Modeling and Designing a Real-Time Event Manager in a Heterogeneous Distributed System

Cop o"i!(h © IF.- C Ili!(i ;,( Compll l' Appli ca io Il s o P o c "ss Con ol. 'ie llll ; l. Au s ia. l ~ I H:) APPLICATION OF A SELF-TUNING REGULATOR TO A SOLAR POWER PLANT F. R. Rubio*, E. F. Camacho* and R. Ca mona** " ~ ) '/)(I iI '/II1(,1I11i ,/" " ~ I/I li ll l/i lll ll /:/ ' ln i ll/ I "11, /: 1S ,/,, / /I,L:" 'l I/I'I"I1I / /I /I/,I n(dn , { '/li" ,' ' l im/ i" S ("'I//(/ , ,I /)(Ii/l ** / /lI I'I"I III II (III( d 7",,1 (//I / [ ,'(//1/(/1 1(111 7i'(/I/I, i" /(/ /,/(1111(/ , IS/'S I(' ,-// 11 1''1/(1, ,1 //(/11/ ~bs ac~. This pape p esen s an applica i on o a sel - uning egula o o he dis ibu ed colle c o ield o a s o la powe plan . The dis ibu ed collec o ield consis s o a se ies o pa abolic mi o s ha e le c sola adia i on on a pipe whe e o il ge s hea ed while ci c ula ing. The p u p o se o he egula o i s o main ain he o il o u le empe a u e as nea as possible o a desi ed le el. This is a cco mpli s hed by a ying h e l ow o h e luid h o ugh h e ield. The ield exhibi s a a iable delay ime ha depends on he co n ol a iable ( low ) . The an s e un c i on o he p oc ess a ies wi h a c o s su c h a s i adian c e le el , mi o s e lec an c e and o il inle empe a u e. The sel - u ning egula o u s e s an iden i i e wi h a a ia ble o g e ing ac o and an adap i e PI c on o lle . The pape also des c ibes he heu i s i c us e d o make he e gula o wo k a n he expe iences o b ain e d a h e SSPS so la plan o Alme ia (Spain ). ~e y w o ds . Adap i e co n o l; Digi al co n ol; Sel -adjus ing sys ems. INTRODUCTION A l o o wo k s showing he ad an ages o adap i e co n olle s (As om 1983, Landau 1974 ) ha e appea ed in he Li e a u e since sel - uning egula o s we e i s in o du c ed by As om and Wi enma k ( 1 973) . M os o hese wo k s a e heo e ical and in spi e o he a c ha adap i e c o n o l has sh own ad an ages when applied o a a ie y o p oc esses ( Bo i s s on 1 97 6, Bu c hh o l 1979, K al l s om 197 9 , Na end a 198 0 . Dum o n 198 2 , Rubi o 1982, e c.) i is no ye widely a cc ep ed a indus ial le el. Thi s pape p esen s an applica i on o an adap i e egula o o c o n o l he o u le empe a u e o he ACUREX dis ibu ed co llec o ield o he SSPS plan o Tabe nas. The obje c i e o he con ol sys em in he dis ibu ed c o llec o ield is o main ain he ou le oil empe a u e a a desi ed le el in spi e o dis u bances such as changes in he sola , i adiance le el (caused by clouds o he ime o day mi o s e lec i i y and inle oil empe a u e. The dis ibu ed collec o s ields is a nonlinea sys em which can be app oxima ed by a linea sys em when conside ing small dis u bances. In he design o any egula o he ope a ing poin o he plan mus be bo n in mind. Howe e in a sola ene gy plan he ope a ing poin a ies acco ding o he ime o day o dis u bances caused by clouds, and 2 6l c on o l; Ene gy con o l; Po we he e o e i i s n o po s s ibl e o design a ixed e g ula o ull y g ua an e ed o wo k. Bec au s e o his sel - unin g egula o seem s o be a g oo d solu i o n o his p o blem. T he p la n o be con o lled i s de s c ibed in s e c io n 2 o hi s a i c le . Se c i on 3 gi es a de sc ip i on o he c o n o l alg o i hm used, se c i on 4 co mm e n s up on he esul s ob a i ne d and s e c i on S is ded ic a e d o he con c lus i o ns. A DESC RIPTI ON OF THE P LANT The s ys em o whi c h h e sel - uning co n o l e e ed o in hi s a i c le has been applied is ha o med by he dis ibu ed co llec o s ield (ACUREX ) o he so l a en e gy plan o Tabe nas. The dis ib u ed c olle c o s ield consis s mainly o a pipe line h ough which o il is l o wing and on o which he s uns a y s a e con c en a ed by means o pa ab o lic mi o s in o de o hea h e o il. I co nsis s o 480 modules a anged in wen y lines which o