CN1694107A - Material data correction method and its system - Google Patents

Abstract

This invention discloses a method and systems of material data correction. In the process of modeling the model, it introduces random arange formula, and then enhances the expatiatory degree. It detects validity of the aboriginal meter data by data-preprocessing model, to recognize and eliminate error. It uses the built concerted model and the preprocessed detecting data, optimizes the compute by concerted model, and then provides the accordant and veracious data for process modeling, control optimizing and the statistical balance of whole factory material. This invention can effectively overcome the difficulty of complexity in detecting net and low expatiatory degree of system. It has a high precision and efficiency. The whole system structure is reasonable, opening neatly, easy to carry out, and it has superiority to work out the problem of high scale application.

Description

A kind of logistic data correction method and system thereof

Technical field

The present invention relates to logistic data correction method and system thereof.

Background technology

Modern times, flow process enterprise was to handle the flexiblesystem of logistics, energy stream and information flow.All have lot of data an every day in the enterprise database.No matter whenever, actual measurement data all has error inevitably yet.The precision of measurement instrument is also because the factors such as technology that adopted in the condition of measured material, measurement and the measurement have insecure characteristic.On the other hand, because the restriction of factors such as economic condition, instrument are aging, equipment failure, make that the data that collect from chemical plant installations are often imperfect.By frequent maintenance, the performance of measurement instrument can be improved or reduce in factory, but this raising is very limited and cost dearly.

For application such as the flowsheeting of chemical process, advanced control, material balance statistics, all wish to use real data rather than raw measurement data.Unreliable and the imperfect meeting of data brings difficulty or even wrong instruction to these application.Although we can't obtain the actual value of data on flows, adjustment of data technology can obtain more the accurate data near actual value.For an operating unit (process units), all pipeline flows that link to each other with it can be measured, use a mass balance principle so, feed rate and that deduct discharging flow and should equal zero, so we can set up a mass balance equation on this device.Because measurement data contains error, make and be not equal to zero that we are defined as the node residual error this difference feed rate and that deduct discharging flow.Thereby each data on flows on this device can be adjusted and reduce this node residual error until equalling zero.Data on flows is exactly an estimation to actual value after adjusting.For the Measurement Network that comprises a plurality of operating units (process units), traditional adjustment of data technology is exactly a quadratic programming problem that satisfies under all mass balance equation of constraint conditions, optimization aim be make all measured values and estimated value to angular variance and minimum.

Some datas on flows on an operating unit (process units) do not have measured, we just do not have enough information to come the computing node residual error yet so, hereto node we just lost the data basis of estimating actual value.For Measurement Network, this is not surveyed variable and may also be connected with other operating units (process units), merges us by node and can set up new mass balance equation and carry out data estimation, and this is not surveyed variable and just can be pushed by the data after estimating and calculate.This has just proposed the notion of system redundancy, and relative adjustment of data algorithm has PROJECTION MATRIX METHOD FOR etc.But these methods will be subjected to the restriction of Measurement Network structure and system redundancy.

The adjustment of data is at first proposed in the measurement data universal time coordinated in chemical process in 1961 by Kuehn and Davidson.Up to now, steady state data alignment technique comparative maturity and the application of existing wide industrial.The commercial Application of the adjustment of data mainly is divided into two classes: 1) process units level or operating unit level data are proofreaied and correct, be mainly flowsheeting, optimization and advanced control the service data that satisfies material, energy equilibrium is provided, as the flowsheeting program MASSBAL of WesternOntario university, the DATACON of Simulation Science company; 2) the full level of factory adjustment of data, other application software for full factory material, energy equilibrium calculating, production statistics, planned dispatching and computerized information integrated system, consistent, streams and energy flow data accurately are provided, Sigmafine as KBC company, the Openyield of Simulation Science company, the DataPro of the Advisor of AspenTech company and SUPCON company.Yet in the commercial Application of reality, the especially full level of factory adjustment of data is used, and all there are some following shortcomings in traditional data correcting method and these software packages:

1) require the Measurement Network model accurate, stable, this proposes very high requirement to the data modeling process.On the one hand, Measurement Network may change along with production run; On the other hand, be difficult to accurately grasp its Measurement Network for some complicated production run.The result of degree of accuracy coordinate to(for) this situation will descend greatly.

2) switching of the production decision in the process industry and random schedule incident happen occasionally, and in each coordination cycle, the system redundancy of Measurement Network can change owing to the influence that scheme is switched so.Above-mentioned adjustment of data technology occurs owing to the too low situation that can not get optimization solution of system redundancy through regular meeting for the adaptive ability deficiency of system redundancy time variation.

