CN103048923A - Consistency restraining method for industrial trough liquid level cascade predictive control system - Google Patents

Abstract

The invention discloses a consistency restraining method for an industrial trough liquid level cascade predictive control system. The consistency restraining method comprises the steps of: 1, respectively establishing a control structure model of an external loop PFC (Power Factor Correction) controller and an internal loop PID (Proportion Integration Differentiation) controller; 2, judging whether the external loop PFC controller and the internal loop PID controller are consistent in restriction; 3, pre-loading the restriction of the internal loop PID controller on an external loop PFC controller by adopting a restriction backspacing algorithm, obtaining a new restriction condition set during optimal computation of the external loop PFC controller; and 4, carrying out optimal computation by the external loop PFC controller under the new restriction condition set, obtaining a set value of the internal loop PID controller and providing the set value to the internal loop PID controller for tracking the set value. According to the invention, the problems of bottom layer loop controller saturation and inconsistency in restriction of an upper layer and a lower layer in a PFC-PID cascade control loop are solved, through the restriction backspacing computation, the set value obtained through optimization is feasible for a bottom layer loop, the control performance of the system is improved, and the economic benefit of a production unit is increased.

Description

The consistency constraint method that is used for service water tank liquor bit string level Predictive Control System

Technical field

The present invention relates to a kind of consistency constraint method, especially a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System.

Background technology

In the industrial processes, because the physical characteristics of topworks and process units is limit, such as the volume of the regulations speed of the linear work range of valve, valve, oil tank and liquid level restriction etc., there is the operability of certain limit in controller, and requires the output of process in certain safe range.Simultaneously, in order under the prerequisite that guarantees product quality, to maximize Business Economic Benefit, usually wish in certain probability constraints scope, to make as much as possible operating point near the constraint limit, realize the edge operation.Therefore, according to the mandatory and optimization of constraint, usually set hard constraint and soft-constraint two-stage constraints policy.The basic demand of production operation is to guarantee each variable without prejudice to the restriction of hard constraint, otherwise the emergency measures such as the stopping production of will taking to stop are with the generation of Accident prevention.In case stop, this will lose millions of units even more economic benefit for the flow industry enterprise of working continuously.Therefore, enterprise usually will stop and process as industrial accident, and avoid as far as possible its generation.For the restriction of soft-constraint, if the production run operating point transfinites, then can report to the police to remind the operation slip-stick artist that production status is regulated.

For traditional PID control program, the slip-stick artist is according to knowhow in operation, guarantees that operating point is without prejudice to the restriction of hard constraint.Operation slip-stick artist's knowhow and very large to the influential effect of control to the understanding of production run.For soft-constraint and economic performance, owing to lack necessary deterministic models information, be difficult to guarantee its optimization, common way is for guaranteeing that security is take the loss economic benefit as cost, therefore the operating process meeting is fluctuateed in a larger scope, and reserves larger backoff value apart from the operation soft-constraint.

PREDICTIVE CONTROL (Model Predictive Control is hereinafter to be referred as MPC) had been widely used in the process industries such as oil, chemical industry, papermaking, pharmacy in the last few years as the Typical Representative of Advanced Control Strategies.One of key property of PREDICTIVE CONTROL is to process easily restricted problem.The process model that obtains according to the identification to production system, following the output of process is predicted, simultaneously the output of process and model prediction output is fed back to controller and be optimized control, improve the control effect, the compression process output pulsation, and then guaranteeing under the prerequisite of hard constraint, allow the soft-constraint in certain probable range to transfinite, make operating point further limit to realize bounder control, maximum economic benefit near constraint.Therefore can say that PREDICTIVE CONTROL depends primarily on two aspect factors for the raising of economic benefit: the first, improve the control effect to reduce fluctuation; The second, rationally process and keep out of the way problem so that operating point is limit near constraint.This shows that no matter for the security of producing, still for the raising of Business Economic Benefit, rationally processing procedure constraint all is important problem in the process industry.

Predictive function control (Predictive Functional Control is hereinafter to be referred as PFC) is called as third generation predictive controller.Because having algorithm, it simply, does not need to process the characteristics such as quadratic form optimization problem, calculated amount are little, therefore can embed easily DCS (Distributed Control System, hereinafter to be referred as DCS) or PLC (Programmable Logic Controller, hereinafter to be referred as PLC) in, not only can treatment scheme industry wait the slow-response process, simultaneously also effectively be applied to quick control procedure, such as robot, aircraft, weapon control system etc.In the process industries such as oil, chemical industry, than general quadratic form predictive control algorithm, PFC can avoid finding the solution the quadratic form optimization problem, and its control structure is simple, thereby more is close to the PID controller and can directly applies to underlying basis loop dynamic control.

PFC-PID tandem control loop is consistent with the control structure of general MPC-PID tandem control loop in being usually used in industrial process.Because the pid control algorithm that the control program in the process industry more than 95% all adopts and the applicating history that decades even last 100 years have been arranged, adopt MPC (PFC)-PID tandem control structure to take into account compatibility and optimality, thereby accepted by industrial enterprise and operation slip-stick artist easilier.

The characteristics of this class PFC-PID tandem control loop are, the object of inner looping PID controller and the control of external loop pfc controller is same process, and this is a kind of typical control program in industrial process, such as mistake! The bookmark self reference is invalid.Shown in.For the external loop controller, the effect of inner looping controller is " transparent ", therefore claims this tandem to be controlled to be " conduction control ".

