Զѧ
ACTA AUTOMATICA SINICA
1999 25 6 Vol.25 No.6 1999



Ԥϵͳģͱʶ
̬Ż1)

ժҪöǰ磬Եվ¯תʽԤϵͳڲͬʱϵͳ̬ģͽ˱ʶ.Ƚ˷¡ѹϵͳӰ.˻ĹϵͳģͣԴģΪ̬Ż㷨ʵ˻תʽԤϵͳ̬ŻƣЧ.
ؼʡ磬ϵͳʶ̬ŻƣԤ.
MODEL IDENTIFICATION AND STEADY-STATE OPTIMIZING
CONTROL OF WIND SYSTEM FOR AIR PREHEATER

LIU Ding
(Xi'an University of Technology,Xi'an710048)
WAN Baiwu
(Xi'an Jiaotong University,Xi'an710049)
AbstractThis paper uses a multi-layer forward neural network to identify the steady-state model of air preheater of different inputs.The effects of the wind temperature and wind pressure on the system model are compared,and then the neural network wind system model is given.Based on this model,an on-line automatical model learning method and steady-state optimizing algorithm are proposed for the wind system of air preheater.
Key wordsNeural network,system identification,steady-state optimizing control,air preheater.

1
תʽԤǴͻ糧й㷺ʹõ豸ԭͼ1ʾ.ڱṹԭʹùдеζ©⣬ʹ.ף1˽һģƷʵʹյõЧ.Ľһоϵͳзѹ븺ɼĹϵʶһӿƶģͣԴΪʵϵͳʵʱŻƣʹ㸺ҪʱһηΪС.ﵽ˽ܽġ߹¯ȼЧʵĿ.


ͼ1תʽԤԭͼ
2ϵͳģͱʶ
ʹʶϵͳ̬Եԭͨɼϵͳ̬롢ݹһѧϰѵʹƽϵͳ̬ԣֱƽҲɿΪһӳ任.BPܹʵӳ任ǰõһ.ԿԤϵͳһǰ磬ڸĽBPѧϰ㷨ʵʲõݣѵõϵͳ̬ģ.
2.1ĽBPѧϰ㷨
ĽBP㷨Ҫ˼ѧϰӦԵģµĹϵƣ2

ЦǦ(n)ĵϵԦ(n)ƫ.
㷨ҪȱǼѵûԵؼ.˿ɿһӦ䲽ĿBP㷨3ݣ˼ǣεݶȵķ෴ζһ½࣬Ӧ䷣С෴εʹͬ½ˣûдﵽС㣬Ӧ佱Ӳ.ͨݶȷű仯ƵϢʵӦ䲽㷨.£
1)Ȩֵֵгʼ
2)(a)㵽μ(b)㡢㷴E/ui;
3)Ȩֵ
(1a)
(1b)
(1c)
(1d)
ʽ0ţ1ȡ0.20.3ʵΪӣȡ0.80.9

Ȩֵ޸ʱÿȨϼһǰμȨ仯ֵʹڹ͹ķ仯İڶ𵽻ƽ.
2.2ѹʱϵͳģͱʶ
ֲֻתʽԤΪһηͶη.һηѹͶηѹı仯Թ¯Ӱܴ󣬶֮ʸӵķԹϵ.Ϊʵ̬ŻƣѡBPϵͳнģϵͳ̬ģ
(2)
x1,x2,x3,x4ֱʾѹͷ£Pʾ¯ɣWΪȨֵ.
ͼ1ʾϵͳУԷ¶Թ¯ɵӰ죬Ƿѹϵͳ.ʶĽṹͼ2ʾ.ԪֱӦһηѹx1Ͷηѹx2һԪӦڹ¯PмԪĿĿѡ4.


