of elephant, Field adjustment device is sent to through Distributed Control System, makes to reach control purpose;The process instantaneous value collected is made
For the significant data source of control system Performance Evaluation.
The process instantaneous value collected is subjected to online evaluation using appraisal procedure, assessment result forming curves are in computer
Shown on interface, and the best curve of history is contrasted, and then field engineer can intuitively find out evaluation effect.Scene
Engineering staff according to evaluation result, be controlled think highly of it is new adjust and search other reasons in time, it is efficient with Guarantee control system
Operation.
As shown in Figure 3-4, the step of gas separation unit control system performance estimating method based on PCA-LSSVM is as follows:
1. for gas separation unit multivariable control system, industry spot process real time data is obtained first.Input and output number
According to being multivariable, data are using the actual temperature of industry spot gas separation unit, pressure, flow etc..
Using data matrix as raw sample data Xn×m(data length is n rows, and input variable number arranges for m) is simultaneously carried out
Standardization:The process data of collection is standardized first, to eliminate the unreasonable influence that dimension different band is come, i.e.,:
Wherein:xijRepresent Xn×mData,For the average value of jth item variable,
For the standard deviation of jth item variable;I, j is respectively the line number and columns of matrix, i=1, and 2 ..., n, j=1,2 ..., m.
Gained standardized value is used for the correlation matrix R between the data after normalized processing, by obtaining at standardization
The covariance matrix of the new matrix obtained after reason obtains, and studies dependent variable (output data Y on the wholen=(y1,y2,...,
yn)) and independent variable (input data Xn×m) between correlation
2. calculate Principal component:Original is calculated using singular value decomposition method is simplified to the correlation matrix R tried to achieve after standardization
The Principal component of beginning sample
That is R=A Σ MT, wherein A and M are two orthogonal matrixes that singular vector is formed, i.e. Σ=diag (λ1,λ2,...,
λn), λ1,λ2,...,λnFor the characteristic value of matrix.Then, the variance of each pivot is calculatedThen total side is calculated
DifferenceFinally, cumulative variance percentage (CPV (0~100%)) is calculated:
3. the pivot number k that all cumulative variance percentages are more than set-point (80%) is found out, so as to obtain pivot, so
N can be tieed up by input data by pivot analysis and be changed into k dimension Principal components.
Data compression and information extraction are carried out by using pca method, eliminates the correlation between variable, then
The principal component of extraction is trained using least square method supporting vector machine, obtains its system model.
4. before vector machine training is supported, it is necessary first to which the form of selected kernel function, the present invention select radial direction base
Kernel function (Radial Basis Function, abbreviation RBF), after selected kernel function, also to select suitable regularization parameter
γ and kernel function width ε.The two parameters largely determine the study of least square method supporting vector machine and extensive energy
Power.The present invention selects the two parameters using three step search algorithm, and this method is simple and easy, is not reducing the feelings of forecast accuracy
Under condition, number of comparisons can be significantly reduced, improves the optimum choice of parameter, it is hereby achieved that accurate modeling result.
The step of three step search algorithm, is as follows:
1. arbitrarily interception γ and ε a bit of section, structure parameter is to [γ, ε];
2. on the two dimensional surface formed on the γ and ε of interception section, find central point, and 8 around it
Point.According to mean square error (Mean Square Error, abbreviation MSE) minimum criteria, (minimum mean-squared error criterion:Corresponding data
Later square of difference is sought, all square values are added so that this and minimum), selection wherein (mean square error minimum) point
As new central point;
3. step-length is reduced into half, then scan for, calculate 8 points around new central point, select the minimum points of MSE
As new central point;
4. repeating calculating 3., until step-length is less than a unit, (unit is 4 grids, and the data after dimensionality reduction are pressed
Ranks form grid), then the parameter pair by this group of parameter [γ, ε] alternatively.