3) quality of coordination data is bigger for the dependence of raw measurement data, and the existence of human error is bigger to coordinating result's influence, and resulting unreasonable data also can't fundamentally overcome among the result for coordinating.

Summary of the invention

The purpose of this invention is to provide a kind of logistic data correction method and system thereof, to improve process industry production run logistics data accuracy.

Complicated and the low characteristics of system redundancy at the process industry Measurement Network, logistic data correction method of the present invention may further comprise the steps:

1) by Coordination Model configuration module single device or full factory logistics progress are carried out data modeling;

2) utilize data input module collection site measurement data and laboratory analysis of data, and raw data is sent into data preprocessing module;

3) data preprocessing module is carried out validity check to raw measurement data, to discern and to reject the human error that wherein contains;

4) data tuning algorithm module receives pretreated data, and according to the data model computation optimization that step 1) is set up, the measurement data on flows is carried out actual value approach estimation, and result of calculation is directly given output module as a result;

5) output module is exported the final optimization pass result by standard data interface as a result, perhaps exports the file of text file format and EXCEL electronic watch form as required.

Among the present invention, Coordination Model configuration module is used to support the physical distribution model of single device or full factory to set up, the model of being set up is the connection of the included main production plant of object, basin and belongings materials flow between them, and the principle of institute's foundation is a law of conservation of mass.At the frequent difficulty of switching the system redundancy time variation that is caused of process industry production decision, the modeling method of Coordination Model configuration module is so: 1) with the tower of each process units or reactor as non-capacitive node processing; 2) will store identical a plurality of of material and jar be defined as a virtual storage tank, with virtual storage tank also as non-capacitive node processing; 3) get rid of and the irrelevant device of material balance; 4) cancellation does not influence the device inner loop flow of material balance relationship under stable situation; 5) for the node that production decision is switched takes place, it is as follows to set up the scheduling equation:

u _i＝θ _ix _m (i＝1，2，Λ，s) (1)

${\mathrm{\θ}}_i=\frac{{\mathrm{\Δt}}_i}{T}---\left(2\right)$

In the formula, u _iBe the unknown flow rate of i scheduling scheme, x _mBe the measurement of discharge at scheme switching node place, Δ t _iBe the execution time of i scheduling scheme, T is the coordination cycle, θ _iBe defined as the parametric variable of i scheduling scheme.

The Coordination Model that said modeling method obtains in the Coordination Model configuration module is:

In M (θ) X=0 (3) formula, M (θ) is the constraint matrix of band scheduling scheme parametric variable θ, and x has surveyed variable vector.

All be furnished with measurement instrument on every pipeline of chemical process and come for we provide the original measurement signal, these signals all are actual expressions to production run pipeline flow.Data input module among the present invention is converted into the discernible data layout of computing machine with these signals, provide database interface and human-computer interaction interface two kinds of forms, on the one hand, the DCS that gathers by real-time data base by database interface visit and the data of data acquisition system (DAS), come flow, temperature, liquid level and the pressure signal of integrated production run, on the other hand, come the laboratory data of integrated production run, basin dipping data, production decision switch data and the oil product mobile data of part by the data typing of man-machine interface.

Ideally the error of measurement instrument all is to satisfy the stochastic error of normal distribution.But, may also contain human error in the raw measurement data owing to reasons such as instrument fault, equipment leakages.Will not coordinate to affect greatly to the data of back if do not deal with, even can obtain irrational coordination result.The Coordination Model that has scheduling parameter variable θ that data preprocessing module among the present invention is set up at Coordination Model configuration module utilizes the node residual test method after improving to come the identification human error.Computing formula is as follows:

$Z_r\left(i\right)=\frac{\left|r_i\right|}{\sqrt{H_r(i,i)}}(i=\mathrm{1,2},\mathrm{\Λ},n)---\left(4\right)$

Z _r(i)-i node balance test statistics after improving

R (i)-i joint constraint residual error

H _r(i, i)-variance of i joint constraint residual error after improving

r＝M(θ ₀)X (5)

H _r=M (θ ₀) VM ^T(θ ₀)+G (x) WG ^T(x) in the formula, θ ₀Be the estimated value of scheduling parameter variable θ, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable, and w is the diagonal angle variance matrix of scheduling parameter variable θ, and the definition of M (θ) and G (X) as shown in Equation (6).

