-
April 2004 15
F E A T U R EF E A T U R E
Electricity generation from solar energy is cur-rently one of
the main research areas in the fieldof renewable energy. Such
systems require reli-able control systems to maintain desired
operat-ing conditions in the face of changes in solarradiation.
Parabolic trough collectors, which are
the most developed line-focus concentrating solar collec-tors,
are used to feed industrial heat processes thermally.At present,
parabolic reflectors can operate at tempera-tures up to around 400
C by concentrating the directsolar radiation onto a tube through
which a fluid ispumped and heated (see Figure 1). Optical
concentrationