CN104881062A - Quick and overregulation-free cooling crystallization reaction kettle temperature control method - Google Patents

Abstract

Provided is a quick and overregulation-free cooling crystallization reaction kettle temperature control method. According to the method, a refrigeration compressor with a variable-frequency power regulation function, an electronic heating tube based on pulse width modulation power, a programmable controller and a monitoring computer are used to construct a temperature control system. First, a square wave excitation experiment is adopted, namely, the temperature change of a crystallization reaction kettle is detected by periodically changing the power of the refrigeration compressor, and a cooling response transfer function model of the crystallization reaction kettle is established according to experimental data; then, a robust closed-loop control system and the form of the controller are designed based on the response model; and finally, the suitable range of the parameters of the controller is set according to the actual executable power of the refrigeration compressor and the electronic heating tube and the working condition constraints of the reaction kettle. By adopting the method of the invention, the cooling rate of solution in the reaction kettle can be regulated quantitatively to ensure no overregulation and reach a specified target value of cooling. Thus, a convenient and reliable method of automatic control is provided for cooling regulation of the crystallization process.

Description

A kind of fast without toning crystallisation by cooling temperature of reaction kettle control method

Technical field

The invention belongs to industrial stokehold technical field, relate to the temperature control system of industrial crystallization reactor, specifically refer to a kind of fast without toning crystallisation by cooling temperature of reaction kettle control method.

Background technology

Chemical reaction still for crystallization and biofermentation etc. generally adopts temperature to control to regulate production run, and temperature controls mainly to comprise heating, cooling, constant temperature.For crystallization process, mainly realize crystalline product by cooling and separate out, because rate of temperature fall and stationarity directly determine crystalline product quality and size, thus cooling control is the core technology of cooling and crystallizing process.Owing to there is reactor volume disunity and the heat transfer characteristic difference of various material and cooled circulated medium is large in Practical Project practice, lack consistent high efficiency cooling control method of generally acknowledging, the relevant temperature control method seldom having document and patent Introduction to promote the use of both at home and abroad, if international crystallization Engineering Control expert Z.K.Nagy is at recent literature " Efficient output feedbacknonlinear model predictive control for temperature control of industrial batch reactors, " (letter is translated: the efficient output feedack nonlinear model predictive control method controlled for industrial mass manufacture temperature of reaction kettle, be published in the international important publication Control Engineering Practice in control engineering field, 2007, 15, 839-859.) in explicitly point out, adopt the problem that conventional unit feedback control structure can cause toning to cool, a kind of nonlinear model predictive control method based on output feedack is proposed for this reason, the phenomenon that toning cools is there is not by regulating rate of temperature fall in real time to guarantee, but it is too conservative that its shortcoming is control performance, also namely rate of temperature fall can not regulate comparatively fast, and controller architecture is too complicated, on-line calculation is large, depend on high-performance computer and perform control algolithm, thus its range of application is limited.

In current crystallization engineering practice, great majority are based on manual operation experience, according to test and historical operation result, repeatedly regulate and optimize cooling control strategy, the major defect of this artificial experience method is: (1) needs repeatedly to adjust for a long time test, does not have unified control method; (2) can not the qualitative assessment cooling control strategy performance index that can reach, once there is system operation conditions or Parameters variation, be difficult to the stability of Guarantee control system; (3) there is no reference design standard, be not easy to the Control System Design of Rapid Popularization for the different production scale of construction or close production system.Therefore, how designing the temperature control system cooled without toning is fast a current investigation and application difficult problem.

Summary of the invention

The technical problem to be solved in the present invention be in industrial cooling and crystallizing process quick without toning cooling control problem.For solving this problem above-mentioned, propose to set up transient temperature response transfer function model to design the technical method of sampled-data control system to crystallization reaction still solution system, to realize fast without toning cooling control effects.

The present invention utilizes refrigeration compressor, the electrons heat pipe based on pulse width modulation power, programmable controller (PLC), the supervisory control comuter structure temperature control system of variable ratio frequency changer regulating power, first square wave excitation is adopted, namely detected the temperature variation of crystallization reaction still by the power of periodically-varied refrigeration compressor, application system identification theory sets up the cooling response pass function model of crystallization reaction still; Secondly the model obtained based on identification adopts robust internal model control Theoretical Design closed-loop control system and controller form; Finally, the reality according to refrigeration compressor and electrons heat pipe can perform power and reactor operating mode constraint condition, controller parameters setting OK range, and determines the adverse effect that steady-state heat balance controling parameters causes to eliminate function of environment heat emission.