$M\left(\mathrm{\θ}\right)=\frac{\∂\left(M\left(\mathrm{\θ}\right)X\right)}{\∂{\left(X\right)}^T}---\left(6\right)$

$G\left(X\right)=\frac{\∂\left(M\left(\mathrm{\θ}\right)X\right)}{{\∂\mathrm{\θ}}^T}$

For the variable that contains human error that calculates, we have two kinds of disposal routes: a kind of is with its direct deletion; Another is to extrapolate actual flow or estimate the data coordination that a value is carried out the back according to historical data according to the balance equation of this variable in Measurement Network.

Among the present invention, data tuning algorithm module is carried out actual value to the measurement data on flows and is approached estimation according to Coordination Model and the pretreated measurement data computation optimization set up.To weighting coefficient of each equation of constraint definition, carry out double optimization by recursive mode then and calculate.Concrete calculation procedure is as follows:

1) at first defines a surplus difference vector of posteriority $\stackrel{\mathrm{\Λ}}{r_i}=m_i\left(\mathrm{\θ}\right)\stackrel{\mathrm{\Λ}}{X},$ And then define an attenuation coefficient β _i, $\stackrel{\mathrm{\Λ}}{r_i}={\mathrm{\β}}_i{r}_i={\mathrm{\β}}_i{m}_i\left(\mathrm{\θ}\right)X,$ β _iComputing formula as shown in Equation (7)

${\mathrm{\β}}_i=1-\frac{{\left(m_i\left({\mathrm{\θ}}_0\right)V{m}_i^T\left({\mathrm{\θ}}_0\right)\right)}^{1/2}}{{(m_i\left({\mathrm{\θ}}_0\right)V{m}_i^T\left({\mathrm{\θ}}_0\right)+g_i\left(X\right)W{g}_i^T\left(X\right))}^{1/2}}---\left(7\right)$

In the formula, θ ₀Be the estimated value of scheduling parameter variable θ, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable, and w is the diagonal angle variance matrix of scheduling parameter variable θ;

2) definition weighting coefficient K and initialization, wherein constant k ₀Select to level off to 0 or level off to infinity respectively

$K=\left(\begin{array}{ccc}{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="A20051005016700172.png"\; state="\{\{state\}\}">k</figure-callout>}}_0& & 0\\ & O& \\ 0& & k_0\end{array}\right)---\left(8\right)$

3) make i from 1 to n, n equals model constrained equation Jie number, i weighting coefficient k _iComputing formula shown in (9)

r _i＝m _iX

$D_i=m_{i}V{M}_{(-i)}^T{(K_{(-i)}^{-2}+M_{-i}V{M}_{(-i)}^T)}^{-1}{M}_{(-i)}X$

$Y_{(-i)}=(I-V{M}_{(-i)}^T{(K_{(-i)}^{-2}+M_{(-i)}V{M}_{(-i)}^T)}^{-1}{M}_{(-i)})V---\left(9\right)$

$A_i=m_i{Y}_{(-i)}{m}_i^T$

$k_i^2=\frac{(1-{\mathrm{\&beta;}}_i)r_i-D_i}{{\mathrm{\&beta;}}_i{r}_i{A}_i}$

In the formula, r _iBe i joint constraint residual error, M _(-i)Equal i constraint m with constraint matrix M _iMatrix after the removal, and K _(-i)It is the matrix after the capable and i row of i with weighting coefficient matrix K are removed;

4) convergence is analyzed: if the relative variation of all weighting coefficients between two continuous estimations makes i add 1 greater than a fixing threshold value so, return computing formula (9), otherwise loop ends enters next step

5) had after the weighting coefficient, coordinated the result

As shown in Equation (10)

${\stackrel{\mathrm{\&Lambda;}}{X}}_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1}){\stackrel{\mathrm{\&Lambda;}}{X}}_i$

$i=0,\mathrm{\&Lambda;},n-1$

$y_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1})y_i---\left(10\right)$

$i=0,\mathrm{\&Lambda;},n-2$

${\stackrel{\mathrm{\&Lambda;}}{X}}_0=\mathrm{<figure-callout\; id="0"\; label="Xand\; y"\; filenames="A20051005016700172.png"\; state="\{\{state\}\}">Xand</figure-callout>}{y}_{}{}_0=V$

In the formula, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable.

Coordinate result and raw measurement data and be saved in the lump, can be application such as flowsheeting, optimal control and full factory material data statistical equilibrium by the standard data interface that output module as a result provided among the present invention consistent, accurate data are provided.These data are meant measurement data, the system coordination value of process double optimization and do not survey the estimated value of variable.Also can export the file of text file format and EXCEL electronic watch form as required.

