Three, summary of the invention
The present invention mainly solves the quality closed-loop control problem of butadiene production device.The present invention adopts multiple soft sensor modeling technology, set up the industrial soft-sensing model of 5 of DA106 tower still water-content, DA106 cat head methylacetylene content, DA107 overhead product divinyl and total alkynes content, DA107 tower still butadiene content at first respectively, and implemented at the DCS device, solved the real-time online estimation problem of DA106 and DA107 tower quality product preferably.Then, on the basis of above-mentioned soft-sensing model, designed and developed the advanced control system that comprises 4 loops.The The whole control scheme comprehensively is made of the control strategies such as optimization of inferring control, constraint control and set(ting)value, and is implemented on DCS by configuration, has finally realized the closed-loop optimization control of two rectifying tower quality producies.Soft-sensing model is succinctly practical, and the on-line correction reliable in function is effective.The whole control conceptual design advanced person, practical reliable has reached obviously and has cut down the consumption of energy, and improves product yield and guarantees the effect of operating safety.
This Controlling System mainly comprises 4 advanced control loops: four Composition Control loops such as DA106 tower still water-content, DA106 cat head methylacetylene concentration, DA107 overhead product concentration, DA107 tower still butadiene concentration.
Technical scheme of the present invention is: a kind of process for advanced controlling rectifying apparatus of butadiene, described rectifier unit comprise first rectifying tower and second rectifying tower, general called after DA106 and DA107 tower, and described method may further comprise the steps:
Utilize metrical instrument to obtain the real time data of DA106 tower measurement point;
Start soft-sensing model, real time data is carried out importing soft-sensing model after the filtering, measure DA106 tower still water-content;
Start soft-sensing model, real time data is carried out importing soft-sensing model after the filtering, measure DA106 cat head methylacetylene content;
The DA106 tower still water-content that draws according to soft-sensing model and the numerical value of cat head methylacetylene content are implemented the real-time online closed-loop optimization control of DA106 quality product on the DCS basis.
The design and the realization of 2 industrial soft-sensing models of DA106 tower
The DA106 rectifying tower is an azeotropy rectification column.Soft measurement be exactly select with by one group of relevant measurable variable of predictor, to construct certain be input with the measurable variable, estimated the mathematical model that variable is output, with the estimation of computer software realization significant process variable.The core content of soft-measuring technique is exactly an accurate and reliable model of setting up measurand, set up soft-sensing model the experience modeling method of normal employing artificial neural network and regression analysis are arranged.Among the present invention, at first adopted artificial neural network and multiple regression analysis modeling method to set up 2 industrial soft-sensing models of methylacetylene content in DA106 tower still water-content and the cat head tail gas, and in the DCS system, implemented.
Wherein: independent variable(s) has tower still hot solvent flow F133, hot solvent inlet temperature T113, hot solvent temperature out T500-25, tower still temperature T 500-26, feed rate F122, quantity of reflux F134, tower still pressure P 116 in the DA106 tower still water-content soft-sensing model.Output variable is: tower still water-content estimated value.
In these 7 parameters, determine that T500-25, T500-26, F122, F134, P116 are chosen to be the process subsidiary variable.Consider because the effect of F133 and T113 definite value control loop do not have anything to get in touch with the variation of tower still water-content, with its rejecting.Because assorted in the flow measurements of F122 and F134 have a large amount of high frequency noises, need in addition filtering process.350 groups of data that collect suitably are divided into two portions: 300 groups of data are as the modeling sample collection, and 50 groups of data are as the cross check sample set.After carrying out stepwise regression analysis with the modeling collection, set up following industrial soft-sensing model:
In the formula: water-content estimated value at the bottom of S130 (t): the t tower constantly;
τ: pure retardation time;
x
i(t-τ) (i=1,2 ..., 5): F122, F134, T500-25, T500-26 and P116
Observed value;
C
i(i=0,1 ..., 5): regression coefficient, value is:
C
0=-84.677507?????C
1=2.841364??????C
2=-0.655388
C
3=-1.731556??????C
4=1.582286??????C
5=231.365314
6 independent variable(s) in the DA106 cat head tail gas methylacetylene content soft-sensing model: quantity discharged F135, the quantity of reflux F134 of temperature T 400-4, cat head between feed rate F122, tower top pressure P117, tower top temperature T300-4, condenser and backflow are irritated; Output variable is methylacetylene in the S132 sampling point analytical value (being a MA) content.The DA106 tower collects 250 groups of related datas altogether, selects 200 groups for use as training data from 250 groups of DA106 tower take off data, and all the other 50 groups as checking data.Set up industrial soft-sensing model:
S132 (t): t cat head MA content sampling analysis value constantly;
x
i(t) (i=1,2 ..., 6): F122, F134, F135, T300-4, T400-4 and
P117t observed value constantly;
B
i(i=0,1 ..., 6): regression coefficient, value is:
B0=27.976935;B1=0.000000;B2=-0.000000;B3=-0.000496;
B4=-0.124183;B5=0.000000;B6=-6.189611
This regression model form is fairly simple, is convenient to Project Realization.After real running, further designed the on-line correction function again, thereby improved the estimated accuracy of soft-sensing model widely.Wherein tower still water-content soft-sensing model evaluated error be controlled in ± the 1.2*10-6 scope in, in the cat head tail gas methylacetylene content model evaluated error be controlled in ± 0.05 scope in.