Liquid level control is typical process control problem in the commercial production, and liquid level is measured accurately and effectively controlled is the important indicator of some equipment good quality and high outputs, low consumption and safety in production.For water level system generalized object, adopt one order inertia to add pure hysteresis and come Approximate Equivalent, consider that its time constant and retardation time are all larger, so adopt strong robustness, be easy to the Predictive function control strategy of engineering construction.Service water tank liquor level controlling system mainly comprises water tank control module, DCS control rack and host computer configuration and monitoring environment.Design PFC-PID conduction control loop mainly comprises external loop pfc controller and inner looping PID controller under this system, and its control flow as shown in Figure 2.Inner looping adoption rate P controls the processing procedure dynamic perfromance, and external loop adopts without the PFC control strategy of explicit integration device and follows the tracks of setting value.External loop is calculated as inner looping by optimization setting value is provided, and inner looping is followed the tracks of this setting value by dynamically controlling.But, owing to the constraint inconsistence problems between the loops at different levels, be difficult to guarantee correctness and the feasibility of external loop optimum results.For this problem, the sandwich construction Predictive function control is retrained consistency analysis, the rollback calculating by loop in series is passed to external loop step by step with the inner looping constraint, thereby guarantees that the Optimal Setting value is practical for basic loop dynamic control.Verified the raising of constraint rollback computing method for the control effect by simulation analysis.

Summary of the invention

In order to overcome the deficiencies in the prior art, the invention provides a kind of for the bottom loop control unit that occurs in the PFC-PID tandem control loop saturated and levels constraint inconsistence problems, guarantee that the external loop controller calculates correct optimal setting, prevent because phenomenons such as the fluctuation that erroneous calculations causes, overshoot, improve control performance, improve the consistency constraint method that is used for service water tank liquor bit string level Predictive Control System of productive unit economic benefit.

Technical solution of the present invention is: a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System may further comprise the steps:

1) sets up the control structure model of the external loop pfc controller that is used for service water tank liquor bit string level Predictive Control System and the control structure model of inner looping PID controller;

2) judge whether the external loop pfc controller retrains consistent with inner looping PID controller, if unanimously carry out step 3), if inconsistent, then return step 1) rebulid the control structure model of external loop pfc controller and the control structure model of inner looping PID controller;

3) retrain in the consistent situation at external loop pfc controller and inner looping PID controller, adopt the constraint backoff algorithm, the constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and acquisition external loop pfc controller is being optimized the stylish constraint condition set of calculating;

4) the external loop pfc controller is optimized calculating under new constraint condition set, obtains the setting value of inner looping PID controller, and this setting value is offered inner looping PID controller, makes inner looping follow the tracks of this setting value.

Further be set to, in step 1) in, set up the control structure model of the external loop pfc controller that is used for service water tank liquor bit string level Predictive Control System and the control structure model of inner looping PID controller,

The control structure model of described external loop pfc controller is

$u_{\mathrm{ext}}\left(k\right)=\frac{({\mathrm{SP}}_{\mathrm{ext}}-y_p\left(k\right))(1-{{\mathrm{\λ}}_{\mathrm{ext}}}^h)-y_m\left(k\right){a_{\mathrm{mext}}}^h+y_m\left(k\right)}{K_{\mathrm{mext}}(1-{a_{\mathrm{mext}}}^h)}---(1-1)$

In the formula, subscript p represents process;

u _Ext(k) be the constantly output of external loop PFC prediction function controller of k;

SP _ExtSetting value for external loop PFC prediction function controller;

y _p(k) and y _p(k+h) represent respectively the k moment and the constantly output of service water tank liquor bit string level Predictive Control System of k+h;

Expression k is deviation ε constantly _Ext(k)=SP _Ext-y _p(k) to k+1 moment deviation ε _Ext(k+1)=SP _Ext-y _p(k+1) attenuation coefficient;

T _sBe the sampling period;

CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

Be external loop PFC Predictive function control model attenuation coefficient;

T _mBe the model time constant;

K _MextBe model gain;

Make inner looping adoption rate P control, establishing controller gain is K _Int, then the control structure model of described inner looping PID controller is

u _int＝(SP _int-y _p(k))K _int (1-2)

In the formula, u _IntOutput for the inner looping controller;

SP _IntSetting value for inner looping;

y _p(k) be the constantly output of service water tank liquor bit string level Predictive Control System of k;

K _IntBe the inner looping controller gain.

Further be set to, in step 2) in, it is consistent to judge whether external loop pfc controller and inner looping PID controller retrain, and adopts following methods:

1. inner looping adoption rate P when control, for the inner looping proportional control rule of control structure model (1-2) expression of inner looping PID controller, the setting value SP ' of an inner looping is set artificially _Int, and make this setting value SP ' _Int＞SP _Int, then corresponding inner looping control law is

u′ _int＝(SP′ _int-y _p(k))K _int (2-1)

In the formula, u _Int' get SP for the inner looping setting value _Int' time, the output valve of inner looping controller;

2. the control structure model (1-2) of inner looping control law (2-1) with inner looping PID controller subtracted each other, obtain

u′ _int-u _int＝(SP′ _int-y _p(k))K _int-(SP _int-y _p(k))K _int (2-2)

Namely

u′ _int-u _int＝(SP′ _int-SP _int)K _int (2-3)

3. work as K _Int＞0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＞0, namely as SP ' _Int＞SP _IntThe time, u ' is arranged _Int(k)＞u _Int(k);

Work as K _Int＜0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＜0, namely as SP ' _Int＞SP _IntThe time, u ' is arranged _Int(k)＜u _Int(k);

Therefore, for the service water tank liquor bit string level Predictive Control System of a ratio P control, controller output and setting value strictly monotone;

4. for PFC-PID tandem conduction control loop, the output of external loop controller is the setting value of inner looping controller, namely

u _ext(k)＝SP _int (2-4)

By the strictly monotone of inner looping controller output with the output of external loop controller, then work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint that is operation constraint and the inner looping PID controller of external loop pfc controller is consistent.