The system that is used for above-mentioned logistic data correction method comprises 5 modules: Coordination Model configuration module, data input module, data preprocessing module, data tuning algorithm module and output module as a result.

-Coordination Model configuration module by model configuration interface, according to the physical connection of single device or the logistics of full factory, is set up and is contained the Coordination Model of dispatching equation and export to data tuning algorithm module;

-data input module, form by database interface and human-computer interaction interface, be used for original measurement signal, the laboratory analysis of data of measurement instrument on every pipeline of chemical process are converted into the discernible data layout of computing machine, integrated data export to data preprocessing module and output module as a result respectively;

-data preprocessing module, be used for data input module integrated raw measurement data carry out validity check, identification and reject the human error wherein contain, pretreated data are directly exported to data tuning algorithm module;

-data tuning algorithm module is used for institute's input Coordination Model and pretreated measurement data are carried out initialization process and computation optimization, and result of calculation is exported to output module as a result;

-output module as a result, the raw measurement data and the computation optimization data that are used for being imported are preserved and are handled, and final double optimization measurement data, system coordination value and the estimated value of not surveying variable are exported.

Advantage of the present invention:

1) the Coordination Model modeling method that proposes among the present invention can effectively solve the complicated and low difficulty of system redundancy of process industry Measurement Network;

2) the data tuning algorithm that is adopted not only can processing production process the uncertainty of material balance relationship model, and have higher computational accuracy and efficient;

3) total system is rational in infrastructure, is convenient to realize, especially has advantage for solving the large-scale industrial application problem;

4) system has opening, dirigibility, and data outputting module provides standard interface.Be convenient to realize data integration with other application system.

Description of drawings

Fig. 1 is a schematic diagram of the present invention, is the logistic data correction system in the frame of broken lines among the figure;

Fig. 2 is a data tuning algorithm modular algorithm process flow diagram;

Fig. 3 is a comparison diagram of using system of the present invention front and back device mean equilibrium rate, and the left side histogram is for using the equal balanced ratio of pre-installed horizontalization, and the right side histogram is device mean equilibrium rate after using.

Embodiment

The invention will be further described below in conjunction with accompanying drawing and example.

With reference to Fig. 1, logistic data correction of the present invention system comprises:

-Coordination Model configuration module 1 by model configuration interface, according to the physical connection of single device or the logistics of full factory, is set up and is contained the Coordination Model of dispatching equation and export to data tuning algorithm module;

-data input module 2, form by database interface and human-computer interaction interface, be used for original measurement signal, the laboratory analysis of data of measurement instrument on every pipeline of chemical process are converted into the discernible data layout of computing machine, integrated data export to data preprocessing module 3 and output module 5 as a result respectively;

-data preprocessing module 3, be used for data input module 2 integrated raw measurement data carry out validity check, identification and reject the human error wherein contain, pretreated data are directly exported to data tuning algorithm module 4;

-data tuning algorithm module 4 is used for institute's input Coordination Model and pretreated measurement data are carried out initialization process and computation optimization, and result of calculation is exported to output module 5 as a result;

-output module 5 as a result, the raw measurement data and the computation optimization data that are used for being imported are preserved and are handled, and final double optimization measurement data, system coordination value and the estimated value of not surveying variable are exported.

The data correction system that important commercial Application of the present invention is full level of factory, be other application software of full factory material, energy equilibrium calculatings, production statistics, planned dispatching and computerized information integrated system, unanimity, streams and energy flow data accurately are provided.Next introduce example and specific implementation process that this system specifically is applied to some refinerys.

This refinery comprises covering device more than 30 and 184 basins of normal pressure, decompression, catalytic cracking, reformation, hydrogenation etc.For this type of flow process enterprise, all can there be every day a large amount of production datas to be integrated by various data acquisition system (DAS)s.Enterprise must carry out statistical study to these data in order to reach the target of steady production, energy-saving and cost-reducing, Optimizing operation.But contain error in the raw measurement data inevitably, this just can influence the accuracy of statistical study, even can obtain wrong conclusion and steering order.System of the present invention in full factory and workshop material balance statistics application just as same " true value is approached the estimation device ", can be plan, scheduling, statistical department consistent, streams and energy flow data accurately are provided, thereby satisfy the needs and the energy saving purposes of the integrated information robotization of full factory.