Technical scheme of the present invention is as follows:

(1) open loop cooling identification

First the temperature (such as 80 DEG C) utilizing heating arrangement solution temperature in crystallization reaction still to be increased to crystal in solution all to dissolve, and keep this temperature stabilization a period of time (as 10-30 minute), then heating arrangement is turned off, adopt square wave test signal to start and regulate refrigerating plant power to lower the temperature to crystallization reaction still, the data of solution temperature change in Real-time Collection crystallization reaction still, until temperature drops to such as room temperature 25 DEG C beyond Tc lower bound);

2. set up temperature-responsive transfer function model: the temperature-responsive transfer function model setting up crystallization reaction still according to the temperature variation data collected;

3. design closed-loop control system: according to the temperature-responsive transfer function model set up, design closed-loop control system structure and controller form;

4. controller parameters setting: carry out operation to crystallization reaction still application closed-loop control system and run, is debugged by the single adjustable parameter increased monotonously or reduces controller, realizes lowering the temperature control effects without toning the most fast.

The present invention utilizes open loop cooling identification, the transfer function model of crystallization reaction still cooling dynamic response characteristic can be set up, for Control System Design provides reference frame, the output error quadratic sum minimum performance index of robust internal model control theory can be reached based on transfer function model design con-trol system, ensure to realize without toning cooling control effects, controller form is the rational expression based on model parameter, be easy to be written as software program or hardware device making, and there is single adjustable setting parameter, can regulate monotonously easily (as increased monotonously or reducing), overcome the time-varying Hurst index of crystallization reaction still in actual mechanical process, reach optimized control performance, realize the most fast without toning cooling control effects.

The present invention can regulate the rate of temperature fall of crystallization reaction still solution quantitatively, guarantee to reach the cooling desired value of specifying without toning, method is simple, do not rely on any priori of crystallization reaction still and operating experience or database, can realize, fast without toning cooling control effects, being convenient to practical application in industry and popularization.

Accompanying drawing explanation

Fig. 1 is control system frame principle figure of the present invention.In Fig. 1, C _srefer to and (show for controller r) in figure, u for following the tracks of set temperature value _crepresent C _soutput control signal; C _frefer to the closed loop controller for eliminating modeling error and load disturbance, u _frepresent C _foutput control signal; U refers to the command signal of regulating and controlling refrigeration/circulator, and as shown in the figure, it is by u _cand u _fmix, i.e. u=u _c-u _f; T _rbe the cooling response pass function model expected, its output is that the temperature-responsive value expected (is designated as y _r), the measured temperature (y) of actual crystallization reaction still and y _rbetween deviation be used as feedback control signal, send to closed loop controller C _f; Refrigeration/circulator is regarded as Generalized Control object together with crystallization reaction still, so that CONTROLLER DESIGN.

Fig. 2 is open loop of the present invention cooling identification schematic diagram.In Fig. 2, the command signal of refrigerating/heating circulator is set when square-wave signal refers to and implements open loop cooling identification test.

Fig. 3 is 10 liters of crystallization reaction still temperature-responsive identification effect figure that application open loop cool-down method of the present invention draws.Wherein heavy line represents the temperature response curve of 10 liters of crystallization reaction still solution, and dotted line represents the identification model temperature response curve that the present invention provides.

The cooling control effects curve that Fig. 4 (a) and Fig. 4 (b) provides for the present invention's (heavy line) and the German Julabo company CF41 automatic temperature-controlled refrigeration/circulator of series (thick dotted line).Wherein, Fig. 4 (a) shows temperature response curve, and Fig. 4 (b) shows refrigeration work consumption and the heating power change curve of refrigeration/circulator.

Embodiment

In order to understand technical scheme of the present invention better, below in conjunction with accompanying drawing, embodiments of the present invention are described in detail.

Embodiment adopts 10 liters of crystallization reaction stills, and in-built 4 liters of concentration are the glutamic acid aqueous solution of 10%, are configured with a refrigerating/heating circulator, is made up of the freon cooling compressor of 2kw refrigeration work consumption and the electrons heat pipe of 2kw heating power.Wherein the refrigeration work consumption of freon cooling compressor opens range of operation is 40-100%, and it is 0-100% that the heating power of electrons heat pipe opens range of operation.