Further be set to, in step 3) in, the operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and obtain the constraint condition set of external loop pfc controller when being optimized calculating, adopt following methods: the Optimized model of setting up the external loop pfc controller

$\underset{{\mathrm{\Δp}}_i,{\mathrm{\Δm}}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\Δp}}_i-{\mathrm{\Δm}}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (3-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

u _extmin≤u _ext≤u _extmax

SP _extmin≤SP _ext≤SP _extmax

In the formula,

Be the optimization aim function, n is the variable number;

In optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, can obtain the control law of external loop pfc controller;

u _Extmin≤ u _Ext≤ u _ExtmaxAnd SP _Extmin≤ SP _Ext≤ SP _ExtmaxRepresent respectively u _ExtAnd SP _ExtConstraint condition, u wherein _Extmin, u _ExtmaxRepresent respectively u _ExtMinimum value and maximal value, SP _Extmin, SP _ExtmaxRepresent respectively SP _ExtMinimum value and maximal value;

Because the operation constraint of external loop pfc controller is consistent with the operation of inner looping PID controller constraint, so work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, namely with Ξ (u _Min)≤u _Ext≤ Ξ (u _Max) be constraint condition, (3-1) is optimized calculating to formula, then

$\left\{\begin{array}{c}{u}_{\mathrm{ext}\min }\≤u_{\mathrm{ext}}\≤u_{\mathrm{ext}\max }\\ \mathrm{\Ξ}\left(u_{\min }\right)\≤u_{\mathrm{ext}}\≤\mathrm{\Ξ}\left(u_{\max }\right)\end{array}\right.---(3-2)$

Obtain thus new constraint condition max{u _Extmin, Ξ (u _Min)≤u _Ext≤ min{u _Extmax, Ξ (u _Max).

Further be set to, in step 4) in, the external loop pfc controller is optimized calculating under new constraint condition set, obtain the setting value of inner looping PID controller, and this setting value offered inner looping PID controller, and make inner looping follow the tracks of this setting value, its method is as follows:

Under new constraint condition, the external loop pfc controller is optimized calculating, its Optimized model is

$\underset{{\mathrm{\Δp}}_i,{\mathrm{\Δm}}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\Δp}}_i-{\mathrm{\Δm}}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (4-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

max{u _extmin，Ξ(u _min)}≤u _ext≤min{u _extmax，Ξ(u _max)}

SP _extmin≤SP _ext≤SP _extmax

Find the solution (4-1) and can get the optimization aim function Obtain minimum value J _MinThe time u _ExtBe the optimum solution u of this optimization problem _Extopt, with its setting value SP as the PID controller _IntOffer inner looping.

Further be set to again, in step 1) in, when the external loop controller is the PFC prediction function controller, at first adopt PFC algorithm computing controller structural model;

1. setting up the single order discrete system, is T when its sampling period _s, process gain is K _p, then the output of process is

CV(k)＝y _p(k)＝y _p(k-1)a _p+(1-a _p)K _pu(k-1) (1-3)

In the formula, subscript p represents process;

CV (k) is k control variable constantly;

y _p(k) be the constantly output of service water tank liquor bit string level Predictive Control System of k;

Be the process attenuation coefficient;

T _pBe the process time constant;

U (k-1) is the constantly output of PFC prediction function controller of k-1;

Setting up the model output function is

y _m(k)＝y _m(k-1)a _m+(1-a _m)K _mu(k-1) (1-4)

In the formula, the subscript m representation model;

y _m(k) be the constantly output of model of k;

Be the model attenuation coefficient;

T _mBe the model time constant;

K _mBe model gain;

2. in optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, adopt one-step optimization at this, even Δ p=Δ m

Wherein,

Δp＝ε(k)-ε(k+h)

＝ε(k)-ε(k)λ ^h

＝ε(k)(1-λ ^h)

＝(SP-y _p(k))(1-λ ^h) (1-5)

In the formula, SP is setting value;

ε (k)=SP-y _p(k) expression setting value SP and k moment service water tank liquor bit string level Predictive Control System output y _p(k) deviation;

ε (k+h)=SP-y _p(k+h)=ε (k) λ ^hExpression setting value SP and k+h be service water tank liquor bit string level Predictive Control System output y constantly _p(k+h) deviation;

Be carved into k+1 attenuation coefficient constantly when representing deviation ε from k;

λ ^hBe h the power of λ, be carved into k+h attenuation coefficient constantly when representing deviation ε from k;

For this single order discrete system, get the system time constant of service water tank liquor bit string level Predictive Control System

CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

And

Δm＝y _m(k+h)-y _m(k) (1-6)

In the formula, y _m(k+h) be constantly model prediction of k+h output, by free response y _MFree(k+h) and forced response y _MForce(k+h) form, namely

y _m(k+h)＝y _mFree(k+h)+y _mForce(k+h) (1-7)

Wherein,

$y_{\mathrm{mFree}}(h+k)=y_m\left(k\right)a_m^h---(1-8)$

$y_{\mathrm{mForce}}(h+k)=u\left(k\right)K_m(1-a_m^h)---(1-9)$

In the formula, y _MFree(k+h) expression free response, y _MForce(k+h) expression forced response;