The design and implementation of 2 advanced control loops of DA106 tower
Based on the soft-sensing model of above-mentioned DA106 rectifying tower, study, developed advanced control loop 1 of tower still water-content and the advanced control loop 2 of cat head methylacetylene concentration.The advanced control loop 1 of tower still has mainly adopted the deduction control strategy, and in control strategy, set up the quantity of reflux constraint and controlled, prevent preferably that not only reflux ratio from changing excessive phenomenon, guarantee that quantity of reflux is controlled in the given zone of reasonableness, and realized making water-content in the product to meet quality control index (under the prerequisite of the water-content of product≤20mg/kg), reaching the purpose that cuts down the consumption of energy; The advanced control loop 2 of cat head has mainly adopted deduction-Optimal Control Strategy, the key of this optimization control scheme design is that the strictness of methylacetylene content is controlled to a suitable level, because accident if the methylacetylene excessive concentration can be blasted, and methylacetylene concentration is crossed the low increase that can cause the divinyl quantity discharged, so that directly reduced the yield of product divinyl, deduction-Optimal Control Strategy is to infer on the basis of control on the basis, increase by one again and optimize mechanism, the set(ting)value of inferring controller is optimized.By implement this control scheme at the DCS system configuration, the exhaust emissions throughput ratio is controlled at a lower level more reposefully, not only can satisfy technic index but also reduce the loss of divinyl significantly.According to according to a preliminary estimate, move the divinyl yield that this Controlling System can reach 200T/y.
Above-mentioned process for advanced controlling rectifying apparatus of butadiene, further comprising the steps of:
Utilize metrical instrument to obtain the real time data of DA107 tower measurement point;
Start soft-sensing model, real time data is carried out importing soft-sensing model after the filtering, measure DA107 cat head divinyl-1,3 and total alkynes content;
Start soft-sensing model, real time data is carried out importing soft-sensing model after the filtering, measure DA107 tower still divinyl-1,3 content;
The numerical value of the cat head divinyl-1,3 that draws according to soft-sensing model and total alkynes content, tower still divinyl-1,3 content is implemented the real-time online closed-loop optimization control of the second rectifying tower quality product on the DCS basis.
The design and the realization of 3 industrial soft-sensing models of DA107 tower
In order to realize advanced person's control of DA107 rectifying tower, 3 industrial soft-sensing models have at first been set up, then on the basis of soft-sensing model to the composition On-line Estimation, adopt deduction control, constraint control, select to control the design of carrying out advanced control with the control strategies such as optimization of set(ting)value, and implemented on DCS by configuration, realized the closed-loop control of composition.For the real-time soft-sensing model that uses of Industrial processes, except needs develop that some pace of learnings are fast, the network structure and learning algorithm of strong robustness, also the sampled data of necessary correct preference pattern training makes the soft-sensing model of being built have well " smoothly " and " extrapolation " ability.