m en pa allel l oo ps as shown in ig. 1. The ield is also p o ided a sun seeking mechanism which causes he mi o s o e o l e a ound an axis pa allel o ha o he pipe line. On pa s sing h o ugh he ield he o il is hea ed and hen in oduced in o a s o age ank o be used o he gene a ion o elec ical ene gy . The cold inle oil o 262 F. R. Rubio, E. F. Camacho and R. Ca mona COLLECTOR FIELDS STORAGE SYSTEM POWER CONVERSION SYSTEM ACUREX FI ELD FI LD PuMPS BuFFER T AlJKS FCV I I I I I I I STEAM TuRBINE C DO:" ING TowER ""'--:::J---" I I FIG. SIMPLIFIED DCS PROCESS FLOW DIAGRAM he ield is ex ac ed om he bo om o he s o age ank. The sys em is p o ided wi h a h ee way al e which allows he oil o be ecycled in he ield un il he i s ou le empe a u e is adequa e o en e in:o :he s o age ank. A mo e de ailed desc Ip Ion o he ield can be ound in (Kal 1982). The empe a u e in he ield can be gi en by he ollowing equa ions: whe e he subindex m e e s o he me al and o he luid and: oil densi y ield capaci y ans e sal a ea ou le empe a u e i adiance op ical e iciency o e all he mal loss coe icien mi o s wid h coe icien ansmission o me al luid ex e io diame e o he pipe line inne diame e o he pipe line oil low a e These equa ions a e only applicable o he ac i e zones o he ield. Tha is, hose pa s o he pipe line whe e sola adia ions a e collec ed. Pa s o he ield, passi e zones, exis s whe e i is no possible o collec sola ene gy due o geome ical condi ions as is he case o he join s be ween he modules. These zones cons i u e a conside able pa o he ield and hey a e cha ac e ized by ha ing nil i adiance and di e en loss cons an s. The abo e equa ions we e used o simUla e he sys em in a compu e di iding one o he loops in o a 100 pieces and using a model o concen a ed pa ame e s o each piece. The model was con as ed agains he eal da a ob ained om he ield. The pa ame e o he model we e adjus ed so ha i ep oduced he beha iou o he sys em (Camacho 1977). The ou le empe a u e o he six h loop is used o con ol he ield as i ac s as he pilo loop, in his way he o al ou le empe a u e is main ained wi hin an accep able ange. THE CONTROL ALGORITHM Because o he a iabili y o cha ac e is ics as has been men ioned an adap i e con ol has been chosen. he plan s p e iously algo i hm The adap i e con ol s uc u e used co esponds o ha o a sel - uning egula o (STR) ( ig. 2), which, in b ie , consis s in calcula ing he pa ame e s o he egula o supposing ha he ields pa ame e s a e hose gi en by means o an iden i ica ion algo i hm. In his case ecu si e leas squa e a iden i ica ion algo i hm has been used. In each sampling pe iod he sel - uning egula o consis s o he ollowing s eps: 1) An es ima ion o he pa ame e s o a linea model by measu ing he inle and ou le alues o he p ocess. 2) The adjus men o he pa ame e s o he egula o . 3) The calcula ion o he con ol signal. 4) The supe ision o he co ec wo king o he con ol. Applica ion o a Sel - uning Regula o 263 The egula o is composed o wo pa s, ( ig.3), a P.I. in he eedback loop and a eed o wa d egula o calcula ed using he a ailable in o ma ion on he ield. We p esume ha he sys em can be modelled as a s able p ocess, in a ian in ime and ha can be linealized wi h jus one inpu and one ou pu , so ha i can be desc ibed by he ollowing linea di e ence equa ion: y(k) + a 1 y(k-l) + .