ѧѧ
ACTA SCIENTIAE CIRCUMSTANTIAE
1998 18 3ڿƼڿ
ISAPEC֪ʶԶȡʵ*
ӱ
(廪ѧܼоԺ 100084)
(貿ˮԴģ 100007)
ժҪ ˹ʵִȾƹ滮ܾ֧ϵͳ(ISAPEC)֪ʶԶȡıҪԣISAPEC֪ʶԶȡʵֹ̣һʵ֤.˹ʵ֪ʶԶȡ֪ʶ˹ȡȼȾиߵʱЧʣܱ֤ߵ.
ؼ ƹ滮ܾ֧ϵͳ˹磻֪ʶԶȡ.
ɳչΪ෢չսԳΪȫĹʶǶȾչơƻʵ־óչĸ֤.ȾԴȾƹ滮ΪȵһΪȾƹ滮ܾ֧ϵͳԴΪҹȾƵ˳ʵʩṩЧļ֧.¶ۺӦþ֧ϵͳרϵͳĴȾƹ滮ܾ֧ϵͳISAPEC(Intelligent Dicision Support System of Air Pollution Total Emission Control Planning)˿о12.ISAPECݿϵͳ֪ʶϵͳģͿϵͳͼοϵͳܿϵͳ.֪ʶϵͳISAPECĺģ֪ʶ⡢ڰ塢֪ʶϵͳԼ֪ʶԶȡȷƣ12.ĶISAPEC֧ͬϵͳһص㼴֪ʶԶȡʵֹ̽н.
1 ISAPEC˹ʵ֪ʶԶȡıҪԷ
֪ʶϵͳȡ֪ʶȻ֪ʶĻȡ൱ѵģѳΪ֪ʶϵͳƿ.Ϊ(1) ֪ʶʦܽᡢרҵ֪ʶ軨Ѵʱ;(2) רҵ֪ʶ֮以ìʱ֪ʶʦ޲ߣ(3) ʱרҽ֪ʶδغܺãҲдʹ֪ʶϵͳõ֪ʶܳ.
չʮŵĽѧơ˹34ݿԺܺõؽ֪ʶϵͳ֪ʶȡ.ν˹ǲʵֵеļģĳЩṹ빦ܣڹ̻.˹۵㲻ȥظϸ磬ǲõĲ˷Ŀǰϵͳܽ⣬ѧϰʶͿƵ.
˹ȡĲŵ㣬˾ص㣺(1) ˹ڽṹɺܶСĴԪӶɣԪġеĻõԤڵʶ𡢼ĽнϿٶȣ(2) ˹кǿݴԣ(3) ˹ϢǴԪ֮ȨУӵȨпϢݣǷֲʽĴʽ(4) ˹ʮǿѧϰܣ˹еȨӵĽṹͨѧϰõ.
Щصʹ˹ʵ֪ʶԶȡͲҪ֪ʶʦܽᡢרҵ֪ʶֻҪרҽʵѵ磬ʹͬ˹ܹרҸĽ𾡿ͬ.
2 ISAPEC˹ʵ֪ʶԶȡʵ
˹ĽṹֶISAPECвRumelhart1986ķԱ任Ԫɵǰ.УԪǷֲеģÿԪֻǰһԪ.һΪ㣬һΪ㣬м㣬һ.Ԫ롢Sigmoid.в÷򴫲(BackPropagation)ѧϰ㷨ҲΪBP.
BPУԪ룬ԪOjNjĹϵΪ
(1)
Цjʾֵ.
ĳһԪNjһԪOiĹϵΪ
(2)
ʽУWjiĳһjԪһĵiԪȨֵ.ȨϵֵĵʽΪ
Wkj(n+1)=ǦkOj+W(n)kj(3)
(n+1)k=Ǧk+(n)k(4)
ȨϵֵĵʽΪ
W(n+1)ji=ǦjOi+W(n)ji(5)
(n+1)j=Ǧj+(n)j(6)
Уk=(tk-Ok)Ok(1-Ok)(7)
j=Oj(1-Oj)kkWkj(8)
ʽУǡֱΪѧϰƽӣ
tkkԪĿֵ.
ISAPECеķ򴫲ѧϰͼ1ʾ.ѧϰʧܣԵ㵥Ԫѧϰʺƽӵȣѧϰ.ѧϰɹͨѧϰԶȡ֪ʶԪȨϵֵʹ֪ʶ.֪ʶе֪ʶʵĹ̲߼һֲмĹ.¼裺(1) ݼӵĸԪ(2) Ԫ֪ʶе֪ʶݼԪԪ(3) Ԫ֪ʶе֪ʶԪԪ.