ͼ2ģͽṹͼ
ѵף3ݸķȨֵ㷨(1)ʽʾ.ֳȡ120ݣȥݼܵĶ̬ݣѡȡ80Ϊѵ30Ϊ.ϵͳı仯ΧȽϼУһηѹ仯ΧΪ12.0kp15.2kp;ηѹ仯ΧΪ1.4kp4.0kpɱ仯ΧΪ184MW302MW.Ϊݽѹ
(3)
(4)
(5)
ʵ֤ݴһĴ=x/xmaxԽʹٶȼӿ.Լ3000εѵѵΪ0.001£ܾΪ10-5.ʱõģͼΪϵͳڷѹʱѧģ.ʣ30Ϊݴ齨ģĽ35%5%£˵õģ;нϸߵľ.
2.3ѹʱϵͳģ
ǰ潨ϵͳ̬ģͣں˷µӰ죬ʲȫطӳϵͳģ͵ڹ.ʵϣڽģпһηºͶηµӰʮֱҪ.ڿǷ¶ϵͳӰʱԲͼ2ʾ磬ϵͳԪĿΪ4ֱӦһηѹηѹһηºͶη£ԪĿΪ1Ӧڹ¯.ѵͬøĽBP㷨ѵҲģͬ.Լ10000ѵʱ3СʱѵΪ0.001£ܾΪ10-5.ʣµûвμѵݴе5%£Ϊ4.98%󲿷ֵ12%.ģȣģ.
ͨԵվ¯Ԥϵͳлģͱʶһϵͳģͻ֤˻ĸɲԭϵ©йأһԤеķѹͷйأΪһʵֶϵͳ̬ŻƵ춨˻.
3Ԥϵͳ̬Ż
ʶõģͣñŻ㷨ͿԵõһµһηѹͶηѹ.ϵͳŻĿڻ鸺Ϊһ£ʹһηѹͶηѹ֮С.ѹ֮СζһηͶηС빦ҲӦСӶﵽ硢Сõ硢ܽĵĿ.
3.1ϵͳŻ
ϵͳģΪP=f(x1,x2,x3,x4)W.ʵʹx1,x2,x3,x4Ӧ±Լx1(12.015.2)x2(1.84.0)x3(300340)x4(310360)ԼP=f(x1,x2,x3,x4)W.
ѡһηѹηѹ֮ΪϵͳŻĿ꺯J=x1+x2.ˣϵͳŻ£
min:J=x1+x2,(6)
ʽԼP=f(x1,x2,x3,x4)W,
ʽԼ12.0x115.2
1.8x24.0
300x3340,
320x4360,
ʽPΪһҪ󸺺ɣһPӦ180P310.
ͨŻ⣬ԵóһµһηѹͶηѹx*1=f1(P),x*2=f2(P).ϵͳſ.ͨʵ֤ڻ̬зµޱ仯ϵͳӰСڶԱϵͳ̬ŻʱһηºͶη½Ϊϵͳ״ֻ̬ҪԼ.
3.2Ż㷨
һŻȣϵͳŻص㣺
1)ϵͳģͼʽԼPϵͳŻǲ仯ģΪڷйУɵɾģһ㲻Ϊı䣻
2)x1,x2ĿԽСüֱѰţ
3)ڱϵͳŻҪߡʵʱУˣŻ㷨ˣнǿĿɶԣԱ.
ص㣬ӼŻ̡ԿϵͳͶʵĿĳڲǷ¶ϵͳӰ£(6)ʽϵͳŻΪ
minJ=x1+x2,(7)
ʽԼP=f(x1,x2)w,
ʽԼ12.0x116.0
1.5x24.5.
Ϊx1X1=12.016.0ݣx2X2=1.54.5ݣԤ©ϵͳŻ㷨£
1)˹ʶõϵͳģP=f(x1,x2)|wװϵͳλУ鸺ȶʱɼѹźţһµ(P,x1,x2)|w,x1X1,x2X2,ԭϵͳ̬ģͽõµģ
P=f(x1,x2)wN;(8)
2)x1x2ĿһֳΪnmС䣬xi1xj2(xi1,xj2)X1X2,in,jm;
3)ÿһ(xi1,xj2),(8)ʽ
PK=f(x1,x2)|wN,K=1,2,nm,(9)
PKPбȽϣPK-PܦţֵСڻĳ趨ֵ򽫴PKֵ
4)PK-PܦŵɸPKУμJ=x1+x2,ҳϵͳָ꺯СֵminJʱӦһ(x*1,x*2)ΪſƱԴ˱Ϊźϵͳ󣬷һηѹͶηѹ
5)ʱ鸺ֵĳһȶʱظ̣ҳӦ.
ڻйУŻ㷨ʵϵͳкʵ飬õ˻ڲͬ(180MW300MW)ʱһηѹߺͶηѹߣͼ3ʾ.


(a)һηѹ/


(b)Ŷηѹ/
ͼ3һηѹͶηѹ
3.3ϵͳŻƵʵ
ƵСͷֲʽϵͳλ80C31ƬΪģݲɼ봦ʾԶлϱͨѶȹ.λУϵͳķѹԼ¯ĸɶȽ̬ģͽѧϰ.ڹ¯ɺͷѹµı仯߱ʶģ͵Ļϣÿѧϰֻһѧϰʱ̣Ӱϵͳ.ڽŻ󣬵óҪĸµһηѹͶηѹ˱ϵͳ(INF90ϵͳ)Ӧźţֱйز(ܷ)Ա֤Ҫʵ.λ˵˱֤ļ϶С⣬Ҫһηη֮ļ϶֤ѹϵʹϵͳ״̬.ͼ4Ϊϵͳͼ.


ͼ4ϵͳͼ
3.4ʵ
ĳ糧AԤΪʵݼ±.ɱɼ̬ŻͶ󣬲ϵͳ©ԼСһηҲв̶ͬȵļСӶﵽ˽ԼõĿ.
1̬ŻͶʵ

ĿͶǰͶ˺
鸺(MW)290290
RO2(%)16.0315.1
RO2(%)14.1113.5
©A1(%)16.411.3
һη(A)9165
ͷ(A)8861
(A)177174


4
Ľ˵վ¯תʽԤϵͳ̬ģͣʵϵͳ̬ŻƵķ.ŻϵͳͶʹúڱ֤ʱʹһηͷֱ½23%20%Чʮ.ͬʱڿƷԭ򵥣ɲǿֳԱգӶ֤˿ϵͳͶ.ϵͳȫ20ͻ糧ͶʵУۼƴֱӾЧԪΪҿƼɹصƹƻĿ1997ҿƼȽ.
1)ҿƼɹصƹƻĿ.
߼飺1957.1997ͨѧϵͳоʿѧλΪѧԶϢѧԺԺ.ڴ¹ҵԶܿӦõȷоĿǰеص񣬷20ƪҼʡƼ4.
塡171.
ߵλѧԶѧԺ710048
塡ͨѧϵͳо710049
ο
1Liu D,Wan B W.On-line coordinative fuzzy control of active leakage control system for power station boiler.IFAC/IFORS/IMACS Symposium on Large-Scale Systems Theory and Application,London:Pergamon press,1995,2:386392
2壬.ҵ̼ߵݽ̬Ż.ѧ磬1998.222229
3ǳ.ѧϰļ㷨().ձ鱨ͨѶѧᴺļձ鱨ͨѧᣬ1992(6)157158
4Hunt K J.Neural networks for control systemA survey.Automatica,1992,28(6):10881111
5壬֡.ģҵ̬ݽŻ.Զѧ199016(2)186192
ոڡ1998-03-11޸ĸڡ1998-10-13
