SMART ADAPTIVE CONTROL OF A SOLAR THERMAL POWER PLANT …

SMART ADAPTIVE CONTROL OF A SOLAR THERMAL POWERPLANT IN VARYING OPERATING CONDITIONS

Esko K. Juuso∗Control Engineering Group, Faculty of Technology, FI-90014 University of Oulu, Finland

Luis J. YebraCIEMAT, Plataforma Solar de Almería, Automatic Control Group

Crta. de Senés s/n, 04200 Tabernas, Almería, Spain

ABSTRACT

Solar thermal power plants collect available solar energy in a usable form at a temperature rangewhich is adapted to the irradiation levels and seasonal variations. Solar energy can be collected onlywhen the irradiation is high enough to produce the required temperatures. During the operation, atrade-off of the temperature and the flow is needed to achieve a good level for the collected power.The storage is needed to keep the heat supply during the nights and heavy cloudy periods. Efficientoperation requires a fast start-up and reliable operation in cloudy conditions without unnecessaryshutdowns and start-ups. Fast and well damped linguistic equation (LE) controllers have been testedin Spain at a collector field, which uses parabolic-trough collectors to supply thermal energy inform of hot oil to an electricity generation system or a multi-effect desalination plant. Control isachieved by means of varying the flow pumped through the pipes in the field during the operation.The LE controllers extend the operation to varying cloudy conditions in a smart way. Varyingirradiation and energy demand during the daytime can be smoothly handled in the LE control systemwith predefined model-based adaptation techniques. The system activates special features whenneeded to facilitate smooth operation. The intelligent state indicators react well to the changingoperating conditions and can be used in smart working point control to further improve the operationin connection with the other energy sources.Keywords: Solar energy, intelligent control, nonlinear systems, adaptation, optimisation, linguisticequations, modelling, simulation

INTRODUCTIONSolar power plants should collect any available ther-mal energy in a usable form at the desired temper-ature range, which improves the overall system ef-ficiency and reduces the demands placed on auxil-iary equipment. In addition to seasonal and dailycyclic variations, the intensity depends also on at-mospheric conditions such as cloud cover, humid-ity, and air transparency. A fast start-up and effi-cient operation in varying cloudy conditions is im-portant. A solar collector field is a good test platformfor control methodologies [1, 2, 3]. Model-basedapproaches are useful since the operation dependsstrongly on weather conditions: the model structure

∗Corresponding author: Phone: +358 294 482463 E-mail:[email protected]

is fairly clear but the properties of the oil depend onthe temperature [4]. Lumped parameter models tak-ing into account the sun position, the field geometry,the mirror reflectivity, the solar irradiation and theinlet oil temperature have been developed for a so-lar collector field [1]. A feedforward controller hasbeen combined with different feedback controllers[5]. Local linearization has been used in feedbackcontrol [6]. Feedforward approaches based directlyon the energy balance can use the measurements ofsolar irradiation and inlet temperature [7]. Energycollection is focused in [8]. Model-based predictivecontrol [9, 10] is suitable for fairly smoothly chang-ing conditions. The classical internal model control(IMC) can operate efficiently in varying time delayconditions [11]. Genetic algorithms have also been

Proceedings from The 55th Conference on Simulation and Modelling (SIMS 55), 21-22 October, 2014. Aalborg, Denmark

111

used for multiobjective tuning [12].Linguistic equation (LE) controllers use model-based adaptation and feedforward features, whichare aimed for preventing overheating, and the con-troller presented in [13] already took care of theactual setpoints of the temperature. The man-ual adjustment of the working point limit has im-proved the operation considerably. Parameters ofthe LE controllers were first defined manually, andlater tuned with neural networks and genetic algo-rithms. Genetic algorithms combined with simula-tion and model-based predictive control have furtherreduced temperature differences between collectorloops [14]. Data analysis methods are based ongeneralised norms [15] and extended to a recursiveversion of the scaling approach was introduced in[16]. New state indicators for detecting cloudy con-ditions and other oscillatory situations by analysingfluctuations of irradiation, temperature and oil flow[17]. The new indicators react well to the chang-ing operating conditions and can be used in smartworking point control. Recent developments includeadvanced model-based LE control are discussed in[18] and intelligent analysers [19].This paper summarises and analyses the overall op-eration of the LE controller in changing operatingconditions. The analysis is based on experimentscarried out in the Acurex Solar Collectors Field ofthe Plataforma Solar de Almeria (PSA) in Spain.

SOLAR COLLECTOR FIELDThe aim of solar thermal power plants is to pro-vide thermal energy for use in an industrial processsuch as seawater desalination or electricity genera-tion. Unnecessary shutdowns and start-ups of thecollector field are both wasteful and time consum-ing. With fast and well damped controllers, the plantcan be operated close to the design limits therebyimproving the productivity of the plant [4].The Acurex field supplies thermal energy (1 MWt) inform of hot oil to an electricity generation system ora multi–effect desalination plant. Parabolic–troughcollectors are used for focusing the irradiation on thepipes and control is achieved by means of varyingthe flow pumped through the pipes (Fig. 1) duringthe operation. In addition to this, the collector fieldstatus must be monitored to prevent potentially haz-ardous situations, e.g. oil temperatures greater than300 oC. The temperature increase in the field may

Figure 1: Layout of the Acurex solar collector field.

rise up to 110 degrees. At the beginning of the dailyoperation, the oil is circulated in the field, and theflow is turned to the storage system (Fig. 1) whenan appropriate outlet temperature is achieved. Thevalves are used only for open-close operation: theoverall flow F to the collector field is controlled bythe pump. [20] The latest test campaigns focusedon achieving a smooth operation in changing oper-ating conditions to avoid unnecessary stress on theprocess equipment.

SMART ADAPTIVE CONTROLThe multilevel control system consists of a nonlin-ear PI-type LE controller with predefined adaptationmodels, some smart features for avoiding difficultoperating conditions and a cascade controller for ob-taining smooth operation (Fig. 2). Controller per-formance is used in the adaptation module. For thesolar collector field, the goal is to reach the nom-inal operating temperature 180− 295 oC and keepit in changing operating conditions [16, 17]. Thefeedback controller is a PI-type LE controller withone manipulating variable, oil flow, and one con-trolled variable, the maximum outlet temperature ofthe loops. The controller provides a compact ba-sis for advanced extensions. High-level control isaimed for manual activating, weighting and closingdifferent actions.

Feedback LE controller

Feedback linguistic equation (LE) controllers use er-ror e j(k) and derivative of the error ∆e j(k). Thesereal values are mapped to the linguistic range [−2,2]


112

Performance

analysis

Adaptation

- adaptation scaling

- model-based

- smart actions

Intelligent analysers

- working point

- fluctuations

- large initial errors

- asymmetry

- fast changes

High-level control

- weighting of strategies, switching

- diagnostics

- cascade control

PI-type

LE Controller

Collector

field

Figure 2: LE controller of the Acurex solar collectorfield.

by nonlinear scaling with variable specific member-ship definitions ( fe) and f∆e), respectively. As allthese functions consist of two second order polyno-mials, the corresponding inverse functions consistof square root functions. The linguistic values ofthe inputs, e j(k) and ˜∆e j(k), are limited to the range[−2,2] by using the functions only in the operatingrange: outside the scaled values are -2 and 2 for lowand high values, respectively.A PI-type LE controller is represented by

˜∆ui j(k) = KP(i, j) ˜∆e j(k)+KI(i, j) e j(k), (1)

which contains coefficients KP(i, j) and KI(i, j). Theeffects of e j(k) and ˜∆e j(k) can be tuned by member-ship definitions ( fe) j and ( f∆e) j, respectively. How-ever, the direction of the control action is fixed in(1). Different directions and strengths can be han-dled with this controller.The output i of a single input single output (SISO)controller is calculated by adding the effect of thecontrolled variable j to the manipulated variable i:

ui(k) = ui(k−1)+∆ui j(k). (2)

In the PI-type LE controller, the error variable isthe deviation of the outlet temperature from the setpoint, and the control variable is oil flow.

Intelligent analysers

Intelligent analyzers are used for detecting changesin operating conditions to activate adaptation and

model-based control and to provide indirect mea-surements for the high-level control. The data anal-ysis is based on the generalised norms

||τMpj ||p = (τMp

j )1/p = [

1N

N

∑i=1

(x j)pi ]

1/p, (3)

where p 6= 0, is calculated from N values of a sam-ple, τ is the sample time. These norms can be calcu-lated recursively [16].

Working point On a clear day, the field couldbe operated by selecting a suitable a balance be-tween the irradiation and the requested temperatureincrease. The working point is represented by

wp = Ie f f − Tdi f f , (4)

where Ie f f and Tdi f f are obtained by the nonlinearscaling of variables: efficient irradiation Ie f f andtemperature difference between the inlet and out-let, Tdi f f = Tout −Tin. The outlet temperature Tout isthe maximum outlet temperature of the loops. Thismodel handles the nonlinear effects: the volumetricheat capacity increases very fast in the start-up stagebut later remains almost constant because the normaloperating temperature range is fairly narrow.The working point variables already define the over-all normal behaviour of the solar collector field,wp = 0, where the irradiation Ie f f and the temper-ature difference, Tdi f f , are on the same level. A highworking point (wp > 0) means low Tdi f f comparedwith the irradiation level Ie f f . Correspondingly, alow working point (wp < 0) means high Tdi f f com-pared to the irradiation level Ie f f . The normal limit(wpmin = 0) reduces oscillations by using slightlylower setpoints during heavy cloudy periods. Higherlimits, e.g. (wpmin = 1), shorten the oscillation peri-ods after clouds more efficiently.

Fluctuations The fluctuation indicators, whichwere introduced to detecting cloudiness and oscil-lations, are important improvements aimed for prac-tical use. The cloudy conditions are detected by cal-culating the difference of the high and the low val-ues of the corrected irradiation as a difference of twomoving generalised norms:

∆xFj (k) = ||KsτMph

j ||ph−||KsτMpl

j ||pl , (5)

where the orders ph ∈ ℜ and pl ∈ ℜ are large pos-itive and negative, respectively. The moments are


113

calculated from the latest Ks + 1 values, and an av-erage of several latest values of ∆xF

j (k) is used as anindicator of fluctuations. [17]

Large initial errors Large steps and irradiationdisturbances may cause very large initial errorswhich are detected by the predictive braking indi-cation. The calculated braking coefficient, bc j(k) isused to emphasise the influence of the derivative ofthe error by means of the following equation:

KP(i, j) = (1+bc j(k)) KP(i, j) (6)

A new solution has been introduced to detecting thelarge error.

Asymmetry The control surface should be asym-metric when the irradiation is strongly increasing ordecreasing. The asymmetry detection is based on thechanges of the corrected irradiation. The action isactivated only close to the set point if there are nostrong fluctuations of the controlled variable evalu-ated by e j

− and e j+. The previous calculation based

on the solar noon does not take into account actualirradiation changes.

Fast changes Fast changes may occur in the tem-peratures (inlet, outlet and difference). Changes ofthe inlet temperature have a strong effect on the out-let temperature. The change is detected with the in-dex

∆T Hin (k) = Tin(k)−

1nL +1

k

∑i=k−nL

Tin(i), (7)

where the average value is calculated from nL + 1.A smooth increase is normal during the daily opera-tion and strong effects are detected after load distur-bances.Fast changes of the outlet temperatures were consid-erably reduced when the fluctuation indicators weretaken into use to adapt the setpoint. Too fast out-let temperature increase represented by the valuerange ∆T R

out(k) obtained as the difference of max-imum and minimum values of Tout(i) in the rangei = k− nL, . . . ,k. This index is activated if Tout hasincreased during the period. An indicator of too hightemperature differences is based on the detection ofan overshoot

∆T Hout(k) = max{0,Tout(k)−T sp

out}. (8)

Adaptive control

Adaptive LE control extends the operating area ofthe LE controller by using correction factors ob-tained from the working point (4) to reduce oscil-lations, when wp is low, and to speedup operation,when wp is high. This predefined adaptation ishighly important since there are not time enough toadapt online when there are strong disturbances.The predictive braking and asymmetrical actions areactivated when needed. Intelligent indicators intro-duce additional changes of control if needed. Thetest campaigns have clarified the events, which acti-vate the special actions. Each action has a clear taskin the overall control system.The additional intelligent features ∆T H

in (k), T Rout(k)

and ∆T Hout(k), which detect anomalies, introduce an

additional change of control:

˜∆uCH

j (k) = c1˜

∆T Hin (k)+c2

˜T Rout(k)+c3

˜∆T H

out(k), (9)

where the coefficients c1, c2 and c3 are chosen fromthe range [0, 1]. The first two actions are predic-tive, and the third one is corrective. If Tdi f f is toohigh, also the set point is corrected correspondinglyto avoid low working point wpi(k)� 0.

Model-based control

Model-based control was earlier used for limitingthe acceptable range of the temperature setpoint bysetting a lower limit of the working point (4). Thefluctuation indicators are used for modifying thelower working point limit to react better to cloudi-ness and other disturbances. This overrides the man-ual limits if the operation conditions require that.The model-based extension is an essential part inmoving towards reliable operation in cloudy condi-tions: the control system can operate without man-ual interventions. The high-level control moves to-wards control strategies to modify intelligent analy-sers and adaptation procedures (Fig. 2).

RESULTSThe control system was tested on a solar collectorfield at PSA in July 2012 to find a suitable solutionfor sustainable almost unattended operation. The re-sults are used in developing optimisation solutionsfor the energy collection.


114

Normal operation

On clear days with high or fairly high irradiation,the setpoint tracking is acceptable: step changesfrom 15-25 degrees are achieved in 20-30 minuteswith minimal oscillation. The working point adap-tation operates efficiently and the temperature canbe increased and decreased in spite of the irradiationchanges (Fig. 3).High setpoints can be used since the working pointlimit activates the setpoint correction when the tem-perature difference exceeds the limit correspondingto the irradiation level in the active working pointlevel. There are three time periods of reduced set-points in Figure 3 from 11:40 to 12:18, from 12:50to 13:30 and from 14:48 to 15:40.The oil flow changes smoothly: the fast changes areat the beginning of the step (Fig. 4). Working pointcorrections and limiting the fast changes are negli-gible. The predictive braking was activated in thesesituations, but the asymmerical action was not yetavailable. This causes positive offset, when the ir-radiation is increasing before the solar noon. Cor-respondingly, the offset is negative after the solarnoon.

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16

180

200

220

240

Time (h)

Tem

pera

ture

( o C

)

Tout

Tref

Tin

Figure 3: Test results of the LE controller on a fairlyclear day [19].

Cloudy conditions

The setpoint correction operates throughout thecloudy periods to reduce oscillations and hazardoussituations caused by the abrupt changes of irradia-tion (Fig. 5(b)).Three cloudy periods occurred on the third day: along period in the morning, a short light one closethe solar noon and a short, but heavy, in the af-ternoon. The temporary setpoint correction oper-ated well in these situations (Fig. 5(a)). During

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16

−20

0

20

x(1)

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16

−20

0

20

x(2)

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16−2

0

2

xlin

g(1)

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16−2

0

2

xlin

g(2)

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16−2

0

2

ylin

g(1)

11 11.5 12 12.5 13 13.5 14 14.5 15 15.5 16

−20

0

20

y(1)

Figure 4: Operation of the LE controller on a fairlyclear day: x = [e, ∆e], xling = [e, ∆e], yling = ∆uand y = ∆u.

the cloudy periods (Fig. 5(b)), the temperature wentdown with 20 degrees but rose back during the shortsunny spells, and finally, after the irradiation dis-turbances, high temperatures were achieved almostwithout oscillations with the gradually changing set-point defined by the working point limit although theinlet temperature was simultaneously rising. Afterthese periods, the field reached the normal operationin half an hour.The controller used high oil flow levels when the skywas clearing up. Also the change of control was re-acting strongly (Fig. 6). The same approach op-erated well for the other two cloudy periods. Theoil flow was changed smoothly also during theseperiods. The working point corrections were nowvery strong, but limiting the fast changes was hardlyneeded. Strong braking was used in the beginningand in the recovery from the first cloudy period.There were problems with some loops during thatday.


115

10 11 12 13 14 15 16

150

200

250

Time (h)

Tem

pera

ture

( o C

)

Tout

Tref

Tin

(a) Outlet temperature.

