CN100541360C - Nuclear power device two-loop multi-variable integrated model fuzzy predication control method - Google Patents

Abstract

The present invention is to provide a kind of nuclear power device two-loop multi-variable integrated model fuzzy predication control method.Utilize measuring system to measure the state parameter information of nuclear power unit; Change digital signal into through behind the wave filter, give controller; Controller is selected the control input of optimum nuclear power unit secondary coolant circuit system; The control signal that controller produced is exported to topworks after producing simulating signal and the enhancing of process signal amplifier through digital/analog converter; Topworks carries out by instruction, total system is changed under the operating mode of appointment.The invention has the advantages that to be applicable to that nuclear power unit has the system of severe nonlinear, coupling, time variation like this, the control accuracy height, robustness is good.

Description

Nuclear power device two-loop multi-variable integrated model fuzzy predication control method

(1) technical field

What the present invention relates to is a kind of control method, specifically a kind of nuclear power unit secondary circuit kinetic controlling equation method.

(2) background technology

The nuclear power unit secondary coolant circuit system owing to strong coupling, the non-linear and equal characteristic of dynamic perfromance, makes the design of its control system have certain challenge.Traditional control method exists each major parameter fluctuation of system bigger aspect solution nuclear power unit variable working condition and the coupling problem, the shortcoming of low-response.This multivariate integrated model Fuzzy Predictive Control method is introduced the integrated model that artificial nerve network model is combined with the multivariate linear prediction model, forecast model is carried out dynamic optimization, and introducing fuzzy theory, set up the fuzzy control performance evaluation table of nuclear power unit secondary circuit control system input quantity, select optimum input quantity by fuzzy decision, realized multivariate integrated model Fuzzy Predictive Control.Reduce the amplitude of control system output pulsation, obtained better control performance.Thereby guaranteed that nuclear power unit moves safely and reliably.

Number of patent application is to disclose a kind of " considering the feed water control system and the control method of water supply control valve pressure drop in the nuclear power station " in 200410069913.5 the application for a patent for invention file.In this control system, adopt the PD control algolithm, be divided into two unit.First detecting unit comprises: the flow rub-out signal generator of flow rub-out signal that produces the difference of corresponding steam flow signal and feedwater flow signal; Water level error correction signal generator with the water level error correction signal that produces corresponding level measuring signal and flow rub-out signal sum.Second detecting unit comprises: the pressure reduction that detects the preceding and rear section of at least one water supply control valve in main water supply control valve and the downcomer water supply control valve also produces the pressure drop sensor unit of the pressure drop signal of the corresponding pressure reduction that detects; Produce the pressure drop duty setting signal generator of the pressure drop duty setting signal of the corresponding water supply control valve pressure drop setting value of setting previously; With the pressure drop rub-out signal generator that compares pressure drop signal and pressure drop duty setting signal and generation pressure drop rub-out signal.Control module is according to controlling main feed pump from the correction control signal of correcting control-signals generator output.But this control system is under high load working condition changes, and systematic parameter exists overshoot big, and the deficiency that stabilization time is long is still waiting to improve.

(3) summary of the invention

The objective of the invention is to have serious coupling and non-linear at the nuclear power unit secondary coolant circuit system, provide a kind of variable working condition particularly during big load variations, can obtain the nuclear power device two-loop multi-variable integrated model fuzzy predication control method of good dynamic control performance.

The object of the present invention is achieved like this:

(1) utilize measuring system to measure the state parameter information of nuclear power unit;

(2) convert the state parameter that obtains to digital signal by analog/digital converter, digital signal is given controller through behind the wave filter;

(3) the multivariate integrated model fuzzy prediction algorithm that comprises in the controller is under the effect of reference controlled quentity controlled variable, the output in future of multistep prediction device prediction controlled system, the predicated error and the predicated error variable quantity of the output of observing nuclear propulsion system secondary circuit, set up the controlled performance evaluation value of control performance evaluation amount question blank according to the fuzzy evaluation rule list, according to this each control effect of control performance evaluation of estimate prediction output evaluation of result with reference to controlled quentity controlled variable, and, select the control input of optimum nuclear power unit secondary coolant circuit system at last through fuzzy decision according to the performance evaluation of control effect is revised current controlled quentity controlled variable;

(4) control signal that controller produced is exported to topworks after producing simulating signal and the enhancing of process signal amplifier through digital/analog converter;

(5) topworks carries out by instruction, total system is changed under the operating mode of appointment.