5. utilize nonlinear function, the raw sample data collection after standardization from input control space reflection to feature it is empty
Between, the nonlinear fitting problem in the input space is changed into the linear fit problem in high-dimensional feature space.According to structure risk
Minimization principle determines decision function parameter w, β, and above-mentioned fitting problems can be converted into constrained optimization problem:
s.t.:yi=ωT·φ(xi)+β+ζi, i=1,2 ..., k
Wherein, ω is weight function, ζiFor coefficient of relaxation, c and β are respectively penalty factor and amount of bias, φ (xi) it is non-linear
Mapping function, yiExported for sample, k is the dimension of Principal component.
Constrained optimization problem is changed into unconstrained optimization problem, establishes Lagrangian.Solve above-mentioned optimization problem
It can be exchanged into solution linear equation:
Wherein, αiLagrange multiplier, i=1,2 ..., k, β be amount of bias, eiFor error, J (ω, e) is loss letter
Number.
According to optimal conditions:
W, e are eliminated, above-mentioned optimization problem is solved and can be exchanged into the following linear equation of solution:
Wherein:Y=(y1,y2,...,yk)T, α=(α1,α2,...,αk)T, Ω is a matrix, and its i-th row j column element isI, j=1,2 ..., k, l=1,2 ..., k, K (xi,xj) it is kernel function, I is unit
Battle array.
If A=Ω+λ-1I, because A is a positive semidefinite matrix, A-1It is existing, solution linear equation (10) obtains:
αi=A-1(yi-βl) (11)
Therefore, the α obtained according to linear equation (10) is solvedi, β can be in the hope of mathematical modeling:
According to the system model tried to achieve, the variance of system can be tried to achieve:
Wherein n is the length of data, and x and y are respectively input sample and output sample.
6. variance can be obtained by above formulaEstimate, reality output data varianceNumber is exported with controller history
According to optimal output varianceRatio, be exactly required controller performance indications η (d).
Performance evaluation can be carried out to the quality of current controller actual motion performance by the stool and urine of the value to η (d).Control
The index η (d) of System Performance Analysis processed scope shows that the current performance of system is poorer between [0,1], closer to 0, shows
System has greatly improved space.Show that the current performance of system is better closer to 1.
For verify inventive method validity, from certain factory's gas fractionation unit mathematical modeling be used as research object, answer
Simulation Evaluation research is done with MATLAB softwares.
The technological process of gas fractionation depropanizing tower unit:Liquefied gas in washing precipitation tank, through depropanizing tower feed pump
Depropanizing tower feed preheater is sent to, 70 DEG C or so are preheating to through steam condensate (SC), into depropanizing tower, depropanizing tower bottom of towe weight
Device is boiled to heat using 0.8MPa steam.C 3 fraction is distillated from tower top, and 35 DEG C of inflow depropanizing towers are cooled to through aerial condenser
Return tank, for the condensate liquid in return tank through depropanizing tower reflux pump, a part returns inflow tower, and a part goes out device as finished product
Remove liquefied petroleum gas storage shipping unit or polypropylene plant.Analyzed more than, using the capacity of returns of depropanizing tower, feed rate as defeated
Enter, using sensitive plate temperature, tower top temperature as output, its control system mathematical modeling T (q) is achieved by the method for the invention
For:
Its interference model N (q) is:
Wherein, q-1Represent backward shift operator;Interference is that average is zero, variance Σa=1.36I white noise sequence.
System controller Q (q) form is:
Interference during system operation is as shown in Figure 5.
2000 group controller output signals, signal sampling period T are produced using MATLAB softwares to above-mentioned models=
1s, unexpected to system between 1400s-1800s to apply disturbance, corresponding MATLAB simulation softwares export Y1、Y2(temperature) signal
As shown in Figure 6,7.
The sample collected is standardized according to formula (1), the process data of selection is standardized first,
To eliminate the unreasonable influence that dimension different band is come, i.e.,Draw correlation matrix
Principal component is calculated according to formula (3).Principal component is obtained using matrix reduction singular value decomposition method.Calculate accumulation
Variance percentage
The pivot number k that all cumulative variance percentages are more than set-point (80%) is found out, it is so logical so as to obtain pivot
N dimension input datas can be changed into k dimension Principal components by crossing pivot analysis.