ͼ1 ISAPEC֪ʶԶȡʵֹ
Fig.1 The process of automatic acquisition of knowledge in ISAPEC
3 ISAPEC˹ʵ֪ʶԶȡʵ
ô(ݶ򡢵Ũȼļд)˵ISAPEC֪ʶԶȡʵֹ̣Զȡ֪ʶ֤ܽ.
ôʵ֪ʶԶȡʱȷṹѧϰ¼棺ȷ㵥ԪΪ4δֵ(mg/m3)ֵ(mg/m3)΢ֵ(mg/m3)ͽ¾ֵĵ(km2*/t)ȷΪ1㵥ԪΪ4ȷ(1 0 0 0)(0 1 0 0)(0 0 1 0)(0 0 0 1)4ģʽֱ4಻ͬĴһģʽĴΪãڶִ֮ΪȷѧϰʺƽӷֱΪ0.80.7ȷΪ20000ȷҪΪ0.00001ϵͳƽҪΪ0.00005.ʵʱЩͨһԻ.Ȼף5ݻȡ1ʾѧϰ.
ͼ1ʾѧϰ̽ѧϰڵΪ17909ʱѧϰɹõĸԪȨϵֵ㹹˴֪ʶе֪ʶ(1).֪ʶԶȡʵ֮󣬱ô֪ʶе֪ʶ210ֵ(10ֵԴף5).ֵΪ轫֮תΪɢӶжĴ.ɱ2֪ʵһµ.
1 ѧϰ
Table 1 Samples for the learning in the neural network of atmospheric environmental quality classification

ֵĿֵ
SO2,mg/m3NOx,mg/m3TSP,mg/m31/ km2*/t
10.4630.0520.3920.0584(0 0 0 1)
20.3030.0920.6410.0384(0 0 0 1)
30.1210.0480.4190.0264(0 0 0 1)
40.1310.0760.3980.0314(0 0 0 1)
50.1170.1020.3400.0564(0 0 0 1)
60.1990.0510.2690.0703(0 0 1 0)
70.1380.0390.3540.0573(0 0 1 0)
80.0600.1070.2970.1013(0 0 1 0)
90.0980.0700.3370.0723(0 0 1 0)
100.1070.0660.2340.0783(0 0 1 0)
110.0500.0320.3310.1103(0 0 1 0)
120.0400.0560.2640.0592(0 1 0 0)
130.0410.0400.2690.0672(0 1 0 0)
140.0350.0180.3680.0432(0 1 0 0)
150.0990.0240.2150.0822(0 1 0 0)
160.0380.0400.3090.1712(0 1 0 0)
170.0860.0110.1880.0832(0 1 0 0)
180.0390.0180.2680.1332(0 1 0 0)
190.0310.0160.1340.2431(1 0 0 0)
200.0070.0140.0900.1341(1 0 0 0)

2 ôԶȡ֪ʶʵĽ
Table 2 Results of the application of neural network of atmospheric environmental quality classification

䡡롡ֵ       ʵʵĴ

  
SO2,
mg/m3NOx,
mg/m3TSP
mg/m31/
km2*/t䡡ֵ

10.3510.0650.0350.056(00.0010690.0421490.962124)44
(0001)
20.1250.0840.3220.019(00.0017020.0149390.986333)44
(0001)
30.0890.0570.6630.028(00.0073240.0007770.999314)44
(0001)
40.1500.0360.4830.023(00.0071420.0008150.999278)44
(0001)
50.1800.0360.2220.079(001.0000000)33
(0010)
60.1280.0350.3070.063(001.0000000)33
(0010)
70.0650.0220.1910.096(0.0015350.9982020.0000170)22
(0100)
80.0540.0460.2120.089(0.0003250.9991100.0000580)22
(0100)
90.0380.0400.3090.171(0.0000010.9822310.0113140)22
(0100)
100.0070.0190.0990.237(0.9995920.00384700)11
(1000)

4 
֪ʶ˹ȡȣ˹ʵֵԶ֪ʶȡȾиߵʱЧʣܱ֤ߵ.֪ʶϵͳгΪƿ֪ʶȡͨ˹͵õ˱ȽϺõĽٲƿ.ISAPECУͨ˹ȡ֮ʾΪ֪ʶͨӦʵʵԶȡ.йشȾʩѡ֪ʶȡͨȾʩѡʵ֣Ӷ֤ISAPECṩĹ滮ӿѧЧ.
*һֿƼչĿ
ο
1 ӱ.Ⱦƹ滮ܾ֧ϵͳ۲ʿѧλģ.廪ѧ1996
2 ӱ.Ⱦƹ滮ܾ֧ϵͳĿо.йѧ199717(1)14
3 .˹ģͼӦ.Ϻѧ磬1992
4 ŵ.뼼˹.ʵ磬1992
5 й༭ίԱ.й1994.й磬1994184186

AUTOMATIC ACQUISITION OF KNOWLEDGE IN ISAPEC
Chen Wenying
(Institute of Nuclear Energy Technology,Tsinghua University,Beijing 100084)
Hou Dun
(Urban Water Resources Center,Ministry of Construction,Beijing 100007)
ABSTRACT The necessity of using the artificial neural network to obtain the knowledge automatically in the intelligent decision support system of air pollution total emission control planning (ISAPEC) is expounded.A process of designing and achieving automatic acquisition of the knowledge in ISAPEC is introduced.The function of the automatic acquired knowledge is tested through a case study.
Keywords total emission control planning,intelligent decision support system,artificial neural network,automatic acquisition of knowledge.
1997-01-14յԭ
1997-09-04յ޸ĸ