•.. + an y(k-n) = bl u(k-d-l) + b Z u(k-d-Z) + bn u(k-d-n) + (k) and also by he ec o ial o m: y(k) = xT(k) p + (k) whe e (-y(k-l),-y(k-Z) .. -y(k-n) u(k-d-l),u(k-d-Z) .. u(k-d-n» pT = (a 1 ' aZ ' u(k) y(l<.) U(k) and U(k) Uoo Y(k) - Y e Y(k) a e he inpu and ou pu alues o he sys em a he ins an k, Uoo is he a e age alue o he inpu signal, Yoo is he alue o he e e ence and (k) is a noise independen and The z ans e be w i en as: y(z) whe e, A(z-l) 1 + a1 B(z-l) b1 The i s signal s a is ically s a iona y o ze o mean. unc ion o his sys em can z-d u(z) + -1 + aZ -2 z z -1 + bZ -2 z z quo ien -------- (z) A(z-l) + + a -n n z + + bn z-n B(z-I)/A(z-l) ep esen s hI model o he ield, and he second l/A(z-) ep esen s he model o he dis u bances. Feed o wa d Con ol. In a concen a ed ep esen a ion o he plan he in e nal ene gy a ia ion o he ield can be gi en by: C dT d N o Fig. 2 Sel - uning egula o FIElD Fig. 3 Basic con ol layou . In he pe manen egime we ha e ha : p Cp ( T - Ti ) This las equa ion can be app oxima ed as: whe e, Re e ence empe a u e. Inle empe a u e. Re lec ance. I adiance. This equa ion gi es he low as a unc ion o he e e ence empe a u e, inle empe a u e, and he i adiance. The cons an s Kl and KZ' ha e been de e mined expe imen ally, ha ing he alues o K1 = 0.93 and KZ = 155 (Ca mona, 1983). Feedback con ol Using a closed loop egula o is necessa y o compensa e o any e o which may be p oduced in he low calcula ion by he eed o wa d con olle . Gi en ha his low is a good es ima ion o he pe manen egime low, he alue o he con ol signal becomes sa u ed a + ZO% o he open loop alue, wi h his, g ea e s abili y is ob ained in he ansien s, as he egula ion band is educed. The plan unde conside a ion p esen s a long delay (be ween 2 and 8 minu es) ha a ies wi h he inle low. Fo his ype o sys em whe e he delay is much g ea e han he o de o he sys em, he con olle desc ibed below can be used (Ise mann, 1981). Supposing ha he plan is modelled using he equa ion y(k) = b u(k-d), whe e d is he delay in sampling pe iods and b he sys em gain, he minimum se ling ime con olle can be exp essed by: u (k) u(k-d) + elk) / b (1) I ins ead o di ec ly using his con olle we app oxima e a P.I. egula o so ha i has he same beha iou , we ob ain: u(k) = u(k-I) + qo elk) + qi e(k-l) whe e / ( 2 b (2 ) qo (d Z) / d 264 F. R. Rubio , E. F. Cam<lcho and R. Cannona I can be shown ha his egula o is less sensi i e o he e o in he accu acy o he delay (Ise mann, 1981), and he e o e i is mo e ad isable o use i a he han he con olle (1), gi en he cha ac e is ics o he p ocess. The e o e a P.I. has been used in he eedback loop. I should also be emembe ed ha when an e o occu s in he de e mina ion o he delay i is be e o conside i as g ea e han he es ima ed, in o de o ensu e he s abili y. The pa ame e s iden i ie is a e y impo an pa o he sel - uning con olle s. I should be ecu si e, as i mus wo k in eal ime wi h he p ocess and he e o e i will a g ea ly in luence he minimum sampling ime we can ob ain o he con ol sys em. In he same way, he s abili y o he sys em depends la gely on he es ima o con e gence. The e a e a ious ypes o iden i ie s which a e deal wi h in he ela i e li e a u e. Gene ally, he mos used me hods is he ecu si e leas squa es a e, because o i s simplici y and i s good con e gence cha a c e is ics. The algo i hm is pe o med by he ollowing s eps: 1. §elec he ini ial alues o P(k) and p(k). 2. Read he new alues o y(k+1) and u(k+1). 3. Calcula e he a p io i e o : e(k+1) = y(k+1) - XT(k+1 ) P (k ) 4. Calcula e exp ession: L(k+l) L(k+1) gi en P(k) X(k+1) by c(k) + XT (k+1) P(k) X(k+1) he 5. Calcula e he new pa ame e es ima ed gi en by: P(k+1) = P(k) + L(k+1) e(k+1) 6. Ac ualize he co a iance ma ix. P(k) P(k+1) (I - L(k+1) XT (k+1) c(k) 7. Calcula e he new o ge ing ac o c(k+1). c(k+1) 1-(1-X T (k+1) L(k+1» So I c(k+1) < c min Then c(k+1) = cmin 8. Ac ualize he measu emen s ec o X(k+2). 