CN100520660C

CN100520660C - Fast stabilizing temperature control method based on characteristic information

Info

Publication number: CN100520660C
Application number: CNB2005100400999A
Authority: CN
Inventors: 刘奇志
Original assignee: Nanjing University
Current assignee: Nanjing University
Priority date: 2005-05-19
Filing date: 2005-05-19
Publication date: 2009-07-29
Anticipated expiration: 2025-05-19
Also published as: CN1719367A

Abstract

The present invention utilizes several dynamic real-time in formations in the control system to control temperature. It includes the following steps: firstly, utilizing two basic characteristic variable quantities of error and error change to produce a group of derivative characteristic information; then utilizing derivative characteristic information to design the control rules of switching, proportion, annealing and retention, etc.; computer utilizes real-time collected data and data in the data base to make characteristic identification, and selects most effective control rule from rule library to act on actuator.

Description

Fast stabilizing temperature control method based on characteristic information

Technical field

The present invention relates to industrial process control field, a kind of method of utilizing various features information to carry out the control of temperature fast and stable is provided.

Background technology

In industrial control system, temperature is one of most important and the most difficult controlled variable.

Controlled temperature system has hysteresis quality (inertia), can be with the order transfer function approximate description with pure-time-delay.Existing temperature-controlled process has PID control (linear combination of ratio, integration, differential control) and multiple improvement project thereof.For the second order controlled system that does not have mathematical models, by the scale-up factor K in the suitable selection PID controller _P, integral coefficient K _IWith differential coefficient K _DThree parameters can obtain satisfied response characteristic to a certain extent.Proportional control is the basic means of accurately controlling, and integration control is used for eliminating static difference, and differential control is in order to accelerate system response time.But have contradiction between stability and the accuracy in this method, parameter tuning bothers very much, and system often will experience long transit time and just can enter and stablize control zone or the approximate control zone of stablizing.For the second order controlled system that mathematical models is arranged, the zero limit of PID controller transfer function can offset with the zero limit of controlled system transport function, so the control effect is very desirable; At delay system, also have some PID improvement projects based on mathematical models structure compensation tache or formula, as Smith (Smith, 0.J.M.) predictive compensation scheme, Da Lin (Dahlin, E.B.) control algolithm, Kalman (Kalman, R.E.) control algolithm and V.E. (Vogel, E.F.﹠amp; Edgar, T.F.) control algolithm etc., they can further improve the control effect to some extent.But because the complicacy of real system, the mathematical models of controlled system is difficult to obtain, so said method is difficult to use in actual engineering.

In the practical application, the engineering staff usually adopts switch control (bang-bang control) method to control temperature, and this control model simply is easy to realize.But the control output quantity of ordinary tap control or be a fixed constant, or be zero, its precision is not high, and stability is bad.

The artificial intelligent method of controlling switch is a kind of intelligent temperature-controlled process, it imitates artificial process, regulate " dutycycle " of control output time according to the trend of error and error variation and control temperature, be about to a control cycle and be divided into control output (opening) and do not export (pass) two stages, regulate the ratio of shared time in two stages and regulate temperature with control.But single artificial intelligent method of controlling switch precision is still not high, and still has many parameters to adjust.

Also have a kind of apery intelligent control method based on the extreme value sampling, this method is divided into a plurality of stages with control cycle, and different phase adopts different control methods, comprises proportional control, suppresses control and retentive control.Retentive control replaces integral action, has eliminated phase lag and integration saturation problem that integral action is brought effectively, suppresses control with the contradiction between the stability that solves control system and accuracy, the rapidity.This method is crossed extreme point in error and adopt inhibition and retentive control when the direction that reduces is changed, if the error ratio of extreme values is bigger, then needs the regular hour just can carry out the transition near the expectation value, has reduced the performance of transit time and these two indexs of overshoot time; In addition, being changed to 0 with error concludes and extreme point and unreliable occurs, because in computer control system, characteristic information is actual to disperse, take place very likely that extreme point has occurred but error to change be not 0, also useful adjacent twice error variation product is less than 0 as criterion, this is also unreliable, because the double sampling value is recurrent in the identical situation in the extreme point left and right sides, change product less than 0 if take place then can not get adjacent twice error.

No. 03100391.5 application documents of Chinese patent disclose " than the control method of large time delay temperature system ", it is a kind of control mode at intermittence of specific period, its essential idea is that switch control is Bang-Bang control, and this simple control method is fit to single thermostatic control occasion.Because the temperature variation characteristics are that programming rate is far longer than cooling rate, so require the occasion high, that operating mode is complicated at temperature control precision, Bang-Bang control is difficult to be competent at.

No. 03154473.8 application documents of Chinese patent disclose " digital intelligent temperature-controlled process and temperature control semiconductor laser ", provide a kind of digital intelligent algorithm to convert detected temperature data to control data, again control data is converted to the impulse ratio signal, open and shut off according to certain dutycycle and frequency with this Signal-controlled switch element, and then the control temperature.Obviously its essence is artificial intelligent switch control mode, be not suitable for operating mode complex engineering system.

U.S. Pat 60/266, No. 926 application documents disclose " inertia temperature control system and method ", a kind of temperature control algorithm that is used for heating furnace is provided, and this algorithm makes temperature with the limited rapid change that reaches by the acceleration and the deceleration of control temperature slope set-point.This algorithm comes down to a kind of artificial intelligent proportional controlling means, promptly constantly adjusts set-point, makes output constantly approach expectation value, thereby improve the systematic steady state precision, can effectively avoid overshoot, relatively be fit to linear stational system, should not be in the omnidistance use of control procedure.

No. 237678/92 application documents of Jap.P. JP disclose a kind of temperature-controlled process, its temperature control need obtain the acceleration of temperature increase and decrease on each time interval, and the load factor of heating part in each time interval is revised by the acceleration of the temperature in previous step increase and decrease, control the temperature of above-mentioned heating part by this load factor, further, in temperature increase process, close well heater, the time point of predicted temperature through for the moment lagging behind and begin to descend reopened well heater before estimating the time point that temperature drops to setting value.This is actually a kind of switch control mode, is not suitable for the high occasion of accuracy requirement.

No. 53133/2002 application documents of Jap.P. JP, relate to a kind of image heating and temperature-controlled process, this temperature control purpose is that the roll coil of strip when preventing high temperature is paid cylinder, stop to make by the heating of heating cylinder to band by the energising from the field circuit to the field coil stopping, the design temperature of band is inferred by a plurality of detection variable.This temperature-controlled process specificity is strong, and with regard to its control method itself, pattern is single.The relevant described temperature-controlled process of patent application document of other temperature control that retrieves also all is concrete grammar and the measure at a certain special application scenario basically.

Summary of the invention

The present invention seeks to: solve the technical matters overcome the slow and poor stability of temperature control system response speed in the Industry Control.

The present invention also aims to: the derivation amount of utilizing error and error to change is controlled temperature, overcomes the technical matters that makes full use of that present temperature-controlled process is ignored real-time dynamic information in the control system.

The object of the present invention is achieved like this: realize that technical scheme of the present invention is summarized as follows:

1, based on the fast stabilizing temperature control method of characteristic information, utilize in the control system multiple dynamic real-time information to control temperature, comprise employing the following step: at first, utilize e (error) and

(error variation) these two characteristic variables the most basic are constructed one group of derived character information; Utilize control laws such as derived character information design switch, ratio, annealing, maintenance then; Computing machine carries out feature identification according to the data in real-time image data and the database, the dynamic behaviour of recognition system, and inference machine is chosen the most effective control law according to identification result from rule base, act on actuator, and controlled device is controlled accurately.

(1) when Error Absolute Value is very big, adopt local switch control, control law is:

If e〉0 and | e| 〉=M, then U=0

If e＜0 and | e| 〉=M, then U=UM

Wherein, M is given threshold value, and U is control output, and UM is a constant, expression output maximal value (standard-sized sheet);

(2) when Error Absolute Value is bigger, adopt proportional control, control law is:

If δ＜| e|≤M, then U=u _H+ K _pe

Wherein, δ is given threshold value, K _pBe scale-up factor, u _HFor exporting retention value last time;

(3) if Error Absolute Value is smaller, but the product that error and error change 0, also carry out proportional control;

(4) smaller and sum of errors error changed product less than 0 o'clock in Error Absolute Value, adopted retentive control, and control law is:

If (| e|≤δ and

e \cdot \dot{e} < 0

) or e=0, THEN U=u _H

If error is 0, then no matter the product that error and error change whether less than 0, also carries out retentive control;

(5) in Error Absolute Value smaller and this error change with last time error change product and be less than or equal at 0 o'clock, adopt annealing control, control law is:

If (0＜| e|≤δ and

{\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} \leq 0

), U=u then _H+ (K _pK _d) e

Wherein, K _d＜1 is annealing coefficient.

2, based on the enhancing stabilizing temperature control method of characteristic information, under above-mentioned (1), (4) condition, revise scale-up factor, for next stage control is prepared:

(1) when error is very big, the recording error peak value goes out now, and error in judgement peak intervals time t _EmAbsolute value with adjacent twice error peak ratio | e _Mi-1/ e _Mi|, if the absolute value of error peak weak point interval time or error peak ratio then reduces proportional action, that is: less than 1

If | e| 〉=M and (t _Em＜T or | e _Mi-1/ e _Mi|＜1), K then _p=K _dK _p

Wherein, τ is threshold value (a given constant), K _d＜1 is annealing coefficient.

(2) and absolute value that error change very big when error is less than δ ', and proportional action can increase by a relatively large margin so, that is:

If (| e| 〉=M and

| \dot{e} | < δ^{,}

), K then _p=K ₁K _p, wherein, K ₁1.

(3) when error is 0, the zero passage speed of recording error ,, then reduce proportional action and (use K if zero passage speed is excessive _pMultiply by K _dAs new K _p, K _dBe annealing coefficient).

3, based on the anti-interference control method of characteristic information, judge multidate information, when the absolute value of adjacent twice error peak ratio and adjacent twice error zero passage speed ratio all less than 1 the time, adopt the control of the one-period of lasting to export.

Derived character information and the implication that the present invention relates to are described as follows:

The absolute value of [| e|] error.With | e|＜δ, | e|〉M can show that system is near the expectation value or near the error peak.Wherein δ, M are threshold value (given constant), δ＜M, and M〉absolute value of permissible error threshold value.

The product that error and error change.Can judge the trend that the system dynamic course error changes, if

e \cdot \dot{e} < 0

Then be away from expectation value, if

e \cdot \dot{e} < 0

Then be near expectation value.

Adjacent twice error changes long-pending.Show whether error extreme value occurs, if

{\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} \leq 0

Show extreme value to occur, if

{\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} > 0

Show no extreme value.

[e _Mi] this error peak.

[u _H] early stage output quantity retention value.Remember this value, can be used as the reference of output next time, if system stability can continue to keep this output, and needn't recomputate.

[ ] the i time zero passage speed of error.This speed is the smaller the better, because little illustrative system tends towards stability, greatly then explanation is tending towards vibration or shake.

[t _Em] error peak interval time.The big more illustrative system of this value oscillation period long and error peak be used in combination can illustrative system stability, the little and t of error peak _CmBig illustrative system good stability, otherwise only error peak is little, and system might be also in shake.

The absolute value that error changes.With

| \dot{e} | < δ^{,}

With

| \dot{e} | > M^{,}

The error pace of change that can show system is slowly or rapidly.δ ', M ' are threshold value.