Owing to can't test full scale plant, obtain to express the data of disturbance and product relation, and the variation range of field data be very narrow again.Therefore in the research project, inherent production operation mechanism characteristic from the DA107 tower, hydro carbons isolation technique in the applied chemistry engineering and the process model building technology in the control engineering, and computer simulation technique have been set up the production technique mechanism mathematical model of DA107 rectifier unit.According to this model not only can the emulation full scale plant performance characteristic, seek inner link between each process operation parameter, and the deep layer operative knowledge storehouse of modelling DA107 rectifying tower in view of the above.Technology mathematics mechanism model according to the DA107 rectifying tower, use orthogonal design method, 3x organizes orthogonal operational condition by the design of L9 (34) orthogonal table, and the actual production operating process of analog D A107 rectification cell makes the performance characteristic of full scale plant obtain fully " excitation ".Good with the more living orthogonalities of mechanism model common property, and the DA107 rectifying production process operation information that the changing operate-point scope is bigger, and with the industrial production collection in worksite to real time data combine, constitute the sample set of DA107 tower soft-sensing model, guaranteed the orthogonality and the completeness of sample space.
In DA107 tower soft sensor modeling, at first adopted artificial neural network and multiple regression analysis modeling method to set up butadiene content in the DA107 overhead product, total alkynes content and 3 industrial soft-sensing models of tower still extraction butadiene content, and in the DCS system these 3 models of actual motion.According to the model estimated accuracy, further designed dynamic compensation and on-line correction module, improve the estimated accuracy of model.
Divinyl and total alkynes content industry soft-sensing model in the DA107 overhead product
According to the technological operation variable analysis of device production and the process mechanism mathematical model simulation research of DA107 rectifying tower, determined the soft-sensing model subsidiary variable of divinyl-1,3 in the DA107 column overhead product and total alkynes content: liquid level L122, tower still temperature T 300-6, tower bottom pressure P119, stripping section column plate temperature T W-111, TW-112 at the bottom of produced quantity F138, the tower top temperature T500-30 of feed rate F122, quantity of reflux F139, tower still and tower top pressure P120, the tower.Output variable: S136 artificial sample analytical value and industrial on-line chromatograph analytical value divinyl-1,3 and total alkynes content.Utilize the pivot analysis technology, at first 500 groups of correlated samples data matrixes to collection in worksite carry out data compression and the extraction of characteristic manipulation variable, to reach sample matrix are carried out orthogonalization and dimensionality reduction purpose.
Utilize above-mentioned multiple stepwise regression method, set up two soft-sensing models of divinyl-1,3 in the DA107 column overhead product and total alkynes content respectively.Consider the completeness of learning sample, choose 350 groups of sampled datas collecting during July 22 to 14 days October in 2003 in 2002 as e-learning and extensive sample, what wherein be used for network model training fits 290 groups in sample, is used for 60 groups of the forecast samples of testing model extrapolability.At first utilize the intelligent data treatment technology that image data is carried out filtering and normalized, utilize the pivot analysis technology that these the 350 groups sampled datas that influence DA107 column overhead butadiene concentration operational variable are carried out data compression and the extraction of characteristic manipulation variable then, utilize multiple regression experience modeling technique at last, set up the soft measurement regression model of following divinyl-1,3:
Wherein: x
1~x
10: the observed value of F122, F138, F139, P119, P120, T500-30, T300_6, TW_111, TW_122 and LICA-122
A
i(i=0,1 ..., 10): regression coefficient, value is:
A
0=1.021286;?A
1=-0.001353;A
2=-0.000002;A
3=0.000172;
A
4=0.071321;?A
5=0.085794;?A
6=-0.001402;A
7=0.000222;
A
8=-0.000601;A
9=-0.000242;A
10=-0.000027
Utilize multiple regression experience modeling technique, set up the soft measurement regression model of following total alkynes:
Wherein: x
1~x
10: the observed value of F122, F138, F139, P119, P120, T500-30, T300_6, TW_111, TW_122 and LICA-122
B
i(i=0,1 ..., 10): regression coefficient, value is:
B
0=-140.225122;B
1=11.655014;?B
2=-0.007929;B
3=-1.488601;
B
4=-683.916151;B
5=215.677896;B
6=7.259641;?B
7=-0.234634;
B
8=0.000000;???B
9=0.257475;??B
10=0.099942
The soft-sensing model on-line correction is that the difference of certain moment S136 sampling assay value of employing and model estimate value is that power carries out.In actual distillation operation, because the sampling chemical examination time is the interior random sampling of section at a fixed time, and for some time (τ) that needs to lag behind just can obtain the assay value, so when model tuning, with per 4 hours assay value (divinyl in the product-1 once, 3 content) deviation between the mean value of soft instrument and in the sampling time section is proofreaied and correct the current y of soft instrument
1Output valve.The correction formula is as follows:
y
out(t)=y
1(t)+α[ y
1.cal(t-τ)-y
1.exp(t-τ)]
In the formula: y
Out(t): the estimated value of t soft instrument correction constantly back divinyl-1,3;
y
1(t): the estimated value of divinyl-1,3 before the t soft instrument correction constantly;
α: correction coefficient, get 0.5 according to actual needs;
y
1.exp(t-τ): divinyl-1, the 3 content laboratory values of taking a sample constantly at t-τ;
y
1.cal(t-τ): the mean value of soft instrument output in this section sample time.