So,

$\mathrm{\Δm}=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-10)$

M gets by Δ p=Δ

$(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-11)$

Obtain the general type of pfc controller after the arrangement

$u\left(k\right)=\frac{(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)-y_m\left(k\right)a_m^h+y_m\left(k\right)}{K_m(1-a_m^h)}---(1-12)$

3. can be got the control structure model of external loop pfc controller by formula (1-12)

$u_{\mathrm{ext}}\left(k\right)=\frac{({\mathrm{SP}}_{\mathrm{ext}}-y_p\left(k\right))(1-{{\mathrm{\λ}}_{\mathrm{ext}}}^h)-y_m\left(k\right){a_{\mathrm{mext}}}^h+y_m\left(k\right)}{K_{\mathrm{mext}}(1-{a_{\mathrm{mext}}}^h)}---(1-1)$

In the formula, u _Ext(k) be the constantly output of external loop PFC prediction function controller of k;

SP _ExtSetting value for external loop PFC prediction function controller;

y _p(k) and y _p(k+h) represent respectively the k moment and the constantly output of service water tank liquor bit string level Predictive Control System of k+h;

Expression k is deviation ε constantly _Ext(k)=SP _Ext-y _p(k) to k+1 moment deviation ε _Ext(k+1)=SP _Ext-y _p(k+1) attenuation coefficient, CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

Be external loop PFC Predictive function control model attenuation coefficient;

T _mBe the model time constant;

K _MextBe model gain;

4. make inner looping adoption rate P control, establishing controller gain is K _Int, obtain thus the control structure model of inner looping PID controller

u _int＝(SP _int-y _p(k))K _int (1-2)

In the formula, inner looping adoption rate P control;

u _IntOutput for the inner looping controller;

SP _IntSetting value for inner looping;

y _p(k) be the constantly output of service water tank liquor bit string level Predictive Control System of k;

K _IntBe the inner looping controller gain.

The invention has the beneficial effects as follows, can be by the saturated control of bottom loop control unit to occurring in the PFC-PID tandem control loop, can avoid levels constraint inconsistence problems to occur, calculate by the constraint rollback, so that the setting value that dynamic optimization is tried to achieve is practical for the bottom control loop, and then improve the control performance of service water tank liquor bit string level Predictive Control System.Particularly in service water tank liquor bit string level Predictive Control System, when being in state of saturation, actuator can calculate correct optimal setting for external loop provides correct inner looping model to guarantee the external loop controller, prevent because phenomenons such as the fluctuation that erroneous calculations causes, overshoot, improve the control performance of service water tank liquor bit string level Predictive Control System, improve the economic benefit that productive unit brings in the service water tank liquor bit string level Predictive Control System.

Description of drawings

Fig. 1 is the process flow diagram of PFC-PID tandem conduction control of the present invention.

Fig. 2 is the structural representation of service water tank liquor bit string level Predictive function control of the present invention system.

Fig. 3 is the block diagram of tandem conduction control loop configuration of the present invention.

Fig. 4 is the synoptic diagram of Predictive function control of the present invention.

Embodiment

The invention will be further described below in conjunction with accompanying drawing:

The CS4000 process control system design industrial water tank that adopts for domestic certain company (can be regarded as is a kind of of tank, namely with the tank of case lid) liquid level cascade control system:

As shown in Figure 1, 2, this industrial water tank tank level control system PFC-PID conduction control loop structure comprises water tank control module, DCS control rack 1 and host computer configuration and monitoring environment.

Among Fig. 1, SP _OptBe the optimal setting of PFC prediction function controller, SP _ExtBe the setting value of external loop PFC prediction function controller, SP _IntMBe the setting value of inner looping PID controller model, MV _ExtBe the performance variable of external loop PFC prediction function controller, MV _IntBe the performance variable of inner looping PID controller, dist is system disturbance, MV _IntMBe the performance variable of inner looping PID controller model, Gp is control object, and CV is system's control variable, and P (ID) is PID controller adoption rate P control.MV is system's input value, refers in the present embodiment valve opening value; CV is system's output valve, refers in the present embodiment high water tank.

Among Fig. 2, DCS control rack 1, water tank 2, solenoid valve 3, reserve tank 4.

As shown in Figure 3, by this PFC-PID conduction control loop of modularization programming environmental structure, wherein main modular comprises external loop pfc controller module and inner looping PID controller module.Inner looping adoption rate P controls the processing procedure dynamic perfromance, and external loop adopts without the PFC control strategy of explicit integration device and follows the tracks of setting value.

1) sets up the control structure model of the external loop pfc controller that is used for industrial water tank liquid level tandem Predictive Control System and the control structure model of inner looping PID controller;

When the external loop controller is the PFC prediction function controller, at first adopt PFC algorithm computing controller structural model.

Setting up the single order discrete system, is T when its sampling period _s, process gain is K _p, then the output of process is

CV(k)＝y _p(k)＝y _p(k-1)a _p+(1-a _p)K _pu(k-1) (1-3)

In the formula, subscript p represents process (process);

CV is the control variable of system;

The control variable of etching system when CV (k) is k;

y _p(k) be the constantly measured value of high water tank of k;

Be the high water tank attenuation coefficient;

T _pBe the process time constant;

U (k-1) is constantly valve opening value of k-1;

Formula (1-3) has been reacted the relation between the control variable CV of system (high water tank) and the performance variable MV (valve opening, the i.e. aperture of solenoid valve 3 valves).The output of process y _p(k) be the constantly measured value of high water tank of k, can in tank liquid level system, obtain by direct-detection.