Adopt specific embodiment of the invention step as follows:

Step 1: open loop cooling identification

First solution temperature in crystallization reaction still is increased to 50 DEG C by unlocking electronic heating tube, and keeps this temperature stabilization about 20 minutes, and glutamic acid crystal in solution is all dissolved.Then turn off electrons heat pipe, adopt square wave test signal u to start and regulate freon cooled compressed acc power to lower the temperature to crystallization reaction still, even

$u\left(t\right)=\{\begin{array}{cc}{h}_1,& 0\&le;t<T_{p1};\\ h_2,& T_{p1}\&le;t<T_{p1}+T_{p2}.\end{array}---\left(1\right)$

Wherein h ₁=50 (Hz) refer to the frequency converter setting value regulating cooled compressed acc power, and it is 50%, h that correspondence arranges compressor output power ₂=100 (Hz) represents that arranging compressor output power is 100%, T _p1=20 (s) and T _p2=100 (s) represent respectively compressor output power be 60% and 100% time be 20 seconds and 100 seconds.Because the sampling period is T _s=0.2 (s), cooled compressed acc power can not complete the conversion from 50% to 100% output power within short like this sampling period, therefore the actual test 20 second time of employing completes the conversion of these two output powers, within this 20 second transition period, cooled compressed acc power is progressively incremented to 100% by 50%, and vice versa.When detecting that solution temperature is down to 40 DEG C, terminate, as shown in Figure 3.

Step 2: set up temperature-responsive transfer function model

According to the solution temperature delta data collected from identification, be designated as Y=[y (t ₀+ 1), y (t ₀+ 2) ..., y (N)] ^t(wherein t ₀represent the starting sample moment, N represents sampled data length), consider that temperature-fall period belongs to integral process, adopt following Disgrete Time Domain integral model structure to carry out data fitting,

$G_m\left(z\right)=\frac{B\left(z\right)}{(1-z^{-1})A\left(z\right)}{z}^{-d}---\left(2\right)$

Wherein z represents sampling time operator, namely has z ^-1u (t)=u (t-1),

$A\left(z\right)=1+a_1{z}^{-1}+a_2{z}^{-2}+...+a_{n_a}{z}^{-n_a}---\left(3\right)$

$B\left(z\right)=b_1{z}^{-1}+b_2{z}^{-2}+...+b_{n_b}{z}^{-n_b}---\left(4\right)$

Therefore, model parameter to be estimated can be written as a vector form

θ _g＝[θ ^T,d] ^T(5)

Wherein $\mathrm{\&theta;}={\&lsqb;a_1,a_2,...,a_{n_a},b_1,...,b_{n_b}\&rsqb;}^r.$

For ease of parameter estimation, make y _e(t)=(1-z ^-1) y (t)=y (t)-y (t-1), can be obtained by formula (2)

$y_e\left(t\right)=\frac{B\left(z^{-1}\right)}{A\left(z^{-1}\right)}{z}^{-d}u\left(t\right)---\left(6\right)$

According to sampled data, can arrange and obtain observation data sequence and matrix,

Y _e＝[y _e(t ₀+1),y _e(t ₀+2),…,y _e(N)] ^T(7)

Wherein represent the Delay Parameters pre-estimated, its initial value an approximate value can be observed from the initial time delayed response of cooling.

First initial samples data amount check is set, is designated as N ₀, application least square method can be done according to a preliminary estimate to parameter vector θ, namely

$\widehat{\mathrm{\&theta;}}={\left({\mathrm{\&psi;\&psi;}}^T\right)}^{-1}\mathrm{\&psi;}{Y}_e---\left(10\right)$

Then along data sampling order, get a moving window, length of window is designated as L, does iterative algorithm progressively convergence estimate best fit parameters, namely

${\widehat{\mathrm{\&theta;}}}_g\left(t\right)={\widehat{\mathrm{\&theta;}}}_g(t-1)+K_L\left(t\right)\&lsqb;Y_{e-L}\left(t\right)-{\widehat{\mathrm{\&psi;}}}_L\left(t\right)\widehat{\mathrm{\&theta;}}(t-1)\&rsqb;---\left(11\right)$

