CN106227699A - A kind of blast furnace throat cross temperature measurer center band temperature predicting method and system - Google Patents

Abstract

The present invention provides a kind of blast furnace throat cross temperature measurer center band temperature predicting method and system, and the method includes: gather the process variable of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature；The process variable gathered is done pretreatment, removes noise spike saltus step data and high frequency noise data；Center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.This system, including: acquisition module, pretreatment module, prediction module.The temperature of the cross temperature measurer center band that the present invention predicts future by the historical data of the input selected by center band temperature prediction model and output exports, can accurately express temperature output and control the relation between input, ensure that in the maintenance of cross temperature measurer center temperature sensor or damage, adjusting blast furnace top and the bottom system in time provides foundation to enable blast furnace operating personnel accurately to judge, and then realizes the direct motion stable, efficient, safe of blast furnace.

Description

A kind of blast furnace throat cross temperature measurer center band temperature predicting method and system

Technical field

The invention belongs to blast furnace temperature monitoring technical field, specifically a kind of blast furnace throat cross temperature measurer center band temperature Degree Forecasting Methodology and system.

Background technology

Iron and steel uses most important basic raw material and the most functional material of annual production as human society, by extensively Apply in all trades and professions such as transportation, machinery manufacturing industry, building and military developments.Blast furnace ironmaking is as steel and iron industry Important procedure, it is ensured that blast furnace ironmaking is efficiently, safety and stability direct motion is to the sustainable and healthy development of steel and iron industry and reduces the energy Consumption suffers from important function.It is the process of the physical-chemical reaction of a many kinds of substance complex shape due to blast furnace process, real Now to its Automated condtrol metallurgy and an automation field unsolved subject difficult problem all the time, particularly abnormal at blast furnace Under the working of a furnace, furnace temperature is predicted accurately and effectively controls, and the intelligent automation realizing blast furnace ironmaking process controls, Current field of metallurgy and automatically control the advanced subject of development in science and technology especially.

Blast furnace temperature mainly comprises three aspects: molten iron temperature in furnace throat temperature, furnace wall temperature, stove.At present, domestic big portion The monitoring means of the blast furnace throat temperature divided is from stock gas CO₂Sample analysis is changed into cross temperature measurer temperature curve Analyze.Owing to stock gas is long for sample time, CO₂Component analysis time lag is big, and error easily occurs in furnace temperature analysis；And cross temperature temperature Write music line and stock gas CO₂Composition profiles has a good corresponding relation: the ground that furnace throat temperature that cross temperature curve shows is high Side's Gas Flow is vigorous, CO₂Content is low.And cross temperature measurer has the advantage that sample frequency is big, data volume big, to stove Condition reacting condition is sensitive, can provide foundation as the adjustment of blast furnace operating person timely top and the bottom system.Utilize cross temperature bent Line instructs the adjustment of blast furnace indices, and particularly when conditions of blast furnace fluctuation is abnormal, for blast furnace operating, person judges top and the bottom system The adjustment direction of degree provides foundation, and the fast quick-recovery of conditions of blast furnace is played positive role, is the effective inspection avoiding furnace condition disorder Survey means.

But, blast furnace cross temperature measuring equipment center band temperature relatively other points for measuring temperature are high, and sensor is easily damaged and changes week Phase is long, thus cannot monitor gas temperature in stove in time, judges that gas flow distribution brings impact to blast furnace operating personnel, and causing cannot The blast furnace operating systems such as the most correct adjustment cloth, air blast, and then affect the direct motion of blast furnace.On the one hand, complicated due to blast furnace Multivariate, the feature of multiple time delay, it is the most difficult for setting up cross temperature point Temperature Mechanism model accurately.On the other hand, base In the development of the modeling method of data-driven, make accurately to predict that cross temperature central temperature is possibly realized.

In view of operate during blast furnace ironmaking the variation of controlled quentity controlled variable the impact of blast furnace temperature is not only had ageing but also The feature of stickiness when having, when therefore carrying out the PREDICTIVE CONTROL of blast furnace temperature, also should on the premise of keeping current furnace condition anterograde Take into account the benign development of next stove working of a furnace.

Summary of the invention

The deficiency existed for prior art, the present invention provides a kind of blast furnace throat cross temperature measurer center band temperature pre- Survey method and system.

The technical scheme is that

A kind of blast furnace throat cross temperature measurer center band temperature predicting method, including:

Gather the process variable of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature；

The process variable gathered is done pretreatment, removes noise spike saltus step data and high frequency noise data；

Center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.

Described center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction, bag Include:

ARMAX is selected to set up center band temperature prediction model；

Determine two polynomial orders of backward shift operator in center band temperature prediction model；

Use recurrent least square method identification cross temperature measurer center band temperature prediction model parameter, after i.e. two, move calculation Polynomial coefficient in submultinomial matrix, and then obtain final cross temperature measurer center band temperature prediction model；

Final center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.

Described selection ARMAX sets up center band temperature prediction model, including:

Gather the process variable of history blast furnace throat temperature, i.e. blast furnace throat cross temperature measurer each measure some temperature and The top temperature of blast furnace four direction；

Choose the process variable conduct of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature The input variable of center band temperature prediction model；

The input variable of center band temperature prediction model is done pretreatment, removes noise spike saltus step data and high-frequency noise Data；

ARMAX is selected to set up center band temperature prediction model, the rank of two backward shift operator polynomial matrix in described model The secondary delayed order representing input variable respectively and the delayed order of output variable, describe between input variable and output variable Time lag relation between time lag relation, output variable the most in the same time, two backward shift operator polynomial matrix in described model Coefficient describes the functional relationship between input variable and output variable.

The described process variable choosing the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature As the input variable of center band temperature prediction model, including:

Use factor-analysis approach master of core out band temperature from the process variable of described history blast furnace throat temperature The factor；

Utilize main gene and process variable to do Pearson correlation analysis, tentatively choose input variable；

The input variable tentatively chosen and center band temperature are done Pearson correlation analysis, rejects and center band temperature Incoherent input variable, the input variable of the center band temperature prediction model finally chosen.

Described determine two polynomial orders of backward shift operator in center band temperature prediction model, including:

According to two backward shift operator polynomial order various combinations, calculate the AIC value that the combination of each order is corresponding；

The codomain of AIC value is divided into N number of node；

Select the AIC value immediate order combination with each node, under different order combinations, calculate cross respectively survey The goodness of fit of temperature device center band temperature prediction model, selects the highest order of the goodness of fit and combines as two backward shift operators Polynomial order.

A kind of furnace throat cross temperature measurer center band temperature prediction system, including:

Acquisition module, for gathering the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature Process variable；

Pretreatment module, for the process variable gathered does pretreatment, removes noise spike saltus step data and high frequency is made an uproar Sound data；

Prediction module, utilizes center band temperature prediction model to carry out blast furnace throat cross temperature measurer center band temperature pre- Survey.

Described prediction module, including:

Model building module, is used for selecting ARMAX to set up center band temperature prediction model；

Order determines module, is used for determining two polynomial orders of backward shift operator in center band temperature prediction model；

Parameter identification module, is used for using recurrent least square method identification cross temperature measurer center band temperature prediction model Parameter, polynomial coefficient in i.e. two backward shift operator polynomial matrix, and then obtain final cross temperature measurer center band Temperature prediction model；

Temperature prediction module, for utilizing final center band temperature prediction model to carry out blast furnace throat cross temperature measurer Center band temperature prediction.

Described model building module, including:

Process variable acquisition module, surveys for gathering the process variable of history blast furnace throat temperature, i.e. blast furnace throat cross The each of temperature device measures some temperature and the top temperature of blast furnace four direction；

Input variable chooses module, for choosing the State of Blast Furnace relevant to blast furnace throat cross temperature measurer center band temperature The process variable of larynx temperature is as the input variable of center band temperature prediction model；

Input variable pretreatment module, for the input variable of center band temperature prediction model is done pretreatment, removal is made an uproar Sound spike saltus step data and high frequency noise data；

Forecast model sets up module, is used for selecting ARMAX to set up center band temperature prediction model, in described model after two Move Operator Polynomial order of matrix and represent the delayed order of input variable and the delayed order of output variable respectively, describe output Time lag relation between time lag relation between variable and input variable, output variable the most in the same time, in described model two The coefficient of backward shift operator polynomial matrix describes the functional relationship between input variable and output variable.

Described input variable chooses module, including:

Main gene chooses module, for using factor-analysis approach from the process variable of described history blast furnace throat temperature The main gene of core out band temperature；

First chooses module, is used for utilizing main gene and process variable to do Pearson correlation analysis, tentatively chooses input and become Amount；

Second chooses module, for the input variable tentatively chosen and center band temperature are done Pearson correlation analysis, Reject and the incoherent input variable of center band temperature, the input variable of the center band temperature prediction model finally chosen.

Described order determines module, including:

AIC value computing module, for according to two backward shift operator polynomial order various combinations, calculates each order group Close corresponding AIC value；

Node division module, for being divided into N number of node by the codomain of AIC value；

Order combination determines module, for selecting the AIC value immediate order combination with each node, at different orders Calculate the goodness of fit of cross temperature measurer center band temperature prediction model under combination respectively, select the order that the goodness of fit is the highest Combination is as two polynomial orders of backward shift operator.