Setting up the model output function is

y _m(k)＝y _m(k-1)a _m+(1-a _m)K _mu(k-1) (1-4)

In the formula, subscript m representation model (model);

Be the model attenuation coefficient;

T _mBe the model time constant;

K _mBe model gain;

The loop in series setting value, namely the input value of external loop pfc controller is determined by the upper strata steady-state optimization, the water box liquid place value of expression actual needs.In the parameter identification process, it doesn't matter to change setting value and steady-state optimization, and the setting value in this time is that how output changes when manually arbitrarily being to observe input and changing to, purpose.Therefore, change the loop in series setting value, analyze the initial output of pfc controller and stable state output, successively PFC closed loop inner looping model is carried out parameter identification, can be in the hope of K _m=1.9, T _m=29

As shown in Figure 4, in optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, even Δ p=Δ m.

Wherein,

Δp＝ε(k)-ε(k+h)

＝ε(k)-ε(k)λ ^h

＝ε(k)(1-λ ^h)

＝(SP-y _p(k))(1-λ ^h) (1-5)

In the formula, SP is the setting value that host computer offers pfc controller;

ε (k)=SP-y _p(k) expression setting value SP and k moment high water tank y _p(k) deviation;

ε (k+h)=SP-y _p(k+h)=ε (k) λ ^hExpression setting value SP and k+h be high water tank y constantly _p(k+h) deviation;

Be carved into k+1 attenuation coefficient constantly when representing deviation ε from k;

λ ^hBe h the power of λ, be carved into k+h attenuation coefficient constantly when representing deviation ε from k.

For this single order discrete system, get system time constant

CLTR is system's closed loop response time, namely arrives the time of 95% setting value;

And

Δm＝y _m(k+h)-y _m(k) (1-6)

In the formula, y _m(k+h) be constantly model prediction of k+h output, by free response y _MFree(k+h) and forced response y _MForce(k+h) form, namely

y _m(k+h)＝y _mFree(k+h)+y _mForce(k+h) (1-7)

Wherein,

$y_{\mathrm{mFree}}(h+k)=y_m\left(k\right)a_m^h---(1-8)$

$y_{\mathrm{mForce}}(h+k)=u\left(k\right)K_m(1-a_m^h)---(1-9)$

In the formula, y _MFree(k+h) expression free response, y _MForce(k+h) expression forced response;

So,

$\mathrm{\Δm}=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-10)$

M gets by Δ p=Δ

$(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-11)$

Obtain after the arrangement

$u\left(k\right)=\frac{(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)-y_m\left(k\right)a_m^h+y_m\left(k\right)}{K_m(1-a_m^h)}---(1-12)$

So get the general type of pfc controller;

(1-12) can get by formula, the control structure model of external loop pfc controller

$u_{\mathrm{ext}}\left(k\right)=\frac{({\mathrm{SP}}_{\mathrm{ext}}-y_p\left(k\right))(1-{{\mathrm{\λ}}_{\mathrm{ext}}}^h)-y_m\left(k\right){a_{\mathrm{mext}}}^h+y_m\left(k\right)}{K_{\mathrm{mext}}(1-{a_{\mathrm{mext}}}^h)}---(1-1)$

In the formula, u _Ext(k) be the constantly output of external loop PFC prediction function controller of k;

SP _ExtSetting value for external loop PFC prediction function controller;

y _p(k) and y _pThe output of etching system when (k+h) representing respectively k constantly with k+h;

Expression k is deviation ε constantly _Ext(k)=SP _Ext-y _p(k) to k+1 moment deviation ε _Ext(k+1)=SP _Ext-y _p(k+1) attenuation coefficient, CLTR is system's closed loop response time, namely arrives the time of 95% setting value;

Be external loop PFC Predictive function control model attenuation coefficient;

T _mBe the model time constant;

K _MextBe model gain;

Inner looping adoption rate P control, establishing controller gain is K _Int, obtain thus the control structure model of inner looping PID controller

u _int＝(SP _int-y _p(k))K _int (1-2)

In the formula, inner looping adoption rate P control, namely valve opening and high water tank are linear;

u _IntBe valve opening value;

SP _IntSetting value for high water tank;

y _p(k) be the constantly measured value of high water tank of k;

K _IntBe inner looping PID controller gain.

2) judge whether the external loop pfc controller retrains consistent with inner looping PID controller:

Inner looping adoption rate P when control, for the inner looping proportional control rule of control structure model (1-2) expression of inner looping PID controller, the setting value SP ' of high water tank is set artificially _Int, and make this setting value SP ' _Int＞SP _Int, then corresponding inner looping control law is

u′ _int＝(SP′ _int-y _p(k))K _int (2-1)

In the formula, u _Int' get SP for the high water tank setting value _Int' time water tank 2 valve opening value;

The control structure model (1-2) of inner looping control law (2-1) with inner looping PID controller subtracted each other, obtain

u′ _int-u _int＝(SP′ _int-y _p(k))K _int-(SP _int-y _p(k))K _int (2-2)

Namely

u′ _int-u _int＝(SP′ _int-SP _int)K _int (2-3)

Work as K _Int＞0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＞0, namely as SP ' _Int＞SP _IntThe time, u ' is arranged _Int(k)＞u _Int(k);

Work as K _Int＜0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＜0, namely as SP ' _Int＞SP _IntThe time, u ' is arranged _Int(k)＜u _Int(k);

Therefore, for the system of a ratio P control, controller output and setting value strictly monotone;

For this industrial water tank liquid level tandem Predictive Control System, the inner looping controller is output as water tank 2 valve opening value, and setting value is the setting value of high water tank.