Using the pivot drawn, least square method supporting vector machine training, identification model are carried out, and entered with the model recognized
Row Performance Evaluation.
In this system, performance indications are calculated using MVC methods and the inventive method respectively, and it is optimal with history
Performance indications compare.From Fig. 8,9 as can be seen that using what the control system performance estimating method based on PCA-LSSVM was drawn
Assessment result error is small, presses close to history optimal criteria value;And the assessment result drawn based on minimum variance method of evaluating performance is then
Suddenly big or suddenly small, fluctuation is larger, when causing assessment result inaccurate, and applying disturbance to system suddenly, is comparatively based on
PCA-LSSVM control system performance estimating method is more more sensitive than minimum variance MVC performance estimating methods.It can be seen that make
It is smaller with the assessment result performance indications amplitude of variation obtained based on PCA-LSSVM performance estimating methods, closer to actual feelings
Condition, reliable control loop information is provided for site operation personnel, control loop can be made adjustment rapidly, makes entirely to control
System reaches optimum state.
Claims (1)
1. the gas separation unit control system performance estimating method based on PCA-LSSVM, it is characterised in that comprise the following steps:
Dimension-reduction treatment is carried out to data using PCA methods;
The Principal component of acquisition is trained and modeled by least square method supporting vector machine;
Performance index value is obtained by the model of gained and is used as assessment result;
Performance evaluation is carried out to the quality of current controller actual motion performance by the size of η (d) value;
By control system performance estimating method, refinery gas's fractionating device control system is applied to;
The Principal component acquisition, which is trained and modeled by least square method supporting vector machine, to be comprised the following steps:
(3-1) selects RBF as kernel function, and finds regularization parameter γ and kernel function width ε by three step search algorithm;
(3-2) establishes Lagrangian, obtains Lagrange multiplier α and amount of bias β, and is substituted into the mathematics of control system
Model:
<mrow>
<mi>y</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<munderover>
<mi>&Sigma;</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msub>
<mi>&alpha;</mi>
<mi>i</mi>
</msub>
<mi>K</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>x</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mi>&beta;</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein K (xi,xj) it is kernel function, α ∈ αi;
(3-3) obtains the variance of system according to mathematical modeling:
<mrow>
<msubsup>
<mi>&sigma;</mi>
<mi>e</mi>
<mn>2</mn>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<msup>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>-</mo>
<mi>&beta;</mi>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mi>T</mi>
</msup>
<mrow>
<mo>(</mo>
<mi>y</mi>
<mo>-</mo>
<mi>&beta;</mi>
<mi>x</mi>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<mi>n</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, n is data length, and x and y are respectively input sample and output sample;
(3-4) is according to the systematic variance obtained, calculation of performance indicatorsWhereinFor the desired output of historical data
Variance;
The three step search algorithm comprises the following steps:
(4-1) arbitrarily interception γ and ε one section of section, structure parameter is to [γ, ε];
(4-2) finds central point and its 8 points of surrounding on the two dimensional surface that γ and ε is formed;It is minimum accurate according to mean square error
A point is then selected as new central point;
(4-3) judges whether step-length is less than a unit;If it is not, step-length is then reduced into half, return to step (4-2);
If it is less, parameter γ, ε corresponding to the central point is parameters obtained;
It is described that data progress dimension-reduction treatment is comprised the following steps using PCA methods:
The real time data that (2-1) chooses gas separation unit control system forms raw sample data Xn×m, n is data length, and m is defeated
Enter variable number, by Xn×mIt is standardized using following formula:
<mrow>
<msubsup>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
<mo>&prime;</mo>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>x</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mi>j</mi>