Wherein

$K_L\left(t\right)=P(t-1){\mathrm{\&Phi;}}_L^{*}\left(t\right){({\mathrm{\&lambda;I}}_{L\&times;L}+{\mathrm{\&Phi;}}_L^T(t\left)P\right(t-1\left){\mathrm{\&Phi;}}_L^{*}\right(t\left)\right)}^{-1}---\left(12\right)$

$P\left(t\right)=\frac{1}{\mathrm{\&lambda;}}(I_{n_m\&times;n_m}-K_L(t){\mathrm{\&Phi;}}_L^T\left(t\right))P(t-1)---\left(13\right)$

${\mathrm{\&phi;}}^{*}\left(t\right)={\&lsqb;1/t^p,...,1/t,1,t,...,t^q,u(t-1),...,u(t-n_b),-\underset{k=1}{\overset{n_b}{\&Sigma;}}{\hat{b}}_k{z}^{-\hat{d}}\mathrm{\&Delta;}u(t-k)\&rsqb;}^T---\left(15\right)$

Δu(t-k)＝u(t-k)-u(t-k-1) (18)

In formula (13), P (t) initial value can be taken as above iteration forgetting factor can be taken as

$\mathrm{\&lambda;}\left(t\right)=max({\mathrm{\&lambda;}}_{\min },1/[1+\left|\right|{\widehat{\mathrm{\&theta;}}}_g\left(t\right)-{\widehat{\mathrm{\&theta;}}}_g(t-1)|{|}_2\&rsqb;)---\left(19\right)$

Wherein λ _mincan choose in [0.85,0.95], get higher value and can improve the susceptibility of iterative algorithm to noise signal.But the speed of convergence of the parameter estimation that can slow down, vice versa.

In formula (15), p and q can be taken as arbitrary integer above, but must meet p+q=n _a-1.

Apply above-mentioned iterative algorithm, until parameter estimation meets precision conditions wherein ε can be taken as a small value as 0.001 etc. according to actual measurement noise level, or till sampled data length, i.e. t=N.

According to the data of above-mentioned 10 liters of crystallization reaction stills, get t ₀=30, N ₀=300, λ _min=0.985, L=150, n=8000 applies this identification algorithm, can obtain cooling response pass function model,

$G_m\left(z\right)=\frac{-0.9368\&times;{10}^{-6}\&times;(z^{-1}-0.9829z^{-2})}{(1-z^{-1})(1-0.9993z^{-1})}{z}^{-566}---\left(20\right)$

Step 3: design closed-loop control system

Build closed-loop control system as shown in Figure 1, wherein setting value tracking controller is following form

Wherein F (z) is with a customized parameter λ _cwave filter, its form is

$F\left(z\right)=\frac{z^l{(1-{\mathrm{\&lambda;}}_c)}^l}{{(z-{\mathrm{\&lambda;}}_c)}^l}---\left(22\right)$

The order l of F (z) can choose according to working control device output violent change condition, and minimum order can be taken as 1.G _mz () is target transfer function model G _mz the minimum phase part (namely zero pole point is all in z-plane unit circle) in (), is namely undertaken drawing by the model shown in formula (2) following decomposition,

$G_m\left(z\right)=G_M\left(z\right)z^{-d}\underset{i=1}{\overset{n_q}{\&Pi;}}(z-z_i)---\left(23\right)$

Wherein z _i(i=1 ... n _q) be G _mthe zero point outside z-plane unit circle is positioned in (z), namely | z _i| > 1.N _gmake the molecule positive integer identical with denominator order.

As can be seen from formula (21), this controller has single customized parameter λ _c, corresponding expectation cooling response pass function is

T _r＝G _m(z)C _s(z) (24)

Convolution (21) and (24) can be seen, increase regulating parameter λ monotonously _csetting point tracking can be made to respond slack-off, thus raising is having the tracking response robustness in the uncertain situation of object, vice versa.

According to 10 liters of crystallization reaction stills cooling response pass function models shown in such as formula (20), can be drawn by controller design method above

$C_s\left(z\right)=\frac{(1-z^{-1})(1-0.9993z^{-1}){(1-{\mathrm{\&lambda;}}_c)}^3}{-0.9368\&times;{10}^{-6}(1-0.9829z^{-1}){(1-{\mathrm{\&lambda;}}_c{z}^{-1})}^3}---\left(25\right)$

$T_r=\frac{{(1-{\mathrm{\&lambda;}}_c)}^3{z}^{-1}}{{(1-{\mathrm{\&lambda;}}_c{z}^{-1})}^3}---\left(26\right)$

Illustratively, choose 3 rank wave filter F (z) (i.e. l=3) here and carry out CONTROLLER DESIGN C _sz () is the peak power output restriction considering cooling compressor.