Multivariate integrated model Fuzzy Predictive Control method of the present invention can also comprise:

1, the state parameter information of described nuclear power unit comprises: steam generator outlet vapor pressure, secondary circuit feedwater flow, steam generator outlet steam flow and turbine speed.

2, described measuring system is temperature sensor, pressure transducer, flowmeter and velocity gauge.

3, described multistep prediction device is based on the prediction device of integrated model.

4, described predicated error is secondary circuit nozzle group valve cam angle aperture and water-supply valve aperture error.

5, described fuzzy evaluation rule list is meant a series of if-then (IF-THEN) language rule according to existing experience and technical know-how structure.

Principle of work of the present invention is: core of the present invention is the design of nuclear power unit secondary circuit multivariate integrated model Fuzzy Predictive Control system.This system is a nuclear power unit secondary circuit multivariate integrated model Predictive Control System and the combining of fuzzy control strategy.

Nuclear power unit secondary circuit multivariate integrated model Fuzzy Predictive Control system is based on the output in future of multistep prediction device prediction controlled system under the effect of reference controlled quentity controlled variable of integrated model, predicated error and predicated error variable quantity by the output of observing nuclear propulsion system secondary circuit, set up the controlled performance evaluation value of control performance evaluation amount question blank according to the fuzzy evaluation rule list, according to this each control effect of control performance evaluation of estimate prediction output evaluation of result with reference to controlled quentity controlled variable, and, select the control input of optimum nuclear power unit secondary coolant circuit system at last through fuzzy decision according to the performance evaluation of control effect is revised current controlled quentity controlled variable.As shown in Figure 1, the multivariate integrated model Fuzzy Predictive Control method of nuclear power unit secondary circuit of the present invention mainly is made up of the links such as multistep prediction device, fuzzy performance evaluation and fuzzy decision based on integrated model.Among Fig. 1, Be the P step predicted value of forecast model, 1≤j≤P wherein, P is a prediction step; E _j(k) and EC _j(k) be the P predicated error and the predicated error variable quantity in step; W (k+j) is a reference locus; Y (k) is the output of secondary coolant circuit system; The structured flowchart of integrated model as shown in Figure 2, integrated model is made up of the two large divisions: nonlinear Static DLF network and dynamic linear CARIMA model.

The first step, the training of nonlinear Static DLF network

The DLF network is on the basis of BP network its input layer and output layer to be contacted directly, and promptly each neuron of input layer all interconnects with output layer.Shown in DLF network structure Fig. 3, this network is three layers of feedforward network structure, and input layer U and output layer X have two nodes, and hidden layer H is a p node, u ₁(k) and u ₂The input of etching system when (k) being k, x ₁(k) and x ₂The output of etching system when (k) failing to k.If the threshold value of j unit of output layer is r _j, the connection matrix W ∈ R between output layer and each unit of hidden layer H ^{P * 2}, and the connection matrix W between output layer and the input layer U _FConnection matrix V ∈ R between hidden layer and each unit of input layer U ^{P * 2}

The DLF Learning Algorithms is as follows:

1) at first provides whole weight w _Il, v _LiWith neuron threshold value r _j, θ _iInitial value (i=1,2, L, p; J=1,2; L=1,2); P is the unit number of hidden layer.

2) by k sample value U (k)=[u ₁(k) u ₂(k)] ^TThe corresponding sequence number unit of input U layer; By weight matrix V, be to the activation value of each unit of H layer:

$h_i\left(k\right)=f_1(\underset{l=1}{\overset{2}{\mathrm{\Σ}}}{v}_{\mathrm{li}}{u}_l\left(k\right)+{\mathrm{\θ}}_i)---\left(1\right)$

In the formula, H (k) is the hidden layer output vector; The Sigmoid function f ₁(x)=(1+e ^-x) ^-1Be the hidden layer transport function.

3) H layer activation value h _i(k) pass through connection matrix W by weight matrix W and U layer _F, to the activation value of each unit of output layer X be:

$x_j\left(k\right)=f_2(\underset{i=1}{\overset{p}{\mathrm{\Σ}}}{w}_{\mathrm{ij}}{h}_i\left(k\right)+r_j)+\underset{m=1}{\overset{2}{\mathrm{\Σ}}}{w}_{\mathrm{Fjm}}{u}_m\left(k\right)---\left(2\right)$

Wherein, $W_F=\left[\begin{array}{cc}{w}_{F11}& w_{F12}\\ w_{F21}& w_{F22}\end{array}\right];$ f ₂(g) be the output layer transport function, same f ₁(g) equally be the Sigmoid function.