Beneficial effect:

High in order to solve cross temperature measurer central temperature point, sensor is easily damaged and the replacing sensor cycle is long, with And person cannot judge the problems such as stock gas flow distribution in time to cause blast furnace operating during sensor degradation, the present invention is based on time discrete Between the dynamic modeling thought of sequence, proposition with seasonal effect in time series auto regressive moving average modeling algorithm under the effect controlling input Set up the center band temperature prediction model of multi output, five, cross temperature measurer center point for measuring temperature temperature is carried out On-line Estimation. Ensureing when cross temperature central temperature point cannot normally be measured, operator can estimate stove according to the temperature prediction value of model Interior gas fluid distrbution, adjusts burden distribution matrix in time, it is ensured that blast furnace stable smooth operation.

The present invention is based on M-ARMAX intelligent modeling method, blast furnace technology mechanism, it is proposed that a kind of blast furnace throat cross temperature Device center band temperature predicting method and system.The present invention is gone through by input selected by center band temperature prediction model and output History data predict the temperature output of the cross temperature measurer center band in future, have method simple, and built center band temperature Forecast model can accurately be expressed temperature output and control the relation between input, and center band temperature prediction model accuracy is high, energy Enough guarantee, in the maintenance of cross temperature measurer center temperature sensor or damage, provides accurate furnace throat temperature for blast furnace operating personnel Data, adjusting blast furnace top and the bottom system in time provides foundation to enable blast furnace operating personnel accurately to judge, and then realizes blast furnace Direct motion stable, efficient, safe.

Accompanying drawing explanation

Fig. 1 is blast furnace throat cross temperature measurer schematic diagram in prior art；

Fig. 2 is blast furnace throat cross temperature measurer center band temperature predicting method flow process in the specific embodiment of the invention Figure；

Fig. 3 is step 3 flow chart in the specific embodiment of the invention；

Fig. 4 is step 3-1 flow chart in the specific embodiment of the invention；

Fig. 5 is step 3-1-2 flow chart in the specific embodiment of the invention；

Fig. 6 is step 3-2 flow chart in the specific embodiment of the invention；

Fig. 7 is step 3-2-1 flow chart in the specific embodiment of the invention；

Fig. 8 is step 3-2-3 flow chart in the specific embodiment of the invention；

Fig. 9 is step 3-3 flow chart in the specific embodiment of the invention；

Figure 10 is stove furnace throat cross temperature measurer center band temperature prediction system block diagram in the specific embodiment of the invention；

Figure 11 is prediction module block diagram in the specific embodiment of the invention；

Figure 12 is that in the specific embodiment of the invention, module frame chart set up by model；

Figure 13 is that in the specific embodiment of the invention, input variable chooses module frame chart；

Figure 14 is that in the specific embodiment of the invention, order determines module frame chart；

Figure 15 is cross temperature central temperature prediction effect figure based on MARMAX, and wherein (a)～(c) is respectively point for measuring temperature 5,6,16 design sketch being shown as 1000 sampled points, (d)～(e) is respectively point for measuring temperature 15,17 and is shown as the effect of 300 sampled points Figure；

In Fig. 1, TT1 is top temperature northeast temperature sensor, and TT2 is top temperature southeast temperature sensor, and TT3 is temperature northwest, top temperature Degree sensor, TT4 is top temperature southwest temperature sensor, and TT5 is cross temperature point 3 temperature sensor, and TT6 is cross temperature point 8 Temperature sensor, point 10 temperature sensor of TT7 cross temperature.

Detailed description of the invention

Below in conjunction with the accompanying drawings the detailed description of the invention of the present invention is elaborated.

As it is shown in figure 1, blast furnace throat cross temperature measurer is arranged on blast furnace throat or sealing cover in the most cross mode, bag Include 4 cross temperature rifles, be separately mounted to blast furnace throat northwest, southwest, the southeast, direction, 4, northeast, thermoelectricity on cross temperature rifle Even sequence number arrangement is as follows: to the southeast be 1～No. 10 successively from northwest；From southwest to northeast, it is 11～No. 21 successively.Cross is surveyed Temperature rifle numbering is respectively No. 1, No. 2, No. 3, No. 4 to northeastward clockwise from southeastern direction.Cross temperature rifle is drawn money on credit and short That props up divides, and wherein direction, northwest (No. 3) are to draw money on credit, and it arranges 6 temperature thermocouples altogether；3 of remaining direction are short, its On all arrange 5 temperature thermocouples.Site layout project has blast furnace throat cross temperature measurer, blast furnace roof top temperature measurement device, number According to harvester, computer system.Data acquisition unit connects blast furnace temperature-measuring system, and connects department of computer science by communication bus System.

Temperature element on blast furnace throat cross temperature measurer uses armouring nickel chromium-nickel silicon thermocouple, and its temperature-measuring range exists Between 0-1100 DEG C, the instantaneous the highest ambient temperature that can bear is less than 1200 DEG C.Thermoelectricity involved in present embodiment Even model is: WRKK-333, Φ 8mm, K Graduation Number, protects tubing matter: GH3030 or 1Cr18Ni9Ti.