Adjacent twice error zero passage speed ratio.If

{\dot{e}}_{0 i - 1} / {\dot{e}}_{0 i} >

Illustrate that the control effect takes a turn for the better,

{\dot{e}}_{0 i - 1} / {\dot{e}}_{0 i} < 1

Then the overshoot of system might than preceding once increase and | e _Mi-1/ e _Mi| be used in combination the confidence level that can increase reasoning, because undesired signal also can influence | e _Mi-1/ e _Mi| value.Characteristics of the present invention are: the characteristic information of having constructed abundant controlled system, improved the transparency of controlled system, can grasp characteristics such as the inertia of error Changing Pattern and controlled device, pure hysteresis and disturbance in real time, according to these rules and characteristic, select different control strategies, realize the fast and stable control of temperature.Can solve existing temperature control parameter adjust difficulty, the transitional period is long, response speed is slow and the problem of poor stability.

The present invention has cancelled differential and integral action, does not have the headache problem of balance compromise between the Several Parameters, is easy to realize.Carry out the aggregative model control that switch, ratio, annealing and open loop keep by various features information, adapt to the continuous variation of controlled system by continuous change control strategy, have the self-adaptation characteristic, and intelligence enhancing property good (but as long as increasing the just control level of intelligence of enhanced system of suitable characteristic information).When error is big, use switch control can reduce error fast.Its proportional action can be considerably beyond the proportional action in other control, because system's output is in case near reference point, change scale-up factor immediately, the reduction output quantity, need not be because of worrying overshoot the consideration scale-up factor of compromising, so shortened rise time and transit time greatly.The maintenance pattern is a kind of open loop approach, and it can reduce static difference even floating, and is effective especially to pure hysteresis industrial system.By characteristic information integrated forecasting error change direction, accelerate system response time, and distinguish the measured signal of undesired signal and system itself.In systematic error hour, adopt the annealing pattern, reduced overshoot, thereby increased stability.Switch to by closed loop open loop keep pattern, by open loop switchback closed loop, even more complex multi-mode formula control all utilizes characteristic information to drive switching, reliability, accuracy that the information that makes information processing rate improve, need to handle reduces, switches increase.

Description of drawings

Fig. 1 is a control system structural drawing of the present invention

Fig. 2 is the dynamic error curve synoptic diagram 1 of electric furnace list temperature control system of the present invention

Fig. 3 is the dynamic error curve synoptic diagram 2 of electric furnace list temperature control system of the present invention

Fig. 4 is a curve of output synoptic diagram 1 of the present invention: wherein 1---switch control, and 2---proportional control, 3---annealing control, 4---retentive control

Fig. 5 is a curve of output synoptic diagram 2 of the present invention

Embodiment

Below in conjunction with accompanying drawing, mode of the invention process is described:

Suppose an electric furnace list temperature control system, require control system the stable setting value R that is controlled at of controlled device temperature Y.The online main source that obtains information of computing machine is exactly R and Y, also can obtain data from real-time or historical data base, and (Y-R), error change therefrom can to calculate error e

And the characteristic information of other characterization system dynamic perfromance.Computing machine is by means of these characteristic informations, and the dynamic behaviour of recognition system is as the foundation of control decision.According to the behavioral characteristics and the dynamic behaviour of system, from rule base, choose the most effective control law, controlled device is controlled accurately.

Control method comprises ^*Might as well establish control output (opening) controlled volume is increased, (pass) do not exported in control reduces controlled volume.：

One, when Error Absolute Value is very big, adopts local switch control.Its control law is:

IF e>0 AND|e|≥M，THEN?U＝0

IF e<0 AND|e|≥M，THEN?U＝UM

Wherein, U is control output, and UM is a constant, expression output maximal value (standard-sized sheet).In addition, most actuators are dragged by motor, and it is 0 can burn out motor that motor passes speed, so during actual the use, should change the output in the rule into a smaller constant by 0, the protection actuator.

Two, when Error Absolute Value is bigger, adopt proportional control.Its control law is:

IFδ < | e | \leq MOR (| e | \leq δANDe \cdot \dot{e} > 0), THENU = u_{H} + K_{p} e

Wherein, K _pBe scale-up factor.If Error Absolute Value is smaller, but the product that error and error change 0 (controlled volume is away from expectation value), so also carry out proportional control.