4) error of output layer X is:

e _j(k)＝x _j(k)(1-x _j(k))(d _j-x _j(k)) (3)

In the formula, d _jDesired output for X layer unit.

5) error of hidden layer H is:

$e_{\mathrm{hi}}\left(k\right)=h_i\left(k\right)(1-h_i\left(k\right))\underset{j=1}{\overset{2}{\mathrm{\Σ}}}{w}_{\mathrm{ji}}{e}_j\left(k\right),(w_{\mathrm{ji}}\∈W^T)---\left(4\right)$

6) being connected the weights increment between hidden layer H unit i and output layer X unit j is:

Δw _ij(k)＝ηh _i(k)e _j(k) (0＜η＜1) (5)

7) input layer U unit j with the weights increment that is connected between the output layer X unit j is:

Δw _Flj(k)＝ηx _j(k)e _j(k) (6)

8) the threshold value increment of output layer X unit j is:

Δγ _j(k)＝ηe _j(k) (7)

9) input layer U unit j with the weights increment that is connected between the hidden layer H unit i is:

Δv _ji(k)＝βu _j(k)e _hi(k) (0＜β＜1) (8)

10) the threshold value increment of hidden layer H unit i is:

Δh _i(k)＝ηe _hi(k) (9)

Double counting 2): the 10) step, until for sample number k=1,2, L, the error e of the output layer X of N _j(k) for enough little or be zero (k=N at this moment) till.

In second step, the on-line identification of the design of the multistep prediction device of linear segment and dynamic linear CARIMA model is as follows:

The CARIMA model that many inputs many defeated (are example with the two dimension) go out is:

A ₀₁(z ^-1)y ₁(k)＝B ₀₁₁(z ^-1)x ₁(k-1)+B ₀₁₂(z ^-1)x ₂(k-1)+ξ ₁(k)/Δ (11)

A ₀₂(z ^-1)y ₂(k)＝B ₀₂₁(z ^-1)x ₁(k-1)+B ₀₂₂(z ^-1)x ₂(k-1)+ξ ₂(k)/Δ (12)

In the formula, A ₀₁(z ^-1), A ₀₂(z ^-1), B ₀₁₁(z ^-1), B ₀₁₂(z ^-1), B ₀₂₁(z ^-1), B ₀₂₂(z ^-1) be z ^-1Polynomial expression;

ξ ₁(k) and ξ ₂(k) be white noise; Δ=1-z ^-1

Forecast model and Diophantine equation by formula (11)

1＝E _1j(z ^-1)A ₀₁(z ^-1)Δ+z ^-jF _1j(z ^-1) (13)

Can get predictive equation:

y ₁(k+j)＝E _1j(z ^-1)B ₀₁₁(z ^-1)Δx ₁(k+j-1)+E _1j(z ^-1)B ₀₁₂(z ^-1)Δx ₂(k+j-1)+F _1j(z ^-1)y ₁(k)+E _1j(z ^-1)ξ ₁(k+j) (j＝1，2，L，n) (14)

The optimum prediction of output is:

${\hat{y}}_1(k+j)=G_{11j}\left(z^{-1}\right)\mathrm{\Δ}{x}_1(k+j-1)+G_{12j}\left(z^{-1}\right)\mathrm{\Δ}{x}_2(k+j-1)+F_{1j}\left(z^{-1}\right)y_1\left(k\right)---\left(15\right)$

$(j=\mathrm{1,2},L,n)$

G _11j＝E _1jB ₀₁₁＝g ₁₁₀+g ₁₁₁z ^-1+L+g _11j-1z ^-j+1+L

In the formula,

G _12j＝E _1jB ₀₁₂＝g ₁₂₀+g ₁₂₁z ^-1+L+g _12j-1z ^-j+1+L

With formula (15)

Value resolves into k known quantity and unknown quantity two parts constantly, uses f ₁(k+j) expression known quantity, that is:

$\left[\begin{array}{c}{\hat{y}}_1(k+1)\\ {\hat{y}}_1(k+2)\\ M\\ {\hat{y}}_1(k+n)\end{array}\right]=\left[\begin{array}{cccc}{g}_{110}& & & 0\\ g_{111}& g_{110}& & \\ M& M& O& \\ g_{11n-1}& g_{11n-2}& L& g_{110}\end{array}\right]\left[\begin{array}{c}\mathrm{\Δ}{x}_1\left(k\right)\\ \mathrm{\Δ}{x}_1(k+1)\\ M\\ \mathrm{\Δ}{x}_1(k+n-1)\end{array}\right]$

$+\left[\begin{array}{cccc}{g}_{120}& & & 0\\ g_{121}& g_{120}& & \\ M& M& O& \\ g_{12n-1}& g_{12n-2}& L& g_{120}\end{array}\right]\left[\begin{array}{c}\mathrm{\Δ}{x}_2\left(k\right)\\ \mathrm{\Δ}{x}_2(k+1)\\ M\\ \mathrm{\Δ}{x}_2(k+n-1)\end{array}\right]+\left[\begin{array}{c}{f}_1(k+1)\\ f_1(k+2)\\ M\\ f_1(k+n)\end{array}\right]---\left(16\right)$

That is: ${\hat{Y}}_1=G_{11}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20071014469700171.png,C20071014469700174.png"\; state="\{\{state\}\}">X</figure-callout>}}_1+{\mathrm{<figure-callout\; id="12"\; label="G"\; filenames="C20071014469700151.png"\; state="\{\{state\}\}">G</figure-callout>}}_{12}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20071014469700171.png,C20071014469700172.png"\; state="\{\{state\}\}">X</figure-callout>}}_2+f_1---\left(17\right)$

${\hat{Y}}_1={[{\hat{y}}_1(k+1),{\hat{y}}_1(k+2),L,{\hat{y}}_1(k+n)]}^T$

f ₁＝H ₁₁Δx ₁(k)+H ₁₂Δx ₂(k)+F ₁y ₁(k)

In the formula, Δ X ₁=[Δ x ₁(k), Δ x ₁(k+1), L, Δ x ₁(k+n-1)] ^T,

F ₁＝[F ₁₁，F ₁₂，L，F _1n] ^T

ΔX ₂＝[Δx ₂(k)，Δx ₂(k+1)，L，Δx ₂(k+n-1)] ^T

$G_{11}=\left[\begin{array}{cccc}{g}_{110}& & & 0\\ g_{111}& g_{110}& & \\ M& M& O& \\ g_{11n-1}& g_{11n-2}& L& g_{110}\end{array}\right],$ ${\mathrm{<figure-callout\; id="12"\; label="G"\; filenames="C20071014469700151.png"\; state="\{\{state\}\}">G</figure-callout>}}_{12}=\left[\begin{array}{cccc}{g}_{120}& & & 0\\ g_{121}& g_{120}& & \\ M& M& O& \\ g_{12n-1}& g_{12n-2}& L& g_{120}\end{array}\right],$

$H_{11}=\left[\begin{array}{c}{G}_{111}-g_{110}\\ Z(G_{112}-g_{111}{Z}^{-1}-g_{110})\\ M\\ Z^{n-1}(G_{11n}-g_{11n-1}{Z}^{-n+1}-g_{110})\end{array}\right],$ ${\mathrm{<figure-callout\; id="12"\; label="H"\; filenames="C20071014469700151.png"\; state="\{\{state\}\}">H</figure-callout>}}_{12}=\left[\begin{array}{c}{G}_{121}-g_{120}\\ Z(G_{122}-g_{121}{Z}^{-1}-g_{120})\\ M\\ Z^{n-1}(G_{12n}-g_{12n-1}{Z}^{-n+1}-g_{120})\end{array}\right].$

In like manner can get by formula (12):

${\hat{Y}}_2=G_{21}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="1"\; label="X"\; filenames="C20071014469700171.png,C20071014469700174.png"\; state="\{\{state\}\}">X</figure-callout>}}_1+G_{22}\mathrm{\&Delta;}{\mathrm{<figure-callout\; id="2"\; label="X"\; filenames="C20071014469700171.png,C20071014469700172.png"\; state="\{\{state\}\}">X</figure-callout>}}_2+f_2---\left(18\right)$

In the formula ${\hat{Y}}_2={[{\hat{y}}_2(k+1),{\hat{y}}_2(k+2),L,{\hat{y}}_2(k+n)]}^T,$

F ₂＝[F ₂₁，F ₂₂，L，F _2n] ^T

f ₂＝H ₂₁Δx ₁(k)+H ₂₂Δx ₂(k)+F ₂y ₂(k)

$G_{21}=\left[\begin{array}{cccc}{g}_{210}& & & 0\\ g_{211}& g_{210}& & \\ M& M& O& \\ g_{21n-1}& g_{21n-2}& L& g_{210}\end{array}\right],$ $G_{22}=\left[\begin{array}{cccc}{g}_{220}& & & 0\\ g_{221}& g_{220}& & \\ M& M& O& \\ g_{22n-1}& g_{22n-2}& L& g_{220}\end{array}\right],$

$H_{21}=\left[\begin{array}{c}{G}_{211}-g_{210}\\ Z(G_{212}-g_{211}{Z}^{-1}-g_{210})\\ M\\ Z^{n-1}(G_{21n}-g_{21n-1}{Z}^{-n+1}-g_{210})\end{array}\right],$ $H_{22}=\left[\begin{array}{c}{G}_{221}-g_{220}\\ Z(G_{222}-g_{221}{Z}^{-1}-g_{220})\\ M\\ Z^{n-1}(G_{22n}-g_{22n-1}{Z}^{-n+1}-g_{220})\end{array}\right].$

Wherein, according to recurrent least square method, can pick out matrix G ₁₁, G ₁₂Parameter, in like manner can get G ₂₁, G ₂₂Parameter.

The 3rd step, fuzzy performance evaluation and fuzzy decision

The predicated error of observing nuclear propulsion system secondary coolant circuit system output is not only wanted in the fuzzy performance evaluation, also should observe its predicated error variable quantity simultaneously.According to expert's control effect assessment experience, constitute the control performance evaluation rule then through further processing, put in order, refine the back.Form the fuzzy evaluation rule list shown in table 1, table 2.Again by between the fuzzy control rule or relation, can calculate the fuzzy performance evaluation amount by the control law in the fuzzy control rule table, adopt method of weighted mean, change the performance evaluation amount into accurate amount by fuzzy quantity, obtain fuzzy performance and estimate question blank.Estimate the controlled performance evaluation value of question blank by control performance at last, select the input quantity of optimum nuclear power unit secondary coolant circuit system according to the control performance evaluation of estimate.

The invention has the advantages that to be applicable to that nuclear power unit has the system of severe nonlinear, coupling, time variation like this, the control accuracy height, robustness is good.

(4) description of drawings

Fig. 1 is the structured flowchart of the multivariate integrated model Fuzzy Predictive Control of nuclear power unit secondary circuit;

Fig. 2 is the structured flowchart of integrated model;

Fig. 3 is the DLF network structure;

Fig. 4 is that nuclear power unit secondary circuit multivariable nonlinearity Predictive Control System is implemented figure;

Fig. 5 is under multivariate integrated model Fuzzy Predictive Control, and load changes from 100%～20% o'clock nuclear power unit secondary circuit parameter.Wherein: Fig. 5-a is that steam generator outlet vapor pressure changes, Fig. 5-b is that feedwater and steam flow change, Fig. 5-d is that steam turbine power changes for the variation of steam generator outlet steam temperature, Fig. 5-c;

Fig. 6 is steam turbine fuzzy set A _vThe subordinate function table;

Fig. 7 is steam turbine fuzzy set B _vThe subordinate function table;

Fig. 8 is steam turbine fuzzy set N _vThe subordinate function table;

Fig. 9 is a turbine speed control performance evaluation rule table;

Figure 10 is a steam turbine control performance evaluation amount question blank;

Figure 11 is steam generator fuzzy set B _pThe subordinate function table;

Figure 12 is steam generator fuzzy set B _pThe subordinate function table;

Figure 13 is steam generator fuzzy set N _pThe subordinate function table;

Figure 14 is a steam generator control performance evaluation rule table;

Figure 15 is a steam generator control performance evaluation amount question blank.

(5) embodiment

For example the present invention is done description in more detail below in conjunction with accompanying drawing:

The training of 1 static DLF neural network

Nuclear power unit secondary circuit nozzle group valve cam angle aperture Δ θ and water-supply valve aperture

Input as the DLF network; The intermediate variable x of nonlinear element ₁And x ₂Be the output of DLF network.BP in the DLF network adopts the 2-10-2 network structure, and after 25000 training, cumulative errors is no more than anticipation error.