DA107 tower still butadiene content industry soft-sensing model
In the design of tower still butadiene content soft-sensing model, adopted nonlinear multielement recurrence and dynamic compensation hybrid modeling to set up this soft measurement industry pattern.By a large amount of l-G simulation tests, finally obtained the leading variable of still liquid butadiene content soft-sensing model and the relational expression between the subsidiary variable:
y
S135(k)=f(x
1(k),x
2(k),x
3(k),x
4(k),x
5(k),x
6(k),x
7(k))
Y (k)---the concentration of divinyl in the DA107 tower bottoms-1,3;
x
1(k)---liquid level L122 at the bottom of the preceding 10 minutes tower of current time
x
2(k)---tower bottom pressure P119;
x
3(k)---tower still temperature T 300-6;
x
4(k)---the quadratic term of tower still temperature;
x
5(k)---the tower still temperature that current time is preceding 5 minutes;
x
6(k)---the tower still temperature that current time is preceding 10 minutes;
x
7(k)---the product term of tower bottom pressure and tower still temperature.
Owing to not only exist nonlinear relationship between the concentration of divinyl and tower still temperature, pressure, the liquid level in the still liquid, and exist bigger hysteresis phenomenon.Therefore utilize non-linear regression modeling technique and dynamic compensation technology to set up following soft-sensing model:
B0=-2.5283????B1=0.0016????B2=46.0836???B3=-0.4005
B4=0.0049?????B5=0.1952????B6=-0.077????B7=-0.704
The wherein comparison of the On-line Estimation value of quality product soft instrument actual motion and experiment laboratory values, data results has shown soft instrument estimation absolute deviation<0.0015, this quality soft instrument can be followed the tracks of quality product preferably and change.
The design and implementation of 2 advanced control loops of DA107 tower
In order further to stablize and improve the operation of divinyl rectification cell, improve the product butadiene quality, cut down the consumption of energy, improve the whole economic efficiency of butadiene product equipment, this project is used above-mentioned industrial soft-sensing model and the automatic control technology that has realized, develop advanced control loop 3 of DA107 cat head and the advanced control loop 4 of tower still, realized DA107 tower product top grade quality control requirement (divinyl in the product-1,3 purity 〉=0.993kg/kg; Controlled target such as total alkynes content≤50mg/kg) and still liquid divinyl loss minimum.
Utilize the On-line Estimation of quality product divinyl and total alkynes content soft-sensing model, and adopted mixture control strategies such as deduction control based on soft-measuring technique, expert's control, selective control, designed and developed advanced control loop 3 and 4.In this advanced control loop, also utilize the intelligent filter technology that field data is handled.And the control strategy of design has been realized the closed-loop optimization control of quality product by configuration on DCS.Actual motion shows: the control accuracy of product divinyl and total alkynes purity can both satisfy controlling index.The estimated accuracy limit of error of divinyl is limited in ± 0.0005 in; Not only can satisfy the requirement of divinyl control of purity, keep relatively steadily (being controlled in the 99.3%-99.6% scope) of quality control, and avoided " mass excess " phenomenon in the production quality control effectively, improve the yield of divinyl.Show that after tuning test run after a while this Controlling System has reached the edge control effect of quality product.