At first need Coordination Model configuration module that process is carried out data modeling.For the production procedure that provides in this example, the resulting Coordination Model of traditional modeling method is as shown in the formula (11).

The existence of not surveying variable U in AX+BU=0 (11) model not only can influence the coordination precision, and has reduced system redundancy and cause can not get optimization solution.The present invention proposes new modeling method, what it was followed is regular as follows:

1) tower of each process units or reactor are the quantity-produced operating units, and the liquid storage quantitative changeization is little, can be with it as non-capacitive node;

2) storage tank is typical capacitive node, and its liquid storage quantitative changeization can not be ignored.We will store material identical jar be defined as a virtual storage tank, need not to consider its inner pot with jar between the tank switching amount, and storage liquid measure variation that will this virtual storage tank is defined as virtual flow;

3) get rid of and the irrelevant device of material balance, as heat interchanger, pump etc.;

4) cancellation does not influence the device inner loop flow of material balance relationship etc. under stable situation;

5) for the node that production decision is switched takes place, to set up the scheduling equation and also be introduced in the initial quality balance equation, original variable of not surveying is surveyed variable and is replaced in the Measurement Network, introduces random schedule equation parameter variable simultaneously.Dispatch equation shown in formula (1), (2), wherein Δ t _iBe the execution time of i scheduling scheme, T is the coordination cycle.

u _i＝θ _ix _m (i＝1，2，Λ，s) (1)

${\mathrm{\&theta;}}_i=\frac{{\mathrm{\&Delta;t}}_i}{T}---\left(2\right)$

In the formula, u _iBe the unknown flow rate of i scheduling scheme, x _mBe the measurement of discharge at scheme switching node place, Δ t _iBe the execution time of i scheduling scheme, T is the coordination cycle, θ _iBe defined as the parametric variable of i scheduling scheme.

By said method, our resulting Coordination Model as shown in Equation (3), M (θ) is the constraint matrix of band random schedule equation parameter variable θ, x has surveyed variable vector.Not survey variable in the system is all picked out, and redundance improves greatly.

M(θ)X＝0 (3)

The various sensors of this refinery provide a large amount of raw measurement datas for us, but these signals need data input module that these signals are converted into the discernible data layout of computing machine just to crossing actual a description of range of flow.According to the characteristics of refinery survey sensor and various data integrated systems, provide database interface and human-computer interaction interface two kinds of forms.On the one hand, the flow of production run, temperature, liquid level, pressure signal are obtained by DCS and the integrated production scene data of data acquisition system (DAS) by real-time data base by the database interface visit.On the other hand, the basin dipping data of the laboratory data of production run, part, production decision switching and oil product mobile data are next integrated by the data typing of man-machine interface.

These raw measurement datas are directly delivered to data preprocessing module and are carried out data validation, to discern and to reject the human error that wherein contains.At the novel Coordination Model of being set up that has model parameter variable θ, not only to consider the influence of measurand error, also to consider the influence of the uncertain degree of model.Therefore we have done some improvement with the node balance test statistics, utilize the node residual test method after improving to come the identification appreciable error.Computing formula is as follows:

$Z_r\left(i\right)=\frac{\left|r_i\right|}{\sqrt{H_r(i,i)}}(i=\mathrm{1,2},\mathrm{\&Lambda;},n)---\left(4\right)$

Z _r(i)--i node balance test statistics after the improvement

R (i)-i joint constraint residual error

H _r(i, i)-variance of i joint constraint residual error after improving

r＝M(θ ₀)X (5)

H _r＝M(θ ₀)VM ^T(θ ₀)+G(x)WG ^T(x)

In the formula, θ ₀Be the estimated value of scheduling parameter variable θ, X has surveyed variable vector, and V is a covariance matrix of having surveyed variable, and W is the diagonal angle variance matrix of scheduling parameter variable θ, and the definition of M (θ) and G (X) as shown in Equation (6).

$M\left(\mathrm{\&theta;}\right)=\frac{\&PartialD;\left(M\left(\mathrm{\&theta;}\right)X\right)}{\&PartialD;{\left(X\right)}^T}---\left(6\right)$

$G\left(X\right)=\frac{\&PartialD;\left(M\left(\mathrm{\&theta;}\right)X\right)}{{\&PartialD;\mathrm{\&theta;}}^T}$

Select level of significance, determine the critical value Z of balance check statistic _cIf, Z _r(i)＞Z _c, think that so i node may contain appreciable error.After determining to contain the node of appreciable error, further infer the variable that contains appreciable error according to the connection situation between network structure and the node again.For the variable that contains appreciable error that finds, we have two kinds of disposal routes.A kind of is with its direct deletion, but can introduce new not survey variable again like this.We advise adopting second method, extrapolate actual flow or estimate the data coordination that a value is carried out the back according to historical data according to the balance equation of this variable in Measurement Network exactly.