If the threshold value of j unit of output layer is r _j, the connection matrix W ∈ R between output layer and each unit of hidden layer H ^{P * 2},

And the connection matrix W between output layer and the input layer U _FConnection matrix V ∈ R between hidden layer and each unit of input layer U ^{P * 2}

1) weight matrix between input layer and the hidden node is as follows:

v[2][10]＝{{39.908，26.379，55.44，13.008，72.505，47.756，22.97，12.49，18.285，52.653}；{17.225，31.66，14.27，5.75，12.91，23.89，18.45，15.01，28.529，37.148}}

2) weight matrix between hidden node and the output layer node is as follows:

w[10][2]＝{{66.73，17.47}；{50.52，31.03}；{13.82，26.73}；{31.031，10.526}；{40.012，21.389}；{88.314，15.133}；{18.961，33.284}；{49.81，68.27}；{72.64，20.142}；{98.67，38.097}}

3) the hidden node threshold value is respectively:

h[10]＝{41.633，23.576，13.283，54.143，31.32，21.47，38.654，13.508，14.453，59.78}

4). output layer node threshold value is:

r[2]＝{46.348，16.816}

5) the connection matrix W between output layer and the input layer U _FFor:

$W_F=\left[\begin{array}{cc}{w}_{F11}& w_{F12}\\ w_{F21}& w_{F22}\end{array}\right]=\left[\begin{array}{cc}128.772& 48.972\\ 114.54& 37.905\end{array}\right]$

2 fuzzy evaluations

Since controlled device have very strong non-linear and the time stickiness, in order to obtain the excellent control performances fuzzy evaluation, not only to observe the predicated error of turbine speed and steam generator outlet vapor pressure, also should observe its predicated error variable quantity simultaneously.

1) fuzzy evaluation of turbine speed control effect

The turbine speed error e _vLinguistic variable be E _v, its domain X _vOn fuzzy set be A _v, domain X _vFor:

X _v＝{-80，-60，-40，-20，0，20，40，60，80}

e _vFuzzy subset A _VlBe divided into: { PB, PM, PS, ZO, NS, NM, NB}

Wherein: PB, PM, PS, O, NS, NM, NB represent respectively honest, center, just little, zero, negative little, negative in, greatly negative.

Fuzzy set A _vThe degree of membership assignment see the table 1 of Fig. 6:

The turbine speed error changes ec _vLinguistic variable be EC _v, domain is:

Y _v＝{-3，-2，-1，0，1，2，3}

Ec _vFuzzy subset B _ViBe divided into: { PB, PS, ZO, NS, NB}, fuzzy set B _vThe degree of membership assignment see the table 2 of Fig. 7.

The turbine speed control performance is estimated c _vLinguistic variable be C _v, domain is: Z _v=0,2,4,6,8,10}.C wherein _vBe worth more for a short time, represent that the control effect of corresponding controlled quentity controlled variable is good more.

c _vFuzzy subset N _VjBe divided into: { ZO, PS, PM, PB}, fuzzy set N _vThe degree of membership assignment see the table 3 of Fig. 8.

According to expert's control effect assessment experience, constitute the control performance evaluation rule through further processing, put in order, refine the back.Usually can be abbreviated as a table, be referred to as the fuzzy evaluation rule list.Turbine speed control performance evaluation rule is seen the table 4 of Fig. 9.

Between the above-mentioned description fuzzy control rule be or relation, can calculate C by the control law in the fuzzy control rule table _Ti(i=1, Λ, 35), then fuzzy set C _TBe expressed as: C _T=C _T1+ C _T2+ Λ+C _T35

The fuzzy performance evaluation amount that is calculated by following formula adopts method of weighted mean, changes the performance evaluation amount into accurate amount by fuzzy quantity.Can obtain capacity then and be 7 * 7 fuzzy performance and estimate question blank, see the table 5 of Figure 10.

2) fuzzy evaluation of steam generator outlet vapor pressure control effect

Steam generator outlet vapor pressure error e _pFuzzy variable be E _p, its domain N _pOn fuzzy set be B _p

Domain is:

N _p＝{-0.6，-0.4，-0.2，0，0.2，0.4，0.6}

Fuzzy set B _pThe degree of membership assignment see the table 6 of Figure 11:

Outlet vapor pressure error is changed to ec _pLinguistic variable be EC _p, its domain Y _pOn fuzzy set be B _p, domain is: Y _p=3 ,-2 ,-1,0,1,2,3}, fuzzy set B _pThe degree of membership assignment see the table 7 of Figure 12.