Blast furnace throat cross temperature measurer is used for measuring charge level gas temperature in State of Blast Furnace, has reliability height, seriality The features such as good, digitized.Blast furnace throat cross temperature measurer instead of furnace throat gas sampling analysis, improves the working of a furnace and the type of furnace is sentenced Disconnected accuracy, plays important directive function to Reasonable adjustment burden distribution system.

Owing to blast furnace throat cross temperature measurer center band temperature exceeds nearly 500 degree than other point for measuring temperature temperature, blast furnace Furnace throat cross temperature measurer center band thermocouple easily damaged and thermocouple the replacing sensor cycle is long, affects the working of a furnace and the type of furnace Timely judgement.5, blast furnace throat cross temperature measurer center temperature is predicted by present embodiment, and model accuracy is high, When thermocouple break or the repair and replacement of blast furnace throat cross temperature measurer center band thermometric, it is possible to ensure cross temperature curve Continuously display, judge that the working of a furnace adjusts burden distribution system in time and provides foundation, to ensure smooth operation of furnace for blast furnace operating personnel.

For convenience of describing, the symbol and the term that use present embodiment are defined as follows:

Point for measuring temperature 3 temperature u₁(k), DEG C；

Point for measuring temperature 8 temperature u₂(k), DEG C；

Point for measuring temperature 10 temperature u₃(k), DEG C；

Top, southeast temperature temperature u₄(k), DEG C；

Top, northwest temperature temperature u₅(k), DEG C；

Top, northeast temperature temperature u₆(k), DEG C；

Top, southwest temperature temperature u₇(k), DEG C；

Point for measuring temperature 5 temperature y₁(k), DEG C；

Point for measuring temperature 6 temperature y₂(k), DEG C；

Point for measuring temperature 16 temperature y₃(k), DEG C；

Point for measuring temperature 15 temperature y₄(k), DEG C；

Point for measuring temperature 17 temperature y₅(k), DEG C；

Present embodiment provides a kind of blast furnace throat cross temperature measurer center band temperature predicting method, as in figure 2 it is shown, bag Include:

Step 1, the process of the blast furnace throat temperature that collection is relevant to blast furnace throat cross temperature measurer center band temperature become Amount；

Step 2, the process variable gathered is done pretreatment, remove noise spike saltus step data and high frequency noise data；

Step 3, center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.

Described step 3, as it is shown on figure 3, include:

Step 3-1, selection ARMAX set up center band temperature prediction model；

Described step 3-1, as shown in Figure 4, including:

Step 3-1-1, each survey of the process variable of collection history blast furnace throat temperature, i.e. blast furnace throat cross temperature measurer Amount point temperature and the top temperature of blast furnace four direction；

Step 3-1-2, choose the mistake of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature Cheng Bianliang is as the input variable of center band temperature prediction model；

Described step 3-1-2, as it is shown in figure 5, include:

During step 3-1-2-1, employing factor-analysis approach are selected from the process variable of described history blast furnace throat temperature The main gene of its central band temperature；

Step 3-1-2-2, utilize main gene and process variable to do Pearson correlation analysis, tentatively choose input variable；

Step 3-1-2-3, the input variable tentatively chosen and center band temperature are done Pearson correlation analysis, reject Input variable incoherent with center band temperature, the input variable of the center band temperature prediction model finally chosen.

In present embodiment, center band temperature prediction model is output as the survey of five, blast furnace throat cross temperature measurer center Warm spot temperature includes: y₁Point for measuring temperature 5, y₂Point for measuring temperature 6, y₂Point for measuring temperature 16, y₄Point for measuring temperature 15, y₅Point for measuring temperature 17.

Blast furnace is a complicated system, the process variable of blast furnace throat temperature up to 30 kinds.If all of data quilt Use, the over-fitting of center band temperature prediction model will necessarily be caused.Therefore, choose 5000 groups of blast furnace real data, use because of Sub-analysis method selects the main gene that can explain five points for measuring temperature in blast furnace throat cross temperature measurer center, and uses this Main gene and process variable do Pearson correlation analysis.Dimension choosing and the output variable phase of incorporation model is reduced for simplified model The closing property variable more than 0.5, selects following 9 input variables: cross temperature point 3, cross temperature point 4, cross temperature point 8, ten Word point for measuring temperature 10, cross temperature point 20, the temperature southeast, top, temperature northwest, top, temperature southwest, top, temperature northwest, top.Again because factorial analysis is selected The main gene gone out does not has practical significance, can not explain the full detail of five temperature spots in center, therefore above selected input Need to do Pearson correlation analysis again with five, cross temperature center temperature spot to reject and the central temperature incoherent variable of point, Choose following 7 variablees the most altogether as input variable: cross temperature point 3, cross temperature point 8, cross temperature point 10, temperature east, top South, temperature northwest, top, temperature southwest, top, temperature northwest, top.

Step 3-1-3, input variable to center band temperature prediction model do pretreatment, remove noise spike saltus step data And high frequency noise data；

Owing to the impact of blast furnace external disturbance makes the input variable of center band temperature prediction model there may be with chance error Difference, to this end, use noise spike filtering algorithm to reject the noise spike saltus step data in input variable；Afterwards, mobile putting down is used During all filtering method rejects input variable temperature value less than 10 DEG C high frequency noise data.

Step 3-1-4, selection ARMAX set up center band temperature prediction model, and in described model, two backward shift operators are multinomial Formula order of matrix represents the delayed order of input variable and the delayed order of output variable respectively, describes output variable and input Time lag relation between time lag relation between variable, output variable the most in the same time, in described model, two backward shift operators are many The coefficient of item formula matrix describes the functional relationship between input variable and output variable.

Temperature in view of five temperature spots in cross temperature center not only has dependence with input variable, goes back and himself There is dependency relation, and the existence of blast furnace internal random interference in historical data, selects ARMAX (Auto Regressive Moving Average) as center band temperature prediction model.Five temperature spots and furnace throat cross centered by output variable simultaneously Also there is between the temperature spot output variable of five, thermometric center dependency, therefore also need during modeling to consider the phase between output variable Pass relation.In addition modeling data be can real-time data collection, the interference caused inside blast furnace has incorporated in blast furnace real data, institute To define the ARMAX in present embodiment for Multi-outputAuto Regressive MovingAverage, i.e. multi output ARMAX, setting up center band temperature prediction model is:

Y (k)=A (z^-1)y(k)+B(z^-1)u(k) (1)

In formula,

U (k)=[u₁(k), u₂(k) ..., u₇(k)]^TFor input vector, y (k)=[y₁(k), y₂(k), y₃(k), y₄(k), y_x(k)]^TFor output vector.Backward shift operator polynomial matrix A (z^-1) the delayed order that order np is output variable, a_ij×1, a_ij×2..., a_ij×npRepresent current k moment output temperature point and history k-1, k-2 ..., the function of the output temperature in k-np moment Relation.Backward shift operator polynomial matrix B (z^-1) the delayed order that order nq is input variable, b_ij×1, b_ij×2..., b_ij×npTable Show current k moment output temperature point and history k-1, k-2 ..., the functional relationship of the input variable temperature in k-nq moment.

Step 3-2, determine two polynomial orders of backward shift operator in center band temperature prediction model；

Described step 3-2, as shown in Figure 6, including:

Step 3-2-1, according to two backward shift operator polynomial order various combinations, calculate each order combination corresponding AIC value；

Described step 3-2-1, as it is shown in fig. 7, comprises:

Step 3-2-1-1: by the center band temperature prediction model parameter of RLS identification l group；

Step 3-2-1-2: the order of l group and the parameter of identification are substituted into formula (1) calculating and records center band temperature is pre- Survey model prediction exportsAnd center band temperature prediction model residual epsilon_l(k, θ_N)；

Step 3-2-1-3: calculate AIC value according to formula (2)；

Owing to the too high easy over-fitting of model order and model are complicated, make output order np and input order respectively at this Nq changes totally 25 groups of orders one by one and combines from 1 to 5, note l=1, and 2,3 ..., 25 is the numbering often organized；Calculate respectively according to formula (2) Often group AIC value corresponding to order；

$AIC=log\left(v\right)+\frac{2d}{N}---\left(2\right)$

$v=\frac{1}{N}\underset{i=1}{\overset{N}{\Σ}}{{\mathrm{\ϵ}}_i}^2$

In formula: centered by d=np+nq band temperature prediction model order and, v is loss function, and N is order number of combinations, ε_iFor the center band temperature prediction model residual error corresponding to i-th group of order combination.

Step 3-2-1-4: judge that l, whether less than 25, meets and goes to step 3-2-1-1, otherwise continue next step.

Step 3-2-2, the codomain of AIC value is divided into N=10 node；Select from 25 AIC codomains according to formula (3) 10 decile nodal value AIC_node；

${\mathrm{AIC}}_{node}={\mathrm{AIC}}_{min}+\frac{{\mathrm{AIC}}_{min}+{\mathrm{AIC}}_{max}}{10}n---\left(3\right)$

Step 3-2-3, selection and AIC value AIC of each node_nodeImmediate order combination (totally 10 groups), in difference Calculate the goodness of fit fit of center band temperature prediction model under order combination respectively according to formula (4), select the goodness of fit the highest Order combines as two polynomial orders of backward shift operator.

$fit=100\×(1-\frac{\left|\right|y-\hat{y}\left|\right|}{\left|\right|y-mean\left(y\right)\left|\right|})---\left(4\right)$

In formula, n=1,2,3 ..., 10 is the AIC node serial number selected,Centered by band temperature prediction model prediction defeated Go out, AIC_max, AIC_min, AIC_nodeRepresent maximum in 25 groups of AIC values that the combination of above-mentioned 25 groups of orders is calculated, the most respectively Little value and the nodal value immediate AIC value with 10 deciles between the minimum and maximum.

Described step 3-2-3, as shown in Figure 8, including:

Step 3-2-3-1: select corresponding A IC_node10 group model orders (np, nq)_nodeAnd extract step 3-2-1-2 The center band temperature prediction model prediction output of middle correspondence

Step 3-2-3-2: according to formula (4) andCalculate the fit value fit of these 10 groups of orders；

Step 3-2-3-3: find out value fit maximum in 10 groups of fit value_max；

Step 3-2-3-4: find out fit_maxCorresponding model order, is model Optimal order (np, nq)_opt；

Step 3-2-3-5: terminate and preserve Optimal order.

Step 3-3, utilization recurrent least square method identification center band temperature prediction model parameter, i.e. two backward shift operators are many Item formula matrix, and then obtain final cross temperature measurer center band temperature prediction model；

Combine based on Optimal order, use recurrence least square (RLS) technology identification center band temperature prediction model parameter. Recurrent least square method utilizes the training data the being newly introduced center band temperature prediction model parameter estimation value root to previous moment It is modified according to recursive algorithm thus obtains the center band temperature prediction model parameter estimation value in this moment, along with training data Continually introducing, center band temperature prediction model parameter estimation value is continuously available to be revised until reaching satisfied precision.Compare Conventional algorithm of common least square, RLS amount of calculation and amount of storage are little, fast convergence rate and be capable of distinguishing online Know.Using RLS need to give initial value θ (0)=0, P (0)=α I, when α is sufficiently large positive number, the corrective action of RLS is big, convergence Speed is fast.α=10 are taken at this⁶。

Described step 3-3, as it is shown in figure 9, include:

Step 3-3-1: center band temperature prediction model parameter initial value θ (0)=0, inverse matrix initial value P (0)=10 are set⁶I, I is unit matrix；

Step 3-3-2: the data vector matrix in structure k-1 moment

θ^T=[θ₁, θ₂, θ₃, θ₄, θ₅] (6)

y_i=[-y_i(k-1) … -y_i(k-np)]

u_j=[u_j(k) … u_j(k-nq)]

θ_i ^T=[A_i1, A_i2, A_i3, A_i4, A_i5, B_i1..., B_i7]

A_ik=[a_ik1, a_ik2..., a_iknp],

B_ij=[b_ij0, b_ij2..., b_ijnq],

Y (k)=[y₁(k), y₂(k), y₃(k), y₄(k), y₅(k)]^T

In formula, i=1,2,3,4,5, j=1,2 ..., 7, k=1,2,3,4,5, θ^TFor parameter matrix.

Step 3-3-3: calculate gain matrix K (k) according to formula (7)；

Step 3-3-4: calculate parameter estimation vector θ (k) according to formula (8)；

Step 3-3-5: judge whether θ (k) meets formula (11) and shut down criterion, meets and goes to step 3-3-8, under otherwise continuing One step.

Step 3-3-6: calculate matrix P of matrix (k) according to formula (9)；

Step 3-3-7: recursion one step k-1 → k, returns the data vector matrix that step 3-3-2 structure is new.

Step 3-3-8: terminate and preserve the parameter of identification.

Formula (5) and formula (6) substitute in formula (1) then center band temperature prediction model is:

The recursive operation of RLS, θ in formula is terminated by the shutdown criterion of formula (11)_kI () is the i-th element of parameter vector θ Recurrence result in kth time.σ is that a certain positive number represents parameters precision requirement, takes σ=0.05 at this.

$\underset{i}{max}\left|\frac{{\mathrm{\&theta;}}_k\left(i\right)-{\mathrm{\&theta;}}_{k-1}\left(i\right)}{{\mathrm{\&theta;}}_k\left(i\right)}\right|<\mathrm{\&sigma;}---\left(11\right)$

Order determines and parameter identification complete after according to M-ARMAX temperature prediction modelObtain center Band temperature prediction model predication value

Introduce relative error δ (%) target function such as formula (12) the modeling effect of center band temperature prediction model is commented Estimating, if meeting standard conditions δ≤5% of actual production, then terminating this center band temperature prediction model training；If be not inconsistent Close, then need to reselect the data that training sample set i.e. collects；

$\mathrm{\&delta;}=\frac{1}{L}\underset{i=1}{\overset{L}{\&Sigma;}}\frac{|y_i-{\hat{y}}_i|}{y_i}\&times;100\%---\left(12\right)$

In formula,The predictive value of w center band temperature prediction model, y_iFor output actual value, L is training sample capacity.

Step 3-4, final center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature Degree prediction.

Present embodiment also provides for a kind of stove furnace throat cross temperature measurer center band temperature prediction system, as shown in Figure 10, Including:

Acquisition module, for gathering the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature Process variable；

Pretreatment module, for the process variable gathered does pretreatment, removes noise spike saltus step data and high frequency is made an uproar Sound data；

Prediction module, utilizes center band temperature prediction model to carry out blast furnace throat cross temperature measurer center band temperature pre- Survey.

Described prediction module, as shown in figure 11, including:

Model building module, is used for selecting ARMAX to set up center band temperature prediction model；

Order determines module, is used for determining two polynomial orders of backward shift operator in center band temperature prediction model；

Parameter identification module, is used for using recurrent least square method identification cross temperature measurer center band temperature prediction model Parameter, i.e. two backward shift operator polynomial matrix, and then obtain final cross temperature measurer center band temperature prediction model；

Temperature prediction module, for utilizing final center band temperature prediction model to carry out blast furnace throat cross temperature measurer Center band temperature prediction.

Described model building module, as shown in figure 12, including:

Process variable acquisition module, surveys for gathering the process variable of history blast furnace throat temperature, i.e. blast furnace throat cross The each of temperature device measures some temperature；

Input variable chooses module, for choosing the State of Blast Furnace relevant to blast furnace throat cross temperature measurer center band temperature The process variable of larynx temperature is as the input variable of center band temperature prediction model；

Input variable pretreatment module, for the input variable of center band temperature prediction model is done pretreatment, removal is made an uproar Sound spike saltus step data and high frequency noise data；

Forecast model sets up module, is used for selecting ARMAX to set up center band temperature prediction model, in described model after two Move Operator Polynomial order of matrix and represent the delayed order of input variable and the delayed order of output variable respectively, describe output Time lag relation between time lag relation between variable and input variable, output variable the most in the same time, in described model two The coefficient of backward shift operator polynomial matrix describes the functional relationship between input variable and output variable.

Described input variable chooses module, as shown in figure 13, and including:

Main gene chooses module, for using factor-analysis approach to select the process variable of described history blast furnace throat temperature In the main gene relevant to center band temperature；

First chooses module, is used for utilizing main gene and process variable to do Pearson correlation analysis, tentatively chooses input and become Amount；

Second chooses module, for the input variable tentatively chosen and center band temperature are done Pearson correlation analysis, Reject and the incoherent input variable of center band temperature, the input variable of the center band temperature prediction model finally chosen.

Described order determines module, as shown in figure 14, and including:

AIC value computing module, for according to two backward shift operator polynomial order various combinations, calculates each order group Close corresponding AIC value；

Node division module, for being divided into N number of node by the codomain of AIC value；

Order combination determines module, for selecting the AIC value immediate order combination with each node, at different orders The goodness of fit of the lower cross temperature measurer center band temperature prediction model respectively of combination, selects the order combination that the goodness of fit is the highest As two polynomial orders of backward shift operator.

The dynamic characteristic complicated due to system for blast furnace ironmaking and easily being affected by raw material, in order to ensure M-ARMAX temperature The precision of forecast model, needs to use new sample data to carry out model when the relative error of temperature prediction value is more than 5% Re-training.

The software that the present invention uses high-level language C# to carry out 5 temperature predicting methods in cross temperature measurer center realizes.Should Software interface achieves data and shows, inquires about and predict the outcome the functions such as display, can allow blast furnace operating personnel easily Obtain its required blast furnace temperature information.Should be equipped with OPC bitcom on the computer of this temperature prediction software additionally, install It is responsible for carrying out data double-way communication with slave computer and data acquisition unit.

Figure 15 (a)～(e) are M-ARMAX temperature model prediction effect figure of the present invention, compared with measured value it can be seen that this 5, invention cross temperature measurer center M-ARMAX temperature prediction is accurate, it is possible to immediately following the variation tendency of measured value.Additionally, this Inventive method training speed is fast, model accuracy is high；A kind of ancillary technique as cross temperature measurer, it is ensured that furnace temperature is monitored Seriality and accuracy.Judge that gas fluid distrbution and conditions of blast furnace provide reliable basis in stove for operator.

Claims (10)

1. a blast furnace throat cross temperature measurer center band temperature predicting method, it is characterised in that including:

Gather the process variable of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature；

The process variable gathered is done pretreatment, removes noise spike saltus step data and high frequency noise data；

Center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.

Method the most according to claim 1, it is characterised in that described utilize center band temperature prediction model to carry out State of Blast Furnace Larynx cross temperature measurer center band temperature prediction, including:

ARMAX is selected to set up center band temperature prediction model；

Determine two polynomial orders of backward shift operator in center band temperature prediction model；

Using recurrent least square method identification cross temperature measurer center band temperature prediction model parameter, i.e. two backward shift operators are many Polynomial coefficient in item formula matrix, and then obtain final cross temperature measurer center band temperature prediction model；

Final center band temperature prediction model is utilized to carry out blast furnace throat cross temperature measurer center band temperature prediction.