For PFC-PID tandem conduction control loop, the output of external loop controller is the setting value of inner looping controller, namely

u _ext(k)＝SP _int (2-4)

By the strictly monotone of inner looping controller output with the output of external loop controller, then work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint that is operation constraint and the inner looping PID controller of external loop pfc controller is consistent.

3) in the situation of not considering the inner looping constraint, calculate external loop PFC control law (the control structure model of external loop pfc controller), will find the solution the controller output that obtains, i.e. optimum water box liquid place value, as the setting value of pid loop, DCS control rack 1 is carried out control operation.

The operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and obtain the constraint condition set of external loop pfc controller when being optimized calculating, adopt following methods:

Retrain in the consistent situation at external loop pfc controller and inner looping PID controller, adopt the constraint backoff algorithm, the constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and obtain the constraint condition set of external loop pfc controller when being optimized calculating.

Set up the Optimized model of external loop pfc controller

$\underset{{\mathrm{\Δp}}_i,{\mathrm{\Δm}}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\Δp}}_i-{\mathrm{\Δm}}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (3-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

u _extmin≤u _ext≤u _extmax

SP _extmin≤SP _ext≤SP _extmax

In the formula, Be the optimization aim function, n is the variable number.In optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, can obtain the control law of external loop pfc controller.

So, u _Extmin≤ u _Ext≤ u _ExtmaxAnd SP _Extmin≤ SP _Ext≤ SP _ExtmaxRepresent respectively u _ExtAnd SP _ExtConstraint condition, u wherein _Extmin, u _ExtmaxRepresent respectively u _ExtMinimum value and maximal value, SP _Extmin, SP _ExtmaxRepresent respectively SP _ExtMinimum value and maximal value;

Because the operation constraint of external loop pfc controller is consistent with the operation of inner looping PID controller constraint, so work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, namely with Ξ (u _Min)≤u _Ext≤ Ξ (u _Max) be constraint condition, (3-1) is optimized calculating to formula, then

$\left\{\begin{array}{c}{u}_{\mathrm{ext}\min }\≤u_{\mathrm{ext}}\≤u_{\mathrm{ext}\max }\\ \mathrm{\Ξ}\left(u_{\min }\right)\≤u_{\mathrm{ext}}\≤\mathrm{\Ξ}\left(u_{\max }\right)\end{array}\right.---(3-2)$

Obtain thus new constraint condition max{u _Extmin, Ξ (u _Min)≤u _Ext≤ min{u _Extmax, Ξ (u _Max).

4) the external loop pfc controller is optimized calculating under new constraint condition set, obtains the setting value of inner looping PID controller, and this setting value is offered inner looping PID controller, makes inner looping follow the tracks of this setting value.

Under new constraint condition, the external loop pfc controller is optimized calculating, its Optimized model is

$\underset{{\mathrm{\Δp}}_i,{\mathrm{\Δm}}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{({\mathrm{\Δp}}_i-{\mathrm{\Δm}}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (4-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

max{u _extmin，Ξ(u _min)}≤u _ext≤min{u _extmax，Ξ(u _max)}

SP _extmin≤SP _ext≤SP _extmax

Find the solution (4-1) and can get the optimization aim function

Obtain minimum value J _MinThe time u _ExtBe the optimum solution u of this optimization problem _Extopt, with its setting value SP as the PID controller _IntOffer inner looping.

Under new constraint condition, calculate external loop PFC control law, constraint consistance rollback calculates so that in the situation of PID controller saturation constraint information is passed to pfc controller, guarantees that the setting value that its Optimization Solution obtains is correctly feasible, thereby improves control performance.

Claims (6)

1. a consistency constraint method that is used for service water tank liquor bit string level Predictive Control System is characterized in that, may further comprise the steps:

1) sets up the control structure model of the external loop pfc controller that is used for service water tank liquor bit string level Predictive Control System and the control structure model of inner looping PID controller;

2) judge whether the external loop pfc controller retrains consistent with inner looping PID controller, if unanimously carry out step 3), if inconsistent, then return step 1) rebulid the control structure model of external loop pfc controller and the control structure model of inner looping PID controller;

3) retrain in the consistent situation at external loop pfc controller and inner looping PID controller, adopt the constraint backoff algorithm, the constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and acquisition external loop pfc controller is being optimized the stylish constraint condition set of calculating;

4) the external loop pfc controller is optimized calculating under new constraint condition set, obtains the setting value of inner looping PID controller, and this setting value is offered inner looping PID controller, makes inner looping follow the tracks of this setting value.

2. a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System according to claim 1, it is characterized in that: in step 1) in, set up the control structure model of the external loop pfc controller that is used for service water tank liquor bit string level Predictive Control System and the control structure model of inner looping PID controller

The control structure model of described external loop pfc controller is

$u_{\mathrm{ext}}\left(k\right)=\frac{({\mathrm{SP}}_{\mathrm{ext}}-y_p\left(k\right))(1-{{\mathrm{\λ}}_{\mathrm{ext}}}^h)-y_m\left(k\right){a_{\mathrm{mext}}}^h+y_m\left(k\right)}{K_{\mathrm{mext}}(1-{a_{\mathrm{mext}}}^h)}---(1-1)$

In the formula, subscript p represents process;

u _Ext(k) be the constantly output of external loop PFC prediction function controller of k;