When three, smaller and controlled volume is near expectation value in Error Absolute Value, adopt retentive control.Its control law is:

IF (| e | \leq δ                                AND e \cdot \dot{e} < 0) OR e = 0,

THEN?U＝u _H

If error is 0, then no matter the product that error and error change whether less than 0, also carries out retentive control.

Four, smaller and when extreme point occurring (controlled volume begins perhaps to begin by changing near expectation value away from expectation value by changing into away from expectation value near expectation value) in Error Absolute Value, adopt annealing control.Its control law is:

IF (0 < | e | \leq δAND {\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} \leq 0),

THEN?U＝u _H+(K _p·K _d)e

Wherein, K _d＜1 is annealing coefficient.

Five, under above-mentioned one, three situations, revise scale-up factor, for next stage control is prepared, to improve the stability of control.

1) when error is very big, the recording error peak value goes out now, and error in judgement peak intervals time t _EmAbsolute value with adjacent twice error peak ratio | e _Mi-1/ e _Mi|, if the absolute value of error peak weak point interval time or error peak ratio then reduces scale-up factor K less than 1 _p, that is:

IF|e|≥M?AND(t _em<τOR|e _mi-1/e _mi|<1)，THEN?K _p＝K _d·K _p

2) and absolute value that error change very big when error is less than δ ', and scale-up factor can increase by a relatively large margin so, that is:

IF (| e | &GreaterEqual; M  AND | \dot{e} | < δ^{,}),

THEN?K _p＝K ₁·K _p

Wherein, K ₁1.

3) when error is 0, the zero passage speed of recording error

,, then reduce scale-up factor K if zero passage speed is excessive _p, that is:

\begin{matrix} IF & {\dot{e}}_{0 i} > ϵ \end{matrix},

THEN?K _p＝K _d·K _p

Wherein, ε is threshold value (a given constant), K _d＜1 is annealing coefficient.

Six, on the basis of said method five, increase anti-interference control law:

When the absolute value of adjacent twice error peak ratio and adjacent twice error zero passage speed ratio all less than 1 the time, adopt the control strategy of last one-period, that is:

IF (| e_{mi - 1} / e_{mi} | < 1 AND {\dot{e}}_{0 i - 1} / {\dot{e}}_{0 i} < 1) THEN U_{i} = U_{i - 1}

The following control law that we will formulate deposits in the rule base as shown in Figure 1.

IF δ < | e | \leq M OR (| e | \leq δANDe \cdot \dot{e} > 0),

THEN U=u _HProportional control

IF (| e | \leq δANDe \cdot \dot{e} < 0) OR e = 0,

THEN U=u _HRetentive control

IF (0 < | e | \leq δAND {\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} \leq 0),

THEN U=u _H+ (K _pK _d) annealing control

IF {\dot{e}}_{0 i} > ϵ, THEN K_{p} = K_{d} \cdot K_{p}

IF (| e_{mi - 1} / e_{mi} | < 1 AND {\dot{e}}_{0 i - 1} / {\dot{e}}_{0 i} < 1) THEN U_{i} = U_{i - 1}

Anti-interference control

The present invention includes as lower member: rule base, inference machine, feature identification, parameter correction, database and actuator.

Parameter in the above-mentioned rule can be revised automatically by the parameter correction device, or manual modification, and rule itself also can manual modification.

In actual use, above-mentioned rule converts the expression formula or the statement of programming language form to, the Be Controlled routine call, and commander's actuator acts on controlled device.

Emotionally the dynamic error curve under the condition is as shown in Figure 2 at unperturbed for above-mentioned electric furnace list temperature control system.Error is at t ₁And t ₃Constantly reach positive and negative error peak respectively, at t ₂And t ₄Constantly be 0, stop heating in the above zone of M dotted line, at-M dotted line with the lower area heating with full power, between M dotted line and the δ dotted line ,-the M dotted line and-the δ dotted line between, and constantly 0 and t ₄Between back δ dotted line and the horizontal ordinate, t constantly ₂Proportional control is adopted in zone between back-δ dotted line and the horizontal ordinate, constantly t ₂And t ₄Before δ ,-retentive control is adopted in zone between δ dotted line and the horizontal ordinate.If graph of errors is very little at longitudinal axis upper span, as shown in Figure 3, then at t ₁And t ₃Adopt annealing control constantly.