Method the most according to claim 2, it is characterised in that described selection ARMAX sets up center band temperature prediction model, Including:

Gather each of the process variable of history blast furnace throat temperature, i.e. blast furnace throat cross temperature measurer and measure some temperature and blast furnace The top temperature of four direction；

Choose the process variable of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature as center Input variable with temperature prediction model；

The input variable of center band temperature prediction model is done pretreatment, removes noise spike saltus step data and high-frequency noise number According to；

Selecting ARMAX to set up center band temperature prediction model, in described model, the order of two backward shift operator polynomial matrix divides Do not represent the delayed order of input variable and the delayed order of output variable, describe the time lag between input variable and output variable Time lag relation between relation, output variable the most in the same time, the coefficient of two backward shift operator polynomial matrix in described model Functional relationship between input variable and output variable is described.

Method the most according to claim 2, it is characterised in that described in choose and blast furnace throat cross temperature measurer center band The process variable of the blast furnace throat temperature that temperature is relevant as the input variable of center band temperature prediction model, including:

Use factor-analysis approach main gene of core out band temperature from the process variable of described history blast furnace throat temperature；

Utilize main gene and process variable to do Pearson correlation analysis, tentatively choose input variable；

The input variable tentatively chosen and center band temperature are done Pearson correlation analysis, rejects and center band temperature not phase The input variable closed, the input variable of the center band temperature prediction model finally chosen.

Method the most according to claim 2, it is characterised in that described determine in center band temperature prediction model move after two The order of Operator Polynomial, including:

According to two backward shift operator polynomial order various combinations, calculate the AIC value that the combination of each order is corresponding；

The codomain of AIC value is divided into N number of node；

The AIC value immediate order combination of selection and each node, under different orders combinations in difference cross temperature measurer The goodness of fit of its central band temperature prediction model, the order combination selecting the goodness of fit the highest is polynomial as two backward shift operators Order.

6. a stove furnace throat cross temperature measurer center band temperature prediction system, it is characterised in that including:

Acquisition module, for gathering the process of the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature Variable；

Pretreatment module, for the process variable gathered does pretreatment, removes noise spike saltus step data and high-frequency noise number According to；

Prediction module, utilizes center band temperature prediction model to carry out blast furnace throat cross temperature measurer center band temperature prediction.

System the most according to claim 6, it is characterised in that described prediction module, including:

Model building module, is used for selecting ARMAX to set up center band temperature prediction model；

Order determines module, is used for determining two polynomial orders of backward shift operator in center band temperature prediction model；

Parameter identification module, is used for using recurrent least square method identification cross temperature measurer center band temperature prediction model to join Number, polynomial coefficient in i.e. two backward shift operator polynomial matrix, and then obtain final cross temperature measurer center band temperature Degree forecast model；

Temperature prediction module, for utilizing final center band temperature prediction model to carry out blast furnace throat cross temperature measurer center Band temperature prediction.

System the most according to claim 7, it is characterised in that described model building module, including:

Process variable acquisition module, for gathering the process variable of history blast furnace throat temperature, i.e. blast furnace throat cross temperature dress The each measurement point temperature put；

Input variable chooses module, for choosing the blast furnace throat temperature relevant to blast furnace throat cross temperature measurer center band temperature The process variable of degree is as the input variable of center band temperature prediction model；

Input variable pretreatment module, for the input variable of center band temperature prediction model is done pretreatment, removes noise point Peak saltus step data and high frequency noise data；

Forecast model sets up module, is used for selecting ARMAX to set up center band temperature prediction model, moves calculation in described model after two Submultinomial order of matrix represents the delayed order of input variable and the delayed order of output variable respectively, describes output variable And the time lag relation between time lag relation between input variable, output variable the most in the same time, moves after two in described model The coefficient of Operator Polynomial matrix describes the functional relationship between input variable and output variable.

System the most according to claim 8, it is characterised in that described input variable chooses module, including:

Main gene chooses module, in using factor-analysis approach to select the process variable of described history blast furnace throat temperature with The main gene that center band temperature is relevant；

First chooses module, is used for utilizing main gene and process variable to do Pearson correlation analysis, tentatively chooses input variable；

Second chooses module, for the input variable tentatively chosen and center band temperature are done Pearson correlation analysis, rejects Input variable incoherent with center band temperature, the input variable of the center band temperature prediction model finally chosen.

System the most according to claim 7, it is characterised in that described order determines module, including:

AIC value computing module, for according to two backward shift operator polynomial order various combinations, calculates the combination of each order right The AIC value answered；

Node division module, for being divided into N number of node by the codomain of AIC value；

Order combination determines module, for selecting the AIC value immediate order combination with each node, in different order combinations The goodness of fit of lower cross temperature measurer center band temperature prediction model respectively, selects the order combination conduct that the goodness of fit is the highest Two polynomial orders of backward shift operator.