Closed loop controller C _fadopt following form

$C_f=\frac{T_d}{G_m(1-T_d)}---\left(27\right)$

Wherein

$T_b\left(z\right)=z^{-d}\frac{({\mathrm{\&beta;}}_0+{\mathrm{\&beta;}}_{1}z){(1-{\mathrm{\&lambda;}}_f)}^{n_f}}{{(z-{\mathrm{\&lambda;}}_f)}^{n_f}}\underset{i=1}{\overset{n_q}{\&Pi;}}\frac{(1-z_i^{-1})(z-z_i)}{(1-z_i)(1-z_i^{-1})}---\left(28\right)$

β ₀＝1-β ₁(29)

${\mathrm{\&beta;}}_1=d+\frac{n_f}{1-{\mathrm{\&lambda;}}_f}-\underset{j=1}{\overset{n_q}{\&Sigma;}}\frac{1+z_j}{1-z_j}---\left(30\right)$

In above formula (28), n _fbe be similar to the filter order l in formula (22), can choose according to working control device output violent change condition, minimum order can be taken as 1.λ _fc _fsingle customized parameter, increase λ monotonously _fcan accelerate close-loop feedback control performance, but can make to be deteriorated than closed loop robust stability having in the uncertain situation of object, vice versa.

Consider the peak power output restriction of cooling compressor, get n here _f=4, determine thus

$T_d\left(z\right)=\frac{{(1-{\mathrm{\&lambda;}}_f)}^4({\mathrm{\&beta;}}_{1}z+{\mathrm{\&beta;}}_0)}{{(z-{\mathrm{\&lambda;}}_f)}^4}{z}^{-d}$

${\mathrm{\&beta;}}_1=\frac{4+d(1-{\mathrm{\&lambda;}}_f)}{1-{\mathrm{\&lambda;}}_f},{\mathrm{\&beta;}}_0=1-{\mathrm{\&beta;}}_1$

Step 4: controller parameters setting

From, therefore the maximum rate of temperature fall of crystallization reaction still solution more than 0.1 DEG C/sec, can not adopt following anti-noise filter feedback signal,

$\hat{y}\left({\mathrm{kT}}_s\right)=\left\{\begin{array}{cc}y\left(\right(k-1)T_s)+\mathrm{\&Delta;}T,& y\left({\mathrm{kT}}_s\right)-y\left(\right(k-1)T_s)\&GreaterEqual;\mathrm{\&Delta;}T;\\ y\left(\right(k-1)T_s)+\mathrm{\&Delta;}T,& y\left({\mathrm{kT}}_s\right)-y\left(\right(k-1)T_s)\&le;-\mathrm{\&Delta;}T;\\ y\left({\mathrm{kT}}_s\right),& else.\end{array}\right.$

Wherein refer to the temperature measurement signal for FEEDBACK CONTROL, y (kT _s) be the temperature value of actual measurement, Δ T=0.01 DEG C is the maximum amplitude limit of filtering.

Because the minimum startup output power of cooling compressor is 40% of rated power, therefore in order to overcome its can not execution requirements output power in the defect of 0-40% rated power, here adopting the method for thermal compensation to carry out this low output power traffic coverage of approximate substitution, is namely the method for 40% and supplementary heating power by the output power of constant cooling compressor.According to emulation and test, draw following thermal compensation reduction formula

h _p＝-1.3385V _c+56.0414

Wherein V _crefer to the refrigeration output power needing to perform, h _prefer to when the output power of maintenance cooling compressor is 40% rated power, the heating power needed for equivalence.

Initial setting up two controller parameter value λ _c=λ _f=0.99, the real output in conjunction with cooling compressor and electrons heat pipe limits, and through increasing this two controling parameters monotonously online, determines that one group obtains fast without the parameter tuning value λ of toning control effects _c=0.99936 and λ _f=0.9978, control effects is as shown in (a) He Fig. 4 (b) in Fig. 4.In Fig. 4, (a) is the temperature response curve of crystallization reaction still solution, is the real output of cooling compressor and electrons heat pipe in Fig. 4 (b).The cooling control effects of the robotization temperature control equipment (product CF41 in 2014) of German JULABO company is also show, to make comparisons in figure.Can see, the control method that the present invention provides can shorten about 400 second time and reach steady-state target temperature province (40 ± 0.1) DEG C, because whole temperature fall time is approximately 2400 seconds, therefore visible the present invention can significantly improve cooling and control rapidity, and ensures to respond without toning.