SP _ExtSetting value for external loop PFC prediction function controller;

y _p(k) and y _p(k+h) represent respectively the k moment and the constantly output of service water tank liquor bit string level Predictive Control System of k+h;

Expression k is deviation ε constantly _Ext(k)=SP _Ext-y _p(k) to k+1 moment deviation ε _Ext(k+1)=SP _Ext-y _p(k+1) attenuation coefficient;

T _sBe the sampling period;

CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

Be external loop PFC Predictive function control model attenuation coefficient;

T _mBe the model time constant;

K _MextBe model gain;

Make inner looping adoption rate P control, establishing controller gain is K _Int, then the control structure model of described inner looping PID controller is

u _int＝(SP _int-y _p(k))K _int (1-2)

In the formula, u _IntOutput for the inner looping controller;

SP _IntSetting value for inner looping;

y _p ^(k)Be the constantly output of service water tank liquor bit string level Predictive Control System of k;

K _IntBe the inner looping controller gain.

3. a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System according to claim 2 is characterized in that: in step 2) in, it is consistent to judge whether external loop pfc controller and inner looping PID controller retrain, and adopts following methods,

1. inner looping adoption rate P when control, for the inner looping proportional control rule of control structure model (1-2) expression of inner looping PID controller, the setting value SP ' of an inner looping is set artificially _Int, and make this setting value SP ' _Int＞SP _Int, then corresponding inner looping control law is

u′ _int＝(SP′ _int-y _p(k))K _int (2-1)

In the formula, u _Int' get SP for the inner looping setting value _Int' time, the output valve of inner looping controller;

2. the control structure model (1-2) of inner looping control law (2-1) with inner looping PID controller subtracted each other, obtain

u′ _int-u _int＝(SP′ _int-y _p(k))K _int-(SP _int-y _p(k))K _int (2-2)

Namely

u′ _int-u _int＝(SP′ _int-SP _int)K _int (2-3)

3. work as K _Int＞0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＞0, namely as SP ' _Int＞SP _IntThe time, u ' is arranged _Int(k)＞u _Int(k);

Work as K _Int＜0 o'clock, owing to supposed SP ' _Int＞SP _IntSo, u ' _Int(k)-u _Int(k)＜0, namely as SP ' _Int＞Sp _IntThe time, u ' is arranged _Int(k)＜u _Int(k);

Therefore, for the service water tank liquor bit string level Predictive Control System of a ratio P control, controller output and setting value strictly monotone;

4. for PFC-PID tandem conduction control loop, the output of external loop controller is the setting value of inner looping controller, namely

u _ext(k)＝SP _int(2-4)

By the strictly monotone of inner looping controller output with the output of external loop controller, then work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint that is operation constraint and the inner looping PID controller of external loop pfc controller is consistent.

4. a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System according to claim 3, it is characterized in that: in step 3) in, the operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, and the constraint condition set of acquisition external loop pfc controller when being optimized calculating, adopt following methods

Set up the Optimized model of external loop pfc controller

$\underset{\mathrm{\Δ}{p}_i,\mathrm{\Δ}{m}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\mathrm{\Δ}{p}_i-\mathrm{\Δ}{m}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (3-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

u _{ext min}≤u _ext≤u _{ext max}

SP _{ext min}≤SP _ext≤SP _{ext max}

In the formula, Be the optimization aim function, n is the variable number;

In optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, can obtain the control law of external loop pfc controller;

u _{Ext Min}≤ u _Ext≤ u _{Ext Max}And SP _Extmin≤ SP _Ext≤ SP _ExtmaxRepresent respectively u _ExtAnd SP _ExtConstraint condition, u wherein _{Ext Min}, u _{Ext Max}Represent respectively u _ExtMinimum value and maximal value, SP _{Ext Min}, SP _{Ext Max}Represent respectively SP _ExtMinimum value and maximal value;

Because the operation constraint of external loop pfc controller is consistent with the operation of inner looping PID controller constraint, so work as u _Min≤ u _Int≤ u _MaxThe time, Ξ (u _Min)≤u _Ext≤ Ξ (u _Max);

The operation constraint of inner looping PID controller is pre-loaded in the computation process of external loop PFC control law, namely with Ξ (u _Min)≤u _Ext≤ Ξ (u _Max) be constraint condition, (3-1) is optimized calculating to formula, then

$\left\{\begin{array}{c}{u}_{\mathrm{ext}\min }\≤u_{\mathrm{ext}}\≤u_{\mathrm{ext}\max }\\ \mathrm{\Ξ}\left(u_{\min }\right)\≤u_{\mathrm{ext}}\≤\mathrm{\Ξ}\left(u_{\max }\right)\end{array}\right.---(3-2)$

Obtain thus new constraint condition max{u _{Ext Min}, Ξ (u _Min)≤u _Ext≤ min{u _{Ext Max}, Ξ (u _Max).

5. a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System according to claim 4, it is characterized in that: in step 4) in, the external loop pfc controller is optimized calculating under new constraint condition set, obtain the setting value of inner looping PID controller, and this setting value offered inner looping PID controller, make inner looping follow the tracks of this setting value, its method is as follows

Under new constraint condition, the external loop pfc controller is optimized calculating, its Optimized model is

$\underset{\mathrm{\Δ}{p}_i,\mathrm{\Δ}{m}_i}{\min }J=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{(\mathrm{\Δ}{p}_i-\mathrm{\Δ}{m}_i)}^2$

s.t.Δp＝[Δp ₁，…，Δp _n] ^T

Δm＝[Δm ₁，…，Δm _n] ^T (4-1)

Δp＝(SP _ext-y _p(k))(1-λ _ext ^h)

Δm＝y _m(k)a _mext ^h+u(k)K _mext(1-a _mext ^h)-y _m(k)

max{u _{ext min}，Ξ(u _min)}≤u _ext≤min{u _{ext max}，Ξ(u _max)}

SP _extmin≤SP _ext≤SP _extmax

Find the solution (4-1) and can get the optimization aim function Obtain minimum value J _MinThe time u _ExtBe the optimum solution u of this optimization problem _Extopt, with its setting value SP as the PID controller _IntOffer inner looping.

6. a kind of consistency constraint method for service water tank liquor bit string level Predictive Control System according to claim 2, it is characterized in that: in step 1) in, when the external loop controller is the PFC prediction function controller, at first adopt PFC algorithm computing controller structural model

1. setting up the single order discrete system, is T when its sampling period _s, process gain is K _p, then the output of process is

CV(k)＝y _p(k)＝y _p(k-1)a _p+(1-a _p)K _pu(k-1)(1-3)

In the formula, subscript p represents process;

CV (k) is k control variable constantly;

y _p(k) be the constantly output of service water tank liquor bit string level Predictive Control System of k;

Be the process attenuation coefficient;

T _pBe the process time constant;

U (k-1) is the constantly output of PFC prediction function controller of k-1;

Setting up the model output function is

y _m(k)＝y _m(k-1)a _m+(1-a _m)K _mu(k-1)(1-4)

In the formula, the subscript m representation model;

y _m(k) be the constantly output of model of k;

Be the model attenuation coefficient;

T _mBe the model time constant;

K _mBe model gain;

2. in optimizing time domain, so that the deviation of the output of process increment Delta p and model prediction output increment Δ m is minimum, adopt one-step optimization at this, even Δ p=Δ m,

Wherein,

Δp＝ε(k)-ε(k+h)

＝ε(k)-ε(k)λ ^h

＝ε(k)(1-λ ^h)

＝(SP-y _p(k))(1-λ ^h)(1-5)

In the formula, SP is setting value;

ε (k)=SP-y _p(k) expression setting value SP and k moment service water tank liquor bit string level Predictive Control System output y _p(k) deviation;

ε (k+h)=SP-y _p(k+h)=ε (k) λ ^hExpression setting value SP and k+h be service water tank liquor bit string level Predictive Control System output y constantly _p(k+h) deviation;

Be carved into k+1 attenuation coefficient constantly when representing deviation ε from k;

λ ^hBe h the power of λ, be carved into k+h attenuation coefficient constantly when representing deviation ε from k;

For this single order discrete system, get the system time constant of service water tank liquor bit string level Predictive Control System

CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

And

Δm＝y _m(k+h)-y _m(k)(1-6)

In the formula, y _m(k+h) be constantly model prediction of k+h output, by free response y _MFree(k+h) and forced response y _MForce(k+h) form, namely

y _m(k+h)＝y _mFree(k+h)+y _mForce(k+h)(1-7)

Wherein,

$y_{\mathrm{mFree}}(k+h)=y_m\left(k\right)a_m^h---(1-8)$

$y_{\mathrm{mForce}}(k+h)=u\left(k\right)K_m(1-a_m^h)---(1-9)$

In the formula, y _MFree(k+h) expression free response, y _MForce(k+h) expression forced response;

So,

$\mathrm{\Δm}=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-10)$

M gets by Δ p=Δ

$(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)=y_m\left(k\right)a_m^h+u\left(k\right)K_m(1-a_m^h)-y_m\left(k\right)---(1-11)$

Obtain the general type of pfc controller after the arrangement

$u\left(k\right)=\frac{(\mathrm{SP}-y_p\left(k\right))(1-{\mathrm{\λ}}^h)-y_m\left(k\right)a_m^h+y_m\left(k\right)}{K_m(1-a_m^h)}---(1-12)$

3. can be got the control structure model of external loop pfc controller by formula (1-12)

$u_{\mathrm{ext}}\left(k\right)=\frac{({\mathrm{SP}}_{\mathrm{ext}}-y_p\left(k\right))(1-{{\mathrm{\λ}}_{\mathrm{ext}}}^h)-y_m\left(k\right){a_{\mathrm{mext}}}^h+y_m\left(k\right)}{K_{\mathrm{mext}}(1-{a_{\mathrm{mext}}}^h)}---(1-1)$

In the formula, u _Ext(k) be the constantly output of external loop PFC prediction function controller of k;

SP _ExtSetting value for external loop PFC prediction function controller;

y _p(k) and y _p(k+h) represent respectively the k moment and the constantly output of service water tank liquor bit string level Predictive Control System of k+h;

Expression k is deviation ε constantly _Ext(k)=SP _Ext-y _p(k) to k+1 moment deviation ε _Ext(k+1)=SP _Ext-y _p(k+1) attenuation coefficient,

CLTR is the single order discrete system closed loop response time, namely arrives the time of 95% setting value;

Be external loop PFC Predictive function control model attenuation coefficient;

T _mBe the model time constant;

K _MextBe model gain;

4. make inner looping adoption rate P control, establishing controller gain is K _Int, obtain thus the control structure model of inner looping PID controller

u _int＝(SP _int-y _p(k))K _int (1-2)

In the formula, inner looping adoption rate P control;

u _IntOutput for the inner looping controller;

SP _IntSetting value for inner looping;

y _p ^(k)Be the constantly output of service water tank liquor bit string level Predictive Control System of k;

K _IntBe the inner looping controller gain.

Images