</msub>
</mrow>
<msub>
<mi>s</mi>
<mi>j</mi>
</msub>
</mfrac>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, xijRepresent Xn×mData,For the average value of jth item variable,For jth item
The standard deviation of variable;
(2-2) asks for the covariance matrix of the matrix, draws output data and input data by obtaining new matrix after standardization
Between correlation matrix R, ∑=diag (λ are obtained using singular value decomposition method to correlation matrix R1,λ2,...,λn), Jin Erqiu
Go out the variance of each pivotWith total varianceWherein δiFor variance, λiValue is characterized, n is
Data length;
(2-3) calculates accumulation contribution rate by following formula:
<mrow>
<mi>C</mi>
<mi>P</mi>
<mi>V</mi>
<mrow>
<mo>(</mo>
<mi>k</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msub>
<mi>&delta;</mi>
<mi>i</mi>
</msub>
</mrow>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msub>
<mi>&delta;</mi>
<mi>i</mi>
</msub>
</mrow>
</mfrac>
<mo>&times;</mo>
<mn>100</mn>
<mi>%</mi>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
</mrow>
The pivot number k that CPV (k) is more than set-point is found out, obtains the Principal component of k dimensions;
The size of the described value by η (d) to current controller actual motion performance quality carry out performance evaluation include with
Lower step:
The index η (d) of control system Performance Evaluation scope shows that the current performance of system is got between [0,1], closer to 0
Difference, show that system has greatly improved space;Show that the current performance of system is better closer to 1;
Described performance estimating method, being applied to implementation steps in refinery gas's fractionating device control system includes:
(1-1) understands gas separation unit industrial process, the main control loop of searching system, is carried out for these process control loops
Analysis, analyzer-controller structure and characteristic, understand the history optimum state in loop, and need to ensure these loops in current work
Steady under condition, safety operation, powerful guarantee is provided to calculate optimal performance;
During live real time data caused by (1-2) gas separation unit industrial processes operation is collected by Distributed Control System
Control room is entreated, selected loop uses PID controller, and the input of PID controller is using loop settings value r and the real-time number in scene
According to y deviation, the input of PID output u as control object, Field adjustment device is sent to through Distributed Control System, makes to reach
To control purpose, the process instantaneous value collected is by the significant data source as control system Performance Evaluation;
The process instantaneous value collected is carried out online evaluation by (1-3) using appraisal procedure, and assessment result forming curves are calculating
Machine is shown on interface, and the best curve of history is contrasted, and then field engineer can intuitively find out evaluation effect, existing
Field engineering staff is controlled to think highly of and newly adjusts and search other reasons in time, with Guarantee control system height according to evaluation result
The operation of effect.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410342442.4A CN104865944B (en) | 2014-07-17 | 2014-07-17 | Gas separation unit control system performance estimating method based on PCA LSSVM |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410342442.4A CN104865944B (en) | 2014-07-17 | 2014-07-17 | Gas separation unit control system performance estimating method based on PCA LSSVM |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104865944A CN104865944A (en) | 2015-08-26 |
CN104865944B true CN104865944B (en) | 2017-11-28 |
Family
ID=53911847
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410342442.4A Expired - Fee Related CN104865944B (en) | 2014-07-17 | 2014-07-17 | Gas separation unit control system performance estimating method based on PCA LSSVM |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104865944B (en) |
Families Citing this family (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105676833B (en) * | 2015-12-21 | 2018-10-12 | 海南电力技术研究院 | Power generation process control system fault detection method |
CN106777684B (en) * | 2016-12-14 | 2019-12-10 | 中国地质大学(武汉) | Method for establishing comprehensive coke ratio prediction model and predicting comprehensive coke ratio |
CN113031430B (en) * | 2021-03-02 | 2023-08-22 | 辽宁石油化工大学 | Design method of singular semi-horse discontinuous hopping system controller based on PD feedback |