Economic benefit: after the DA106 cat head is implemented optimal control, with the butadiene content in the tower top discharge tail gas, by the past average 85% be reduced to present 55%, consider that discharging tail gas amount is about 80-100kg/h, the year of device is 8000 hours a working time, and then optimal control system can reclaim the divinyl meter from the DA106 tower: (85%-55%) * 0.1 (T/h) * 8300 (h/y)=250 (T/y); The DA107 cat head is after adopting optimal control, both guaranteed the effective of quality product, steadily control, reduced the surplus of overhead product purity again effectively, relatively improved the output of divinyl, reduce to 99.55% present (the mass concentration standard is 〉=99.3%) by this overhead product butadiene concentration by 99.65% of the mean concns of history, then output divinyl amount be in the past 0.9965/0.9945 doubly, product extraction flow calculates with average 7.0t/h, and the then optimal control system callable divinyl of control loop thus is: (0.9965/0.9945-1) * 7.0 (T/h) * 8000 (h/y)=117 (T/y); The DA107 cat head is after adopting optimal control, owing to reduced the butadiene content in the still liquid, carry secretly in the high boiling material in the still liquid divinyl amount by in the past 12% be reduced to 10%, average discharge can get by 250kg/h: (12%-10%) * 0.25 (T/h) * 8300 (h/y)=42 (T/y).Total economic benefit: the average settlement price of divinyl is by 70500 yuan of/ton calculating, and therefore the direct economic benefit that can be obtained by the optimal control system of DA-106 and DA-107 tower is: (250+117+42) (ton/year) * 0.705 (ten thousand yuan/ton)=288.5 (ten thousand yuan/year).
Social benefit: after implementing the distillation system optimal control, raising by a relatively large margin the technic index of device and the equipment rate that operates steadily, reduce the risk of keeping the safety in production in the technological process; Also alleviated labor intensity of operating personnel relatively.Owing to reduced discharging, also be favourable for environment protection., fast development long-term, stable, healthy for enterprise laid a good foundation.
Five, embodiment
The present invention successfully puts into operation raising on the sub-petro-chemical corporation butadiene production device, has solved the engineering construction problem well, has realized the quality product closed-loop control.This Controlling System has not only guaranteed productive capacity preferably in the steady running of industry spot, and obtains economic benefit thus and can reach 288.5 ten thousand yuan/year.
The present invention has developed the advanced control system that comprises 4 loops raising on the divinyl rectifier unit of sub-petro-chemical corporation (DA106 rectifying tower and DA107 rectifying tower).Employing is based on the deduction control of soft measurement, and the auxiliary control strategy of optimizing and revising with constraint control and set(ting)value.Realized the closed-loop control of the rectifying tower quality product under the manual monitoring.Be described further below in conjunction with accompanying drawing:
Embodiment 1:DA106 Tata still water-content advanced control system
Referring to Fig. 1, the DA106 tower belongs to azeotropy rectification column, is mainly used in the lower-boiling impurity (methylacetylene, saturation water) that removes in the crude butadiene.This tower has 70 blocks of column plates, and the 11st block of column plate is feed tray.Nearly 1200mg/kg saturation water in the charging after DA106 tower and divinyl form the lower boiling azeotrope, distillates from cat head, after the condensation in return tank by sedimentation, two-layer about being divided into naturally.The all phlegmas in upper strata are sent into the DA106 tower entirely, as the backflow of this tower; The free-water that passes through sedimentation formation of lower floor's minute quantity is discharged from the battery limit (BL), and technical process is seen shown in Figure 1.
July 06:15 to 2002 year on the 1st July in 2003, the operation result of 288 hours soft-sensing model of 06:15 on the 11st was as shown in table 1.Wherein soft measurement estimated value is compared with measured result, and maximum absolute error 2.886 (mg/kg), maximum relative error are 17.5%, and mean absolute error 0.9259 (mg/kg), average relative error are 5%.Generally speaking, can follow the tracks of DA106 tower still water-content preferably changes.
Table 1 water-content soft instrument On-line Estimation value and laboratory values are relatively
Time | Laboratory assay value (mg/kg) | Soft instrument estimated value (mg/kg) | Absolute deviation (mg/kg) |
??2002-07-01?06:15:01 ??2002-07-01?22:15:09 ??2002-07-02?06:15:05 ??2002-07-02?14:15:04 ??2002-07-02?22:15:00 ??2002-07-03?06:15:02 ??2002-07-03?14:15:05 ??2002-07-03?22:15:03 ??2002-07-04?06:15:04 ??2002-07-04?14:15:03 ??2002-07-04?22:15:04 ??2002-07-05?06:15:03 ??2002-07-05?14:15:09 ??2002-07-05?22:15:00 ??2002-07-06?06:15:01 ??2002-07-06?14:15:07 ??2002-07-06?22:15:02 ??2002-07-07?06:15:02 ??2002-07-07?14:15:05 ??2002-07-07?22:15:05 ??2002-07-08?06:15:08 ??2002-07-08?14:15:09 ??2002-07-08?22:15:02 ??2002-07-09?06:15:04 ??2002-07-09?14:15:01 ??2002-07-09?22:15:03 ??2002-07-10?06:15:08 ??2002-07-10?14:15:05 ??2002-07-10?22:15:00 ??2002-07-11?06:15:05 ??2002-07-11?14:15:03 ??2002-07-11?22:15:07 ??2002-07-12?06:15:04 | ??20 ??17 ??19 ??19 ??19 ??19 ??19 ??19 ??20 ??19 ??19 ??19 ??19 ??18 ??15 ??15 ??16 ??18 ??16 ??19 ??19 ??18 ??18 ??16 ??18 ??17 ??17 ??16 ??17 ??20 ??19 ??20 ??19 | ??19.5929 ??19.7181 ??19.1049 ??19.2519 ??18.3023 ??18.4110 ??18.5607 ??18.6359 ??19.5897 ??18.9517 ??19.3067 ??19.7491 ??19.2795 ??18.8339 ??16.9873 ??17.26275 ??17.312 ??18.8419 ??18.886 ??18.0106 ??18.0216 ??17.5929 ??16.8618 ??18.0204 ??17.5951 ??17.0076 ??17.566 ??17.4059 ??16.9531 ??17.3922 ??18.4026 ??18.6174 ??18.4873 | ??0.4071 ??2.7181 ??0.1049 ??0.2519 ??0.6977 ??0.589 ??0.4393 ??0.3641 ??0.4103 ??0.0483 ??0.3067 ??0.7491 ??0.2795 ??0.8339 ??1.9873 ??2.26275 ??1.312 ??0.8419 ??2.886 ??0.9894 ??0.9784 ??0.4071 ??1.1382 ??2.0204 ??0.4049 ??0.0076 ??0.566 ??1.4059 ??0.0469 ??2.6078 ??0.5974 ??1.3826 ??0.5127 |
Mean value | ??18.12 | ??18.31 | ??0.9259 |
The elementary object of this scheme control is: guarantee that water-content is less than under the 20mg/kg quality index in the product, make the energy consumption minimum of DA106 tower.In this advanced control subsystem design, at first utilize water-content soft measuring instrument function in the product that the front realized in the DCS system, with the estimated value of soft instrument and the deviation of quality product set(ting)value,, form closed loop Composition Control loop as the input signal of controller.Fig. 2 provides the skeleton diagram of this advanced person's control strategy.
But when fluctuation of operating conditions was big, the variation of quantity of reflux also increased thereupon, made the distillation operation condition of whole tower take place than storm and stress thus, and this is unallowed on technology.Increase selector switch among the figure, exactly in order to realize that reboiler adds the constraint control of heat, to guarantee that quantity of reflux is controlled in the scope of processing requirement (quantity of reflux is 1.3~1.7 with the ratio of inlet amount) by increase quantity of reflux constraint control in inferring the control scheme, prevent that preferably reflux ratio from changing excessive phenomenon, guarantees that the quantity of reflux of this tower is controlled in the given zone of reasonableness.In Fig. 3, provided the comparison that increases constraint control front and back implementation result.
After should inferring-retrain the operation of control scheme, its control effect has satisfied the water-content quality index well, has realized the closed-loop control of tower still water-content quality.Before Fig. 4 a has provided advanced control system and has put into operation actual motion effect (8:00 to 2002 year on the 24th September in 2002 on September, 29,8:00 amounted to 120 hours).Fig. 4 b provided after advanced control system puts into operation actual motion effect (May 8:00 to 2003 year on the 18th May in 2003 8:00 on the 25th, amount to 140 hours), after as can be seen from the figure advanced control system put into operation, water-content changed smoothness significantly better than the operating performance before putting into operation.And water-content is controlled in the controlling index claimed range just (promptly in 17~19mg/kg scope, fluctuates), and this control accuracy satisfied quality control index well, realized the edge control of tower still water-content.
Embodiment 2:DA106 column overhead exhaust emissions amount advanced control system
In the design of this industry soft-sensing model, still adopt above-mentioned soft sensor modeling technology, set up the industrial soft-sensing model of cat head MA content.Fig. 5 has shown the DA106 rectifying tower during on December 09th, 01 day 1 December in 2002 real time execution, DA106 tower tail gas MA concentration soft measuring instrument real time execution effect.Soft-sensing model is succinctly practical, and the on-line correction reliable in function is effective.Control scheme advanced design, practical reliable reaches obviously and cuts down the consumption of energy, and improves product yield and guarantees the effect of operating safety.