9. Make k = k+1 and e u n o s ep 2. Sampling an~ delay ime The sampling ime used has been se o en seconds. This ime is easonable in espec o he ime cons an o he sys em, being in he ange: T95 : Time slgnal. aken T m : Sampling ime. o each 95% o he The pa ame e (d) in he model unde conside a ion is an in ege which ep esen s he delay o he sys em. In his p ocess i is a iable wi h he low, and can be calcula ed as a unc ion o he physical dimensions o he plan and o he low, esul ing ha : 80 + 187.5 " 5.3 d ---------- - -------- + 1.5 " T m The p e i o us alues o he con ol signal a e s o ed in a ec o , in which hey shi . Because he delay is a iable he i s idea was o shi he in o ma ion wi hin his ec o a a cons an a e and o ake as he ou pu he componen o his ec o co esponding o he delay. This me h od causes p oblems in he iden i ie . To a oid his p oblem a la ge ec o in which he inpu en e in one end, and g o ou o he o he a ying hei shi ing a e ac c o ding o he delay was used. This me hod ep oduces wha eally happens in he sys em because an elemen o luid is displaced al o ng he ube a a speed whi c h a ies acco ding o he low a a gi en me men o In his way sa is ac o y beha iou o he iden i ie has been ac hie ed. Gi en he ac ha he plan is a no nlinea sys em and ha i is con inu o usly being iden i ied as a linea sys em, he pa ame e o he linea model a y wi h ime. In o de o educe he memo y o he iden i ie we use a a iable o ge ing ac o ( c ) (Fo escou 1981), as has been des c ibed p e iously. No ice ha o c 1 we ha e he leas squa es iden i ie wi h in ini e memo y and he ma ix P(k) dec eases mono o nously so he es ima o gain can ea c h ze o. As he s y s em mu s o llow he pa ame e a ia i o ns a o ge ing ac o (c < 1 ) mus be u se d. On he o he hand i he wo king poin does no change, he p oduc P ( k ) X(k ) can be ze o and he e o e P ( k+1 ) = P(k )/c. I c < 1, P ( k ) can g ow g ea ly, making he iden i ie e y sensi i e o any c hange. To a oid his a a iable o ge ing ac o is used. This ac o is made o equal 1 i he ace o P ( k ) is g ea e han a ce ain alue. A he same ime, he ace o P ( k ) is no allowed o go below a p e ixed alue by adding a ma ix R o P(k+l). Supe ision. A basic equi emen o he adap i e con ol algo i hm o be s able is ha he es ima ed pa ame e s con e ge on ce ain alues. Ce ain condi ions a e necessa y in o de o and hese dis u bances gua an ee his con e gence, depend on he ype o and whe he he inpu is pe sis en ly exci ing. A g ea imp o emen in he beha iou o he closed loop sys em can be achie ed in many cases by using a le el o supe ision which can check hings such as: - The es ima ed pa ame e s. Applica ion o a Sel - uning Regula o 265 The con ol a iable. The e olu ion o he iden i ie . In he conc e e case o plan he ollowing s eps he sola ene gy a e aken: a) Wi h ega ds o he plan iden i ica ion we know ha he sys em gain is nega i e and in e io o a ce ain alue. Because o his he iden i ied gain is il e ed be o e being conside ed alid, in o de o a oid a mis aken iden i ica ion. This can be p oduced in cases whe e he i adiance dec eases suddenly. The iden i ie wo ks using he a ia ions in he empe a u e and low wi h espec o he alues o he pe manen egime. These alues a e gi en by he e e ence empe a u e and by he low calcula ed by he eed o wa d con olle . b) The ace o he co a iance ma ix o he iden i ie is con inuously checked, aking he s eps conside ed p e iously c) The con ol a iable is sa u a ed a + 20% o he alue o he low calcula ed by he eed o wa d con olle . Fu he mo e a cau ions con ol is used in he beginning modi ying he exp ession (2) as ollows: qo = 1 / ( 2 * b + ace(P(k» * b ) RESULTS. Be o e applying he egula o o he p ocess a se ies o es s we e ca ied ou by simula ion in o de o alida e and p oo i s use ullness. These es we e aimed a analysing he beha iou o he egula o when aced wi h changes in e e ence and all ypes o dis u bances. Di e en wo king poin s ha e been conside ed and he sys em has been subjec ed o small changes o e e ence and small dis u bances as well as sudden dis u bances. A model o dis ibu ed pa ame e s has been used o simula e he p ocess, which has been alida ed and iden i ied wi h eal da a om he sola ene gy plan . Fig. 4 shows he beha iou o he egula o when aced wi h a slow a ia ion o he i adiance, keeping he empe a u e e e ence a 280 deg ees. Fig. 5 shows he sys em eac ion o e e ence changes o be ween 280 and 290 deg ees beha ing well a e he i s e e ence change, which belongs o he i s phase o adap ion. 300 QC 27oUOL-----------se--c-.------~-----------8--000 Figu e 4. I adiance changes 300 270~----------~~----------------~ o sec. 8000 Figu e 5. Se p ain changes 300 sec. 8000 Fi g u e 6 Regula o s comp. 250~----------~--~~--------------- o sec. 3900 Figu e 7. Se poin changes Cloud pe u ba ion Fig. 6 shows he beha iou o he sel uning egula o and o a ixed egula o when he i adiance change suddenly and andomly. I has been p o ed ha he mean squa e e o o he ou pu in he case o he sel uning egula o is hal ha o he co esponding one o he ixed egula o . Fig. 7 o 10 co espond o es s done in he SSPS plan o Alme la. Fig. 7 shows he ou le oil empe a u e when he e e ence is changed om 290 o 280 deg ees, a slow a ia ion in he i adiance and sha p change in his due o a small cloud passing, can be seen. Fig. 8 co esponds o he s a ing-up phase o he p ocess and egula o . The e e ence empe a u e was se a 280 deg ees. The e olu ion o he iden i ied pa ame e b can be seen in his igu e. No ice ha he ou le oil empe a u e has a small o e shoo . 266 F. R. Rubio, E. F. Camacho and R C . a mon a T. (· e) V( lI s) : : -1 -2 -3 S a ing-up p hase 10 8 ~ o Fig.-~l~O-~S~::~~~~--~~~~L-I ? e pOin changes Applicalion o a Sel -LUning RegulaLO 267 Fig. 9 shows he esponse o he plan when he oil inside he pipe line has been cool down by ge ing he mi o s ou o ocus. I can be seen ha he ou le oil empe a u e eaches he se poin (280 deg ees) sa is ac o ily. The esponse o he sys em when he se poin is changed om 280 o 290 deg ees and he i adiance is dec easing slowly can be seen in ig. 10. CONCLUSIONS An adap i e egula o o con ol he ield o collec o s in a sola ene gy plan is p esen ed in his pape . This egula o has been success ully ied ou unde a ious wo king condi ions in he ACUREX ield o he sola ene gy plan SSPS o Alme ia. In he same way o he ypes o sel - uning egula o s a e being es ed, one o which being he design o a egula o using he pole assignmen me hod. REFERENCIAS As om, K.J. and Wi enma k, P.D. (1973), On Sel - uning Regula o s, Au o~~~ica Vol 9 pp 185-199. As om, K.J. (1983), Theo y and Applica ions o Adap i e Con ol: A Su ey, .Au. om_ a i ca Vol 19-5 pp 471-486. Bo isson, U. and Syding, R. (1976), Sel - uning Con ol o an O e C ushe , Au oma ica Vol !~ pp 1-7. Buchhol , F. and Sel - uning Kummel, Con ol M. (1979), o a PH-Neu aliza ion P ocess, Au oma ica Vol !~ pp 665-671. Camacho, E.F. (1977), Iden i icacion de Sis emas no Lineales, AJus e y Es imacion de pa Ame os, ~ _ ~ _ ~ doc o al. ETSII Uni . de Se illa. Ca mona, R. y Ma in, J. (1983), Digi al DCS Con ols: The SSPS Expe ience, Fi s Te m _ ___ . Wo l<s _ l} _ o~. Tabe nas - (Aiiiie ia) - ~ - . Dumon , G.A. (1982), Sel - unins Con ol o a Chip Re ine Mo o Load, Au oma ic~ Vol 18-3 pp 307-314. Ise mann, R. (1981), Q.iLl, al Sys ems, Sp inge -Ve lag Heidelbe g New Yo k. Con ol Be lin Kalls om,C.G. e a1. (1979), Adap i e Au opi lo s o Tanke s, Au oma _ ,,-.! <e _'!_ Vol 15 pp 241-254. Kal , A. e al. Collec o Sys em Repo , IEA-SSPS (1982), D ...! . l ib~ .! ,~c1 Plan Cons uc ion ~~e~a ing Agen iFVLR, Cologne, FRG. Landau, 1.D. (1974), A Su ey o Model Re e ence Adap i e Techniques - Theo y and Applica ions, Au oma ica Vol lQ pp 353-379. Na end a, K.S. and Monopoli, R.V. (1980), Applica ions o Adal" i _e Con~ ol, Academic i -ess. Rubio, F.R. e al. (1982), S udies o he Aplica ion and Adap i e Con olle o Hid o u bine Gene a o s, P oc. o he IFAC So wa e o Compu e Con ol - (5OCOCO), Mad id. - .