10 11 12 13 14 15 160

200

400

600

800

Time (h)

Irra

diat

ion

( W

/m2 )

HighMeanLow

(b) Irradiation.

Figure 5: Test results of the LE controller on acloudy day [19].

The fourth day had two very different periods: thestart was very bright and the irradiation was risingsmoothly, but everything was changed just beforethe solar noon, and the heavy cloudy period contin-ued the whole afternoon. The field was in tempera-tures 160 - 210 oC for more than two hours althoughthe loops were not tracking the sun all the time. Theworking point corrections were during this periodvery strong, but limiting the fast changes was hardlyneeded [18].

Load disturbances

During a day, the temperature increases more or lesssmoothly in the storage tank (Fig. 1), but the useenergy may cause fast disturbances. The controllershould also hande these disturbances. On the fifthday, the start-up followed the setpoint defined by theworking point limit. In addition, there was an unin-tentional drop of 16.9 degrees in the inlet tempera-ture. The disturbance lasted 20 minutes. The con-troller reacted by introducing a setpoint decrease of19.8 degrees. The normal operation was retained in50 minutes with only an overshoot of two degrees,but with some oscillations.This kind of disturbance was repeated on the sixth

10 11 12 13 14 15 16−50

0

50

x(1)

10 11 12 13 14 15 16

−20

0

20

x(2)

10 11 12 13 14 15 16−2

0

2

xlin

g(1)

10 11 12 13 14 15 16−5

0

5

xlin

g(2)

10 11 12 13 14 15 16−2

0

2

ylin

g(1)

10 11 12 13 14 15 16

−20

0

20

y(1)

Figure 6: Operation of the LE controller on a cloudyday: x = [e, ∆e], xling = [e, ∆e], yling = ∆u and y =∆u.

day (Fig. 7(a)): the inlet temperature was droppedmaximum 13.5 degrees during a period of 15 min-utes. Now the setpoint was changed when the in-let temperature reached the minimum. The workingpoint limit was changed to allow a higher setpointin the recovery. The change of control is operatingin a similar way as in the setpoint change (Fig. 8).The temperature drop was smaller (7.5 degrees) butthe overshoot slightly higher (2.5 degrees). Also therecovery took less time (30 minutes).

Asymmetrical correction

The new asymmetrical correction was activated inseveral periods on the sixth day. There were goodresults on two previous days, but now the operationwas better tuned for the afternoon as well. The set-points were achieved in the range± 0.5 degrees withhardly any offset (Fig. 7(a)). The change is con-siderable to the first days, when the outlet tempera-


116

10 11 12 13 14 15 16

180

200

220

240

260

Time (h)

Tem

pera

ture

( o C

)

Tout

Tref

Tin

(a) Outlet temperature.

10 11 12 13 14 15 160

2

4

6

8

Time (h)

Oil

flow

( l/

s )

F

∆ F

(b) Oil flow.

Figure 7: Test results of the LE controller on a clearday: asymmetral action [19, ].

ture exceeded the setpoint with 0.5-1 degrees, whenthe irradiation was increasing, and remained about1.0 degrees lower when the irradiation decreased.Around the solar noon, the setpoint was achievedvery accurately even for high temperatures corre-sponding negative working points. The increase ofthe inlet temperature is smoothly compensated withsmall changes of the oil flow and the setpoint is alsoaccurately achieved after the load disturbance (Fig.7(b)). In this case, the asymmetrical action increasesthe positive changes before the solar noon and thenegative changes after that (Fig. 8).

Optimisation

The temperature increase in the collector field natu-rally depends on the irradiation, which is the high-est close to the solar noon. As the inlet tempera-ture often increases slightly during the day, there isa possibility to use even higher outlet temperatures.The temperatures increase with decreasing oil flow,which can be controlled smoothly in a wide range.A trade-off of the temperature and the flow is neededto achieve a good level for the collected power. Theworking point is chosen from the high power range

10 11 12 13 14 15 16−50

0

50

x(1)

10 11 12 13 14 15 16−50

0

50

x(2)

10 11 12 13 14 15 16−2

0

2

xlin

g(1)

10 11 12 13 14 15 16−2

0

2

xlin

g(2)

10 11 12 13 14 15 16−2

0

2

ylin

g(1)

10 11 12 13 14 15 16

−20

0

20

y(1)

Figure 8: Operation of the LE controller on aclear day, including asymmetry action: x = [e, ∆e],xling = [e, ∆e], yling = ∆u and y = ∆u.

and used in the model-based control to choose orlimit the setpoint. The power surface is highly non-linear because of the properties of the oil. Distur-bances of the inlet temperatures introduce fluctua-tion to the outlet temperature (Fig. 7(a)). The ac-ceptable working point is limited by the oscillationrisks and high viscosity of the oil during the start-up.In the latest tests, the inlet temperatures are high al-ready in the start-up, since the oil flow was not firstcirculated in the field. During high irradiation peri-ods, high outlet temperatures are avoided by keepingthe working point high enough.

CONCLUSIONThe intelligent LE control system, which is basedon intelligent analysers and predefined model-basedadaptation techniques, activates special featureswhen needed. Fast start-up, smooth operation andefficient energy collection is achieved even in vary-


117

ing operating condition. The state indicators reactwell to the changing operating conditions and canbe used in smart working point control to furtherimprove the operation. The controller reacts effi-ciently on the setpoint changes, clouds and load dis-turbances. The setpoint is achieved accurately withthe new asymmetrical action. The working point canbe chosen in a way which improves the efficiency ofthe energy collection. A trade-off of the temperatureand the flow is needed to achieve a good level for thecollected power.

ACKNOWLEDGEMENTSExperiments were carried out within the project "In-telligent control and optimisation of solar collectionwith linguistic equations (ICOSLE)" as a part of theproject "Solar Facilities for the European ResearchArea (SFERA)" supported by the 7th FrameworkProgramme of the EU (SFERA Grant Agreement228296).

REFERENCES[1] E. Camacho, M. Berenguel, F. R. Rubio, Adap-

tive Control of Solar Plants, Springer, London,1997.

[2] E. K. Juuso, Fuzzy control in process indus-try: The linguistic equation approach, in: H. B.Verbruggen, H.-J. Zimmermann, R. Babuška(Eds.), Fuzzy Algorithms for Control, Inter-national Series in Intelligent Technologies,Vol. 14 of International Series in IntelligentTechnologies, Kluwer, Boston, 1999, pp. 243–300, doi: 10.1007/978-94-011-4405-6_10.

[3] E. F. Camacho, M. Berenguel, F. R. Rubio,D. Martinez, Control of Solar Energy Systems,Advances in Industrial Control, Springer, Lon-don, 2012.

[4] E. K. Juuso, P. Balsa, L. Valenzuela, Multi-level linguistic equation controller applied toa 1 MW solar power plant, in: Proceedingsof the ACC’98, Philadelphia, PA, June 24–26,1998, Vol. 6, ACC, 1998, pp. 3891–3895.

[5] L. Valenzuela, P. Balsa, Series and parallelfeedforward control schemes to regulate the

operation of a solar collector field, in: Pro-ceedings of the 2nd User Workshop Train-ing and Mobility of Researchers Programmeat Plataforma Solar de Almeria, November 10-11, 1998, Ciemat, 1998, pp. 19–24.

[6] C. M. Cirre, M. Berenguel, L. Valenzuela,E. Camacho, Feedback linearization controlfor a distributed solar collector field, Controlengineering Practice 15 (2007) 1533–1544,doi: 10.1016/j.conengprac.2007.03.002.

[7] E. Camacho, F. Rubio, F. Hughes, Self-tuningcontrol of a solar power plant with a distributedcollector field, IEEE System Control Magazine12 (2) (1992) 72–78.

[8] T. A. Johansen, C. Storaa, Energy-based con-trol of a distributed solar collector field, Auto-matica 38 (2002) 1191–1199.

[9] D. Limon, I. Alvarado, T. Alamo, M. Ruiz,E. F. Camacho, Robust control of the dis-tributed solar collector field ACUREX usingMPC for tracking, in: Proceedings of 17thIFAC World Congress, Seoul, Korea, July 6-11, 2008, Vol. 17, IFAC, 2008, pp. 958–963,http://www.ifac-papersonline.net/.