Claims (1)

1., fast without a toning crystallisation by cooling temperature of reaction kettle control method, it is characterized in that following steps:

(1) open loop cooling identification

First the temperature utilizing heating arrangement solution temperature in crystallization reaction still to be increased to crystal in solution all to dissolve, keep this temperature stabilization, then turn off heating arrangement, adopt square wave test signal u to start and regulate refrigerating plant power to lower the temperature to crystallization reaction still, namely $u\left(t\right)=\left\{\begin{array}{cc}{h}_1,& 0\&le;t<T_{p1};\\ h_2,& T_{p1}\&le;t<T_{p1}+T_{p2};\end{array}\right.$ Wherein, h ₁refer to the Lower Limit Amplitude of setting cooled compressed acc power, h ₂represent the upper limit magnitude of setting cooled compressed acc power, T _p1and T _p2represent that compressor output power is the time of lower limit and upper limit magnitude respectively;

The data of solution temperature change in Real-time Collection crystallization reaction still, until solution temperature drops to beyond Tc lower bound, terminate cooling;

(2) temperature-responsive transfer function model is set up

According to the solution temperature delta data that identification collects, be designated as Y=[y (t ₀+ 1), y (t ₀+ 2) ..., y (N)] ^t, wherein t ₀represent the starting sample moment, N represents sampled data length; Following Disgrete Time Domain integral model structure is adopted to carry out data fitting,

$G_m\left(z\right)=\frac{B\left(z\right)}{(1-z^{-1})A\left(z\right)}{z}^{-d}$

Wherein z represents sampling time operator, namely has z ^-1u (t)=u (t-1),

$A\left(z\right)=1+a_1{z}^{-1}+a_2{z}^{-2}+\mathrm{...}+a_{n_a}{z}^{-n_a}$

$B\left(z\right)=b_1{z}^{-1}+b_2{z}^{-2}+\mathrm{...}+b_{n_b}{z}^{-n_b}$

Remember that model parameter to be estimated is vector form a: θ _g=[θ ^t, d] ^t, wherein

Note y _e(t)=(1-z ^-1) y (t)=y (t)-y (t-1), obtain following form

$y_e\left(t\right)=\frac{B\left(z^{-1}\right)}{A\left(z^{-1}\right)}{z}^{-d}u\left(t\right)$

According to sampled data, arrange and obtain observation data sequence and matrix:

Y _e＝[y _e(t ₀+1),y _e(t ₀+2),…,y _e(N)] ^T

Wherein represent the Delay Parameters pre-estimated, its initial value an approximate value is observed from the initial time delayed response of cooling;

First set initial samples data amount check, be designated as N ₀, application least square method is done according to a preliminary estimate to parameter vector θ, namely

$\widehat{\mathrm{\&theta;}}={\left({\mathrm{\&psi;\&psi;}}^T\right)}^{-1}{\mathrm{\&psi;Y}}_e$

Then along data sampling order, get a moving window, length of window is designated as L, does iterative algorithm progressively convergence estimate best fit parameters, that is: ${\widehat{\mathrm{\&theta;}}}_g\left(t\right)={\widehat{\mathrm{\&theta;}}}_g(t-1)+K_L\left(t\right)\&lsqb;Y_{e-L}\left(t\right)-{\widehat{\mathrm{\&psi;}}}_L\left(t\right)\widehat{\mathrm{\&theta;}}(t-1)\&rsqb;$

Wherein, $K_L\left(t\right)=P(t-1){\mathrm{\&Phi;}}_L^{*}\left(t\right){({\mathrm{\&lambda;I}}_{L\&times;L}+{\mathrm{\&Phi;}}_L^T(t\left)P\right(t-1\left){\mathrm{\&Phi;}}_L^{*}\right(t\left)\right)}^{-1}$

$P\left(t\right)=\frac{1}{\mathrm{\&lambda;}}(I_{n_m\&times;n_m}-K_L\left(t\right){\mathrm{\&Phi;}}_L^T\left(t\right))P(t-1)$