The key that this advanced person controls conceptual design is the cat head exhaust emissions amount that calculates an optimum, because this quantity discharged is crossed the loss that conference causes the target product divinyl, but if the emptying amount is low excessively, might cause the methylacetylene excessive concentration and the accident of blasting, this quantity discharged can guarantee that the methylacetylene amount of DA106 tower top discharge satisfies the safety in production index.Based on MA content soft measuring instrument On-line Estimation value in the above-mentioned DA106 cat head tail gas, study, developed DA106 cat head exhaust emissions amount and infer Controlling System (advanced control subsystem 2).This system is according to the estimated value of MA content in the DA106 cat head tail gas, by inferring the set(ting)value of the online given tower top discharge amount controller of controller, the concentration of MA in the cat head distillate gas is controlled in the optimum range, make that the divinyl loss is reduced to minimum degree in the tail gas, thereby improve the yield of product.Its control scheme is seen Fig. 6.
Behind the input actual motion of The whole control loop, receive good economic benefits.Shown in the lower curve among Fig. 7 DA106 tower top discharge amount F135.Before this control loop that do not put into operation, F135 is bigger than normal always, often is in the situation of outrange.By implementing this control scheme, this throughput ratio of F135 is controlled at a lower level more reposefully, not only can satisfy technic index but also reduce the loss of divinyl significantly.Tool moves the divinyl yield that this Controlling System can reach 50kg/h according to a preliminary estimate.
The advanced control subsystem of embodiment 3:DA107 tower quantity of reflux
The soft-sensing model of divinyl and total alkynes content in the DA107 column overhead product has been set up in the technological operation variable analysis of producing according to device and the process mechanism mathematical model simulation research of DA107 rectifying tower.This project development product quality soft-sensing model has higher precision, lays a good foundation thereby realize the advanced control of quality product closed loop for next step.Fig. 8 has shown the DA107 rectifying tower during on October 1,16:15 to 2002 year on the 20th September in 2002 16:15 real time execution, DA107 column overhead quality product divinyl-1,3 concentration soft measuring instrument real time execution effect.Table 2 has provided the On-line Estimation value of quality product soft instrument actual motion and the comparative result of laboratory values, and data results has shown soft instrument estimation absolute deviation<0.0015, and this quality soft instrument can be followed the tracks of quality product preferably and change.
The On-line Estimation value of table 2 soft instrument and laboratory values are relatively
Absolute deviation | ??<0.0001 | ??<0.0005 | ??<0.001 | ??<0.0015 | ??>0.0015 |
(sum 200) | ??122 | ??45 | ??21 | ??8 | ??2 |
Frequency (%) | ??61 | ??22.5 | ??10.5 | ??4 | ??1 |
In the design of the advanced control strategy of DA107 tower quantity of reflux (F139), adopted deduction control and expert to control the mixture control strategy that combines based on soft-measuring technique, be applied to multiple complicated control strategy simultaneously, as selective control, tandem control etc.This system in this control subsystem, also utilizes the intelligent filter technology that field data is handled according to the estimated value of overhead product butadiene concentration.Fig. 9 has provided the sketch of should the advanced person controlling scheme.
In the DeltaV of Fisher-Rosemount company DCS system, behind the commissioning test by a stage, prove that this scheme can actual at the scene preferably smooth running, and satisfied product quality indicator preferably, realized the control of quality product closed loop edge.Before Figure 10 a has provided advanced control system and has put into operation actual motion effect (8:00 to 2002 year on the 20th September in 2002 on October 1,8:00, amount to 240 hours) and Figure 10 b provided after advanced control system puts into operation actual motion effect (May 8:00 to 2003 year on the 18th May in 2003 8:00 on the 25th, 160 hours altogether).From Figure 10 a, 10b contrast as seen, this advanced control strategy of operation has been obtained excellent control effect on device.
Actual motion shows: the control accuracy of product divinyl and total alkynes purity has reached controlling index, the estimated accuracy limit of error of divinyl-1,3 is limited in ± 0.0005 in; Not only can satisfy the requirement of divinyl control of purity, keep relatively steadily (being controlled in the 99.3%-99.6% scope) of quality control, and avoided " mass excess " phenomenon in the production quality control effectively, relatively improve the yield of divinyl.Show that after tuning test run after a while this Controlling System has reached the edge control effect of quality product.