[10] C. O. Ayala, L. Roca, J. L. Guzman, J. E.Normey-Rico, M. Berenguel, L. Yebra, Lo-cal model predictive controller in a so-lar desalination plant collector field, Re-newable Energy 36 (2011) 3001–3012, doi:10.1016/j.renene.2011.03.037.

[11] I. Farkas, I. Vajk, Internal model-based con-troller for a solar plant, in: L. Basanez, J. A.de la Puente (Eds.), Proceedings of the 15thTriennial World Congress, Barcelona, Spain,July 21-26, 2011, IFAC, 2002, pp. 49–54,http://www.ifac-papersonline.net/.

[12] J. Bonilla, , L. J. Yebra, S. Dormido,E. Zarza, Parabolic-trough solar thermal powerplant simulation scheme, multi-objective ge-netic algorithm calibration and validation,Solar Energy 86 (2012) 531–540, doi:10.1016/j.solener.2011.10.025.

[13] E. K. Juuso, L. Valenzuela, Adaptive intelli-gent control of a solar collector field, in: Pro-


118

ceedings of Eunite 2003 - European Sympo-sium on Intelligent Technologies, Hybrid Sys-tems and their implementation on Smart Adap-tive Systems, July 10-11, 2003, Oulu, Finland,Wissenschaftsverlag Mainz, Aachen, 2003, pp.26–35.

[14] E. K. Juuso, Modelling and simulation in de-velopment and tuning of intelligent controllers,in: I. Troch, F. Breitenecker (Eds.), Proceed-ings of 5th Vienna Symposium on Mathemat-ical Modelling, February 8-10, 2006, Vienna,Austria, Argesim Report 30, Argesin Verlag,Vienna, 2006, pp. 6–1 – 10, iSBN 3-901608-30-3.

[15] E. Juuso, S. Lahdelma, Intelligent scaling offeatures in fault diagnosis, in: Proceedings ofthe 7th International Conference on ConditionMonitoring and Machinery Failure PreventionTechnologies, 22-24 June, 2010, Stratford-upon-Avon, UK, Vol. 2, Curran Associates,NY, USA, 2010, pp. 1358–1372, ISBN 978-1-61839-013-4.

[16] E. K. Juuso, Recursive tuning of intelli-gent controllers of solar collector fields inchanging operating conditions, in: S. Bit-tani, A. Cenedese, S. Zampieri (Eds.), Pro-ceedings of the 18th World Congress TheInternational Federation of Automatic Con-trol, Milano (Italy) August 28 - September2, 2011, IFAC, 2011, pp. 12282–12288, doi:10.3182/20110828-6-IT-1002.03621.

[17] E. K. Juuso, Model-based adaptation of intel-ligent controllers of solar collector fields, in:I. Troch, F. Breitenecker (Eds.), Proceedings of7th Vienna Symposium on Mathematical Mod-elling, February 14-17, 2012, Vienna, Austria,Part 1, Vol. 7, IFAC, 2012, pp. 979–984, doi:10.3182/20120215-3-AT-3016.00173.

[18] E. K. Juuso, L. Yebra, Model-based intelli-gent control of a solar energy collector field,in: Proceedings of 8h EUROSIM Congress onModelling and Simulation, 10-13 September,2013, Cardiff, UK, IEEE, 2013, pp. 513–518,doi:10.1109/EUROSIM.2013.92.

[19] E. K. Juuso, L. Yebra, Optimisation of so-lar energy collection with smart adaptive con-

trol, in: USB Proceedings of IECON 2013 -39th Annual Conference of the IEEE IndustrialElectronics Society, 10-14 November, 2013,Vienna, Austria, IEEE, 2013, pp. 7930–7935,doi: 10.1109/IECON.2013.6700459.

[20] E. K. Juuso, P. Balsa, K. Leiviskä, LinguisticEquation Controller Applied to a Solar Col-lectors Field, in: Proceedings of the EuropeanControl Conference -ECC’97, Brussels, July 1- 4, 1997, Vol. Volume 5, TH-E I4, paper 267(CD-ROM), 1997, p. 6 pp.


119

SMART ADAPTIVE CONTROL OF A SOLAR THERMAL POWER PLANT …

Documents

Transcript of SMART ADAPTIVE CONTROL OF A SOLAR THERMAL POWER PLANT …