Accompanying drawing 2 has provided the concrete workflow of data tuning algorithm module, and wherein the concrete calculation procedure of data coordination is as follows:

1) at first defines a surplus difference vector of posteriority ${\stackrel{\mathrm{\&Lambda;}}{r}}_i=m_i\left(\mathrm{\&theta;}\right)\stackrel{\mathrm{\&Lambda;}}{X},$ And then define an attenuation coefficient β _i, ${\stackrel{\mathrm{\&Lambda;}}{r}}_i={\mathrm{\&beta;}}_i{r}_i={\mathrm{\&beta;}}_i{m}_i\left(\mathrm{\&theta;}\right)X,$ β _iComputing formula as shown in Equation (7)

${\mathrm{\&beta;}}_i=1-\frac{{\left(m_i\left({\mathrm{\&theta;}}_0\right)V{m}_i^T\left({\mathrm{\&theta;}}_0\right)\right)}^{1/2}}{{(m_i\left({\mathrm{\&theta;}}_0\right)V{m}_i^T\left({\mathrm{\&theta;}}_0\right)+g_i\left(X\right)W{g}_i^T\left(X\right))}^{1/2}}---\left(7\right)$

In the formula, θ ₀Be the estimated value of scheduling parameter variable θ, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable, and w is the diagonal angle variance matrix of scheduling parameter variable θ;

2) definition weighting coefficient K and initialization, wherein constant k ₀Select to level off to 0 or level off to infinity respectively

$K=\left(\begin{array}{ccc}{\mathrm{<figure-callout\; id="0"\; label="k"\; filenames="A20051005016700172.png"\; state="\{\{state\}\}">k</figure-callout>}}_0& & 0\\ & O& \\ 0& & k_0\end{array}\right)---\left(8\right)$

3) make i from 1 to n, n equals model constrained equation number, i weighting coefficient k _iComputing formula shown in (9)

r _i＝m _iX

$D_i=m_{i}V{M}_{(-i)}^T{(K_{(-i)}^{-2}+M_{-i}V{M}_{(-i)}^T)}^{-1}{M}_{(-i)}X$

$Y_{(-i)}=(I-V{M}_{(-i)}^T{(K_{(-i)}^{-2}+M_{(-i)}V{M}_{(-i)}^T)}^{-1}{M}_{(-i)})V---\left(9\right)$

$A_i=m_i{Y}_{(-i)}{m}_i^T$

$k_i^2=\frac{(1-{\mathrm{\&beta;}}_i)r_i-D_i}{{\mathrm{\&beta;}}_i{r}_i{A}_i}$

In the formula, r _iBe i joint constraint residual error, M _(-i)Equal i constraint m with constraint matrix M _iMatrix after the removal, and K _(-i)It is the matrix after the capable and i row of i with weighting coefficient matrix K are removed;

4) convergence is analyzed: if the relative variation of all weighting coefficients between two continuous estimations makes i add 1 greater than a fixing threshold value so, return computing formula (9), otherwise loop ends enters next step

5) had after the weighting coefficient, coordinated the result As shown in Equation (10)

${\stackrel{\mathrm{\&Lambda;}}{X}}_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1}){\stackrel{\mathrm{\&Lambda;}}{X}}_i$

$i=0,\mathrm{\&Lambda;},n-1$

$y_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1})y_i$

$i=0,\mathrm{\&Lambda;},n-2$

${\stackrel{\mathrm{\&Lambda;}}{X}}_0=\mathrm{<figure-callout\; id="0"\; label="Xand\; y"\; filenames="A20051005016700172.png"\; state="\{\{state\}\}">Xand</figure-callout>}{y}_{}{}_0=V$

In the formula, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable.

Coordinate result and raw measurement data and be saved in the lump, the standard data interface that provides by output module as a result writes results among the relation data platform.The data that the statistical department of this refinery utilizes system handles of the present invention to cross are made the material balance statistical form in full factory and each workshop, and accompanying drawing 3 is comparison diagrams of using all device mean equilibrium rates of the full factory in front and back of system of the present invention.As seen from the figure, adopt the present invention, its logistics data accuracy and precision increase substantially.