Conventional PID control curve of output down as shown in Figure 4, the curve of output under the described control model of employing this patent is as shown in Figure 5.The control effect is obviously improved.Adopt different control models to further specify to zones of different among Fig. 4.

Control method can be carried out multi-form combination and be applied as the case may be.

More than the disclosed temperature-controlled process that relates to is specifically described, those skilled in the art can understand and implement, do not departing under the scope of the invention situation, can carry out the various modifications of form and details to control law, such as increasing anti-interference control law: IF|e for ringing _n| | e _N-1| AND|e _N-1| e _N-2| ANDe _nE _N-10 AND e _N-1E _N-20, THEN U (k)=U (k-1)+(K _pK _D) e, K _D1, to eliminate static difference.Therefore above advised but the modification that do not limit all within the scope of the present invention.

Claims

1, based on the fast stabilizing temperature control method of characteristic information, it is characterized in that utilizing in the control system multiple dynamic real-time information to control temperature, be controlled at setting value R with controlled device temperature Y is stable, therefrom calculating error e is that Y-R, error change

Be de/dt; Comprise employing the following step: at first, utilize error e and error to change

These two characteristic variables the most basic are constructed one group of derived character information; Utilize derived character information design switch, ratio, annealing, retentive control rule then; Computing machine carries out feature identification according to the data in real-time image data and the database, the dynamic behaviour of recognition system, and inference machine is chosen the most effective control law according to identification result from rule base, act on actuator, and controlled device is controlled accurately:

(1) when error is very big, adopt local switch control, control law is: when Error Absolute Value was very big, error was that canonical is closed actuator, error is exported with peak power for negative then actuator, that is:

If e〉0 and | e| 〉=M, then U=0; If e＜0 and | e| 〉=M, then U=UM

Wherein, M is given threshold value, and U is control output, and UM is a constant, expression output maximal value;

(2) when error ratio is big, adopt proportional control, control is output as output last time and several times error sum, that is:

If δ＜| e|≤M, then U=u _H+ K _pe

(3) if error is smaller, but the product that error and error change 0, also carry out proportional control;

(4) less and sum of errors error changed product less than 0 o'clock in error ratio, adopted retentive control, and control law is:

If | e|≤δ and

e \cdot \dot{e} < 0,

Or | e|≤δ and e=0, then U=u _H

(5) less and controlled volume begins by changing into away from expectation value near expectation value in error ratio, perhaps begin when changing into away from expectation value near expectation value, adopt annealing control, control law is: when error ratio less, and this error changes and error variation last time product is less than or equal at 0 o'clock, be output as output valve and several times error sum last time, the multiple of error multiply by one less than 1 annealing coefficient, that is: on above-mentioned scale-up factor basis here

If 0＜| e|≤δ and

{\dot{e}}_{n} \cdot {\dot{e}}_{n - 1} \leq 0,

U=u then _H+ (K _pK _d) e

Wherein, K _dBe annealing coefficient.

2, fast stabilizing temperature control method according to claim 1 is characterized in that under above-mentioned (1), (4) condition, revises scale-up factor:

If | e| 〉=M and t _Em＜τ or | e| 〉=M and | e _Mi-1/ e _Mi|＜1, K then _p=K _dK _p

Wherein, τ is a threshold value, and threshold value is given constant, K _d＜1 is annealing coefficient;

(2) and absolute value that error change very big when error is less than δ ', and proportional action increases by a relatively large margin so, that is:

If | e| 〉=M and

| \dot{e} | < δ^{,},

K then _p=K ₁K _p, wherein, K ₁1;

(3) when error is 0, the zero passage speed of recording error

If zero passage speed is excessive, then reduce proportional action; Use K _pMultiply by K _dAs new K _p, K _dBe annealing coefficient.