CN114611416B (en) * | 2022-05-12 | 2022-08-02 | 中国空气动力研究与发展中心计算空气动力研究所 | LS-SVM modeling method for nonlinear unsteady aerodynamic characteristics of missile |
CN117519067A (en) * | 2023-10-20 | 2024-02-06 | 东北大学 | Multi-frame control performance evaluation method in continuous rolling process |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1191017A (en) * | 1995-07-14 | 1998-08-19 | 卡斯考产品公司 | Prediction of the properties of board by using a spectroscopic method combined with multivariate calibration |
CN101140458A (en) * | 2006-09-06 | 2008-03-12 | 费舍-柔斯芒特***股份有限公司 | Process plant monitoring based on multivariate statistical analysis and on-line process simulation |
CN103853152A (en) * | 2014-03-21 | 2014-06-11 | 北京工业大学 | Batch process failure monitoring method based on AR-PCA (Autoregressive Principal Component Analysis) |
-
2014
- 2014-07-17 CN CN201410342442.4A patent/CN104865944B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1191017A (en) * | 1995-07-14 | 1998-08-19 | 卡斯考产品公司 | Prediction of the properties of board by using a spectroscopic method combined with multivariate calibration |
CN101140458A (en) * | 2006-09-06 | 2008-03-12 | 费舍-柔斯芒特***股份有限公司 | Process plant monitoring based on multivariate statistical analysis and on-line process simulation |
CN103853152A (en) * | 2014-03-21 | 2014-06-11 | 北京工业大学 | Batch process failure monitoring method based on AR-PCA (Autoregressive Principal Component Analysis) |
Non-Patent Citations (1)
Title |
---|
《PCA-LSSVM方法的控制***性能评估》;蔡宏斌等;《自动化仪表》;20140120;第35卷(第1期);第0-5节 * |
Also Published As
Publication number | Publication date |
---|---|
CN104865944A (en) | 2015-08-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104865944B (en) | Gas separation unit control system performance estimating method based on PCA LSSVM | |
CN110303380B (en) | Method for predicting residual life of cutter of numerical control machine tool | |
CN104537415B (en) | A kind of non-linear process industry failure predication and recognition methodss based on compressed sensing and DROS ELM | |
CN104932488B (en) | Model predictive control performance evaluation and diagnosis method | |
Deng et al. | Modified kernel principal component analysis using double-weighted local outlier factor and its application to nonlinear process monitoring | |
CN103488091A (en) | Data-driving control process monitoring method based on dynamic component analysis | |
CN109491358B (en) | Control performance monitoring method for boiler dynamic information of million-kilowatt ultra-supercritical unit | |
CN110262450B (en) | Fault prediction method for cooperative analysis of multiple fault characteristics of steam turbine | |
CN103500281B (en) | Nonlinear system modeling method of process of sugar boiling crystallization | |
CN110991043B (en) | Energy consumption modeling and evaluating method of energy consumption system | |
Liu et al. | Unevenly sampled dynamic data modeling and monitoring with an industrial application | |
CN101872181A (en) | Prediction method for monitoring performance of power plant instruments | |
CN108647373A (en) | A kind of industrial process flexible measurement method based on xgboost models | |
CN110285403A (en) | Main Steam Temperature Control method based on controlled parameter prediction | |
CN108667069B (en) | Short-term wind power prediction method based on partial least squares regression | |
CN109507972B (en) | Industrial production process fault monitoring method based on layered non-Gaussian monitoring algorithm | |
CN104914723A (en) | Industrial process soft measurement modeling method based on cooperative training partial least squares model | |
CN109471420B (en) | CVA-SFA-based method for monitoring control performance of air preheater of large coal-fired power generator set of intelligent power plant | |
CN109538311B (en) | Real-time monitoring method for control performance of steam turbine in high-end power generation equipment | |
CN104899425A (en) | Variable selection and forecast method of silicon content in molten iron of blast furnace | |
CN107704958A (en) | A kind of thermal power plant's generated energy Forecasting Methodology of multivariable modeling | |
CN107704962A (en) | A kind of smelter steam flow interval prediction method based on imperfect time series data collection | |
CN109032117B (en) | ARMA model-based single-loop control system performance evaluation method | |
CN113420500B (en) | Intelligent atmospheric and vacuum system | |
CN112597430A (en) | Method for optimizing operation parameters of complex rectifying tower |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
EXSB | Decision made by sipo to initiate substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20171128 |