Claims (8)

1. logistic data correction method is characterized in that may further comprise the steps:

1) by Coordination Model configuration module single device or full factory logistics progress are carried out data modeling;

2) utilize data input module collection site measurement data and laboratory analysis of data, and raw data is sent into data preprocessing module;

3) data preprocessing module is carried out validity check to raw measurement data, to discern and to reject the human error that wherein contains;

4) data tuning algorithm module receives pretreated data, and according to the data model computation optimization that step 1) is set up, the measurement data on flows is carried out actual value approach estimation, and result of calculation is directly given output module as a result;

5) output module is exported the final optimization pass result by standard data interface as a result, perhaps exports the file of text file format and EXCEL electronic watch form as required.

2. logistic data correction method according to claim 1 is characterized in that said modeling method in the Coordination Model configuration module: 1) with the tower of each process units or reactor as non-capacitive node processing; 2) will store identical a plurality of of material and jar be defined as a virtual storage tank, with virtual storage tank also as non-capacitive node processing; 3) get rid of and the irrelevant device of material balance; 4) cancellation does not influence the device inner loop flow of material balance relationship under stable situation; 5) for the node that production decision is switched takes place, it is as follows to set up the scheduling equation:

u _i＝θ _ix _m (i＝1，2，Λ，s) (1)

${\mathrm{\&theta;}}_i=\frac{\mathrm{\&Delta;}{t}_i}{T}----\left(2\right)$

In the formula, u _iBe the unknown flow rate of i scheduling scheme, x _mBe the measurement of discharge at scheme switching node place, Δ t _iBe the execution time of i scheduling scheme, T is the coordination cycle, θ _iBe defined as the parametric variable of i scheduling scheme.

3. logistic data correction method according to claim 2 is characterized in that the Coordination Model that said modeling method obtains in the Coordination Model configuration module is:

M(θ)X＝0 (3)

In the formula, M (θ) is the constraint matrix of band scheduling scheme parametric variable θ, and x has surveyed variable vector.

4. logistic data correction method according to claim 1, it is characterized in that said data input module: the original measurement signal on every pipeline in the chemical process is converted into the discernible data layout of computing machine, provide database interface and human-computer interaction interface two kinds of forms, on the one hand, the DCS that gathers by real-time data base by database interface visit and the data of data acquisition system (DAS), come the flow of integrated production run, temperature, liquid level and pressure signal, on the other hand, come the laboratory data of integrated production run by the data typing of man-machine interface, the basin dipping data of part, production decision switch data and oil product mobile data.

5. logistic data correction method according to claim 1, it is characterized in that saidly carrying out data validation:, utilize the node residual test method after improving to come the identification human error at the Coordination Model that has scheduling parameter variable θ that Coordination Model configuration module is set up by data preprocessing module.Computing formula is as follows:

$Z_r\left(i\right)=\frac{\left|r_i\right|}{\sqrt{H_r(i,i)}}(i=\mathrm{1,2},\mathrm{\&Lambda;},n)----\left(4\right)$

z _r(i)-i node balance test statistics after improving

R (i)-i joint constraint residual error

H _r(i, i)-variance of i joint constraint residual error after improving

r＝M(θ ₀)X

H _r＝M(θ ₀)VM ^T(θ ₀)+G(x)WG ^T(x) (5)

In the formula, θ ₀Be the estimated value of scheduling parameter variable θ, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable, and w is the diagonal angle variance matrix of scheduling parameter variable θ, and the definition of M (θ) and G (X) as shown in Equation (6).

$\begin{array}{c}M\left(\mathrm{\&theta;}\right)=\frac{\&PartialD;\left(M\left(\mathrm{\&theta;}\right)X\right)}{\&PartialD;{\left(X\right)}^T}\\ G\left(X\right)=\frac{\&PartialD;\left(M\left(\mathrm{\&theta;}\right)X\right)}{\&PartialD;{\mathrm{\&theta;}}^T}\end{array}----\left(6\right)$

6. logistic data correction method according to claim 1 is characterized in that said data tuning algorithm module: to weighting coefficient of each equation of constraint definition, carry out double optimization by recursive mode then and calculate.Concrete calculation procedure is as follows:

1) at first defines a surplus difference vector of posteriority ${\stackrel{\mathrm{\&Lambda;}}{r}}_i=m_i\left(\mathrm{\&theta;}\right)\stackrel{\mathrm{\&Lambda;}}{X},$ And then define an attenuation coefficient β _i,