${\mathrm{\&phi;}}^{*}\left(t\right)={\&lsqb;1/t^p,...,1/t,1,t,...,t^q,u(t-1),...,u(t-n_b),-\underset{k=1}{\overset{n_b}{\&Sigma;}}{\hat{b}}_k{z}^{-\hat{d}}\mathrm{\&Delta;}u(t-k)\&rsqb;}^T$

Δu(t-k)＝u(t-k)-u(t-k-1)

The initial value of P (t) is taken as above iteration forgetting factor is taken as

$\mathrm{\&lambda;}\left(t\right)=max({\mathrm{\&lambda;}}_{\min },1/\&lsqb;1+|\left|{\widehat{\mathrm{\&theta;}}}_g\right(t)-{\widehat{\mathrm{\&theta;}}}_g(t-1)|{|}_2\&rsqb;)$

Wherein λ _minchoose in [0.85,0.95]; P and q is taken as arbitrary integer, but must meet p+q=n _a-1;

Apply above-mentioned iterative algorithm, until parameter estimation meets precision conditions wherein ε is taken as a small value according to actual measurement noise level or till sampled data length, i.e. t=N;

(3) closed-loop control system is designed

Have two controllers in control system, one of them is setting value tracking controller C _s, its input end is setting value input instruction signal r, exports the positive terminal of a termination two paths of signals mixer, the output signal of another controller of negative pole termination of this signal mixer; Another controller is close-loop feedback controller C _f, its input end is the output of another two paths of signals mixer, the measuring tempeature signal y of the brilliant reactor solution of positive terminal access node of this signal mixer, its consequent pole termination preferred temperature output response signal y _r, this signal is by preferred temperature response pass function T _rproduce, the input termination setting value input instruction signal r of this transport function; Setting value tracking controller is following form

Wherein F (z) is with a customized parameter λ _cwave filter, its form is

$F\left(z\right)=\frac{z^l{(1-{\mathrm{\&lambda;}}_c)}^l}{{(z-{\mathrm{\&lambda;}}_c)}^l}$

The order l of F (z) chooses according to working control device output violent change condition, and minimum order is taken as 1; G _mz () is target transfer function model G _mminimum phase part in (z), the transfer function model namely set up by above-mentioned steps (2) carries out drawing following decomposition,

$G_m\left(z\right)=G_M\left(z\right)z^{-d}\underset{i=1}{\overset{n_q}{\&Pi;}}(z-z_i)$

Wherein z _i(i=1 ... n _q) be G _mthe zero point outside z-plane unit circle is positioned in (z), namely | z _i| > 1; n _gmake the molecule positive integer identical with denominator order;

Setting value tracking controller has single customized parameter λ _c, corresponding expectation cooling response pass function is

T _r＝G _m(z)C _s(z)

Close-loop feedback controller C _ffor following form

$C_f=\frac{T_d}{G_m(1-T_d)}$

Wherein

$T_d\left(z\right)=z^{-d}\frac{({\mathrm{\&beta;}}_0+{\mathrm{\&beta;}}_{1}z){(1-{\mathrm{\&lambda;}}_f)}^{n_f}}{{(z-{\mathrm{\&lambda;}}_f)}^{n_f}}\underset{i=1}{\overset{n_q}{\&Pi;}}\frac{(1-z_i^{-1})(z-z_i)}{(1-z_i)(z-z_i^{-1})}$

β ₀＝1-β ₁

${\mathrm{\&beta;}}_1=d+\frac{n_f}{1-{\mathrm{\&lambda;}}_f}-\underset{j=1}{\overset{n_q}{\&Sigma;}}\frac{1+z_j}{1-z_j}$

Wherein, λ _fc _fsingle customized parameter, n _fbe a filter order chosen according to working control device output violent change condition, minimum order is taken as 1;

(4) controller parameters setting

Increase setting value tracking controller C monotonously _sregulating parameter λ _csetting point tracking is responded slack-off, improve and having the tracking response robustness in the uncertain situation of object, vice versa; Increase close-loop feedback controller C monotonously _fregulating parameter λ _faccelerate close-loop feedback control performance, but can make to be deteriorated than closed loop robust stability having in the uncertain situation of object, vice versa; In conjunction with the output power amplitude limit condition of actual refrigerating plant, by increasing monotonously or reduce this two controling parameters λ online _cand λ _f, reach the most fast without toning cooling control effects.