Steam generator outlet vapor pressure control performance is estimated c _pLinguistic variable be C _p, domain is: Z _p=0,2,4,6,8,10}.C wherein _pBe worth more for a short time, represent that the control effect of corresponding controlled quentity controlled variable is good more.c _pFuzzy subset N _PjBe divided into: { ZO, PS, PM, PB}, fuzzy set N _pThe degree of membership assignment see the table 8 of Figure 13.Then obtain steam generator outlet vapor pressure control performance evaluation rule table and control performance and estimate the table 9 that question blank is seen Figure 14, the table 10 of Figure 15.

Last controlled according to the method described above amount and corresponding prediction output are estimated the controlled performance evaluation value of question blank by control performance, select the input quantity of optimum nuclear power unit secondary coolant circuit system according to the control performance evaluation of estimate.According to shown in Figure 4, the multivariate integrated model Fuzzy Predictive Control system of nuclear power unit secondary circuit carries out simulation study under load down (100%-20%) disturbance, get PREDICTIVE CONTROL parameter n=10, m=2, λ=[1,0; 0,1], α=0.3; Sampling period is 0.1 second, obtains simulation curve as shown in Figure 5.

Present embodiment shows, when nuclear power unit carries out load variations, use the multivariate integrated model Fuzzy Predictive Control system of nuclear power unit secondary circuit after, the control accuracy height of whole Nuclear Power System, robustness is good.

Claims (6)

1, a kind of nuclear power device two-loop multi-variable integrated model fuzzy predication control method, it is characterized in that: mainly form by links such as multistep prediction device, fuzzy performance evaluation and fuzzy decision based on integrated model, described integrated model is made up of the two large divisions, i.e. nonlinear Static DLF network and dynamic linear CARIMA model; The DLF network is on the basis of BP network its input layer and output layer to be contacted directly, and promptly each neuron of input layer all interconnects with output layer; Come identification by recurrent least square method, can obtain dynamic linear CARIMA model;

(1) utilize measuring system to measure the state parameter information of nuclear power unit;

(2) convert the state parameter that obtains to digital signal by analog/digital converter, digital signal is given controller through behind the wave filter;

(3) the multivariate integrated model fuzzy prediction algorithm that comprises in the controller is under the effect of reference controlled quentity controlled variable, the output in future of multistep prediction device prediction controlled system, the predicated error and the predicated error variable quantity of the output of observing nuclear propulsion system secondary circuit, set up the controlled performance evaluation value of control performance evaluation amount question blank according to the fuzzy evaluation rule list, according to this each control effect of control performance evaluation of estimate prediction output evaluation of result with reference to controlled quentity controlled variable, and, select the control input of optimum nuclear power unit secondary coolant circuit system at last through fuzzy decision according to the performance evaluation of control effect is revised current controlled quentity controlled variable;

(4) control signal that controller produced is exported to topworks after producing simulating signal and the enhancing of process signal amplifier through digital/analog converter;

(5) topworks carries out by instruction, total system is changed under the operating mode of appointment.

2, nuclear power device two-loop multi-variable integrated model fuzzy predication control method according to claim 1 is characterized in that: the state parameter information of described nuclear power unit comprises: steam generator outlet vapor pressure, secondary circuit feedwater flow, steam generator outlet steam flow and turbine speed.

3, nuclear power device two-loop multi-variable integrated model fuzzy predication control method according to claim 2 is characterized in that: described measuring system is temperature sensor, pressure transducer, flowmeter and velocity gauge.

4, nuclear power device two-loop multi-variable integrated model fuzzy predication control method according to claim 3 is characterized in that: described multistep prediction device is based on the prediction device of integrated model.

5, nuclear power device two-loop multi-variable integrated model fuzzy predication control method according to claim 4 is characterized in that: described predicated error is secondary circuit nozzle group valve cam angle aperture and water-supply valve aperture error.

6, nuclear power device two-loop multi-variable integrated model fuzzy predication control method according to claim 5 is characterized in that: described fuzzy evaluation rule list is meant a series of if-then language rules according to existing experience and technical know-how structure.

Images