${\stackrel{\mathrm{\&Lambda;}}{r}}_i={\mathrm{\&beta;}}_i{r}_i={\mathrm{\&beta;}}_i{m}_i\left(\mathrm{\&theta;}\right)X,$ β _iComputing formula as shown in Equation (7)

${\mathrm{\&beta;}}_i=1-\frac{{\left(m_i\left({\mathrm{\&theta;}}_0\right)V{m}_i^T\left({\mathrm{\&theta;}}_0\right)\right)}^{1/2}}{{(m_i\left({\mathrm{\&theta;}}_0\right)V{m}_i^T\left({\mathrm{\&theta;}}_0\right)+g_i\left(X\right)W{g}_i^T\left(X\right))}^{1/2}}----\left(7\right)$

In the formula, θ ₀Be the estimated value of scheduling parameter variable θ, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable, and w is the diagonal angle variance matrix of scheduling parameter variable θ;

2) definition weighting coefficient K and initialization, wherein constant k ₀Select to level off to 0 or level off to infinity respectively

$K=\left(\begin{array}{ccc}{k}_0& & 0\\ & O& \\ 0& & k_0\end{array}\right)----\left(8\right)$

3) make i from 1 to n, n equals model constrained equation number, i weighting coefficient k _iComputing formula shown in (9)

r _i＝m _iX

$D_i=m_i{\mathrm{VM}}_{(-i)}^T{(K_{(-i)}^{-2}+M_{-i}{\mathrm{VM}}_{(-i)}^T)}^{-1}{M}_{(-i)}X$

$Y_{(-i)}=(I-{\mathrm{VM}}_{(-i)}^T{(K_{(-i)}^{-2}+M_{(-i)}{\mathrm{VM}}_{(-i)}^T)}^{-1}{M}_{(-i)})V----\left(9\right)$

$A_i=m_i{Y}_{(-i)}{m}_i^T$

$k_i^2=\frac{(1-{\mathrm{\&beta;}}_i)r_i-D_i}{{\mathrm{\&beta;}}_i{r}_i{A}_i}$

In the formula, r _iBe i joint constraint residual error, M _(-i)Equal i constraint m with constraint matrix M _iMatrix after the removal, and K _(-i)It is the matrix after the capable and i row of i with weighting coefficient matrix K are removed;

4) convergence is analyzed: if the relative variation of all weighting coefficients between two continuous estimations makes i add 1 greater than a fixing threshold value so, return computing formula (9), otherwise loop ends enters next step

5) had after the weighting coefficient, coordinated the result

As shown in Equation (10)

${\stackrel{\mathrm{\&Lambda;}}{X}}_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1}){\stackrel{\mathrm{\&Lambda;}}{X}}_i$

i＝0，Λ，n-1

$y_{i+1}=(I-y_i{m}_{i+1}^T{(k_{i+1}^{-2}+m_{i+1}{y}_i{m}_{i+1}^T)}^{-1}{m}_{i+1})y_i----\left(10\right)$

i＝0，Λ，n-2

${\stackrel{\mathrm{\&Lambda;}}{X}}_0=\mathrm{X\; and}{y}_0=V$

In the formula, x has surveyed variable vector, and v is a covariance matrix of having surveyed variable.

7. logistic data correction method according to claim 1 is characterized in that final optimization pass result that output module is as a result exported is meant through measurement data, the system coordination value of double optimization and does not survey the estimated value of variable.

8. adopt the system of logistic data correction method as claimed in claim 1, it is characterized in that comprising:

-Coordination Model configuration module (1) by model configuration interface, according to the physical connection of single device or the logistics of full factory, is set up and is contained the Coordination Model of dispatching equation and export to data tuning algorithm module;

-data input module (2), form by database interface and human-computer interaction interface, be used for original measurement signal, the laboratory analysis of data of measurement instrument on every pipeline of chemical process are converted into the discernible data layout of computing machine, integrated data export to data preprocessing module (3) and output module (5) as a result respectively;

-data preprocessing module (3), be used for data input module (2) integrated raw measurement data carry out validity check, identification and reject the human error wherein contain, pretreated data are directly exported to data tuning algorithm module (4);

-data tuning algorithm module (4) is used for institute's input Coordination Model and pretreated measurement data are carried out initialization process and computation optimization, and result of calculation is exported to output module (5) as a result;

-output module (5) as a result, the raw measurement data and the computation optimization data that are used for being imported are preserved and are handled, and final double optimization measurement data, system coordination value and the estimated value of not surveying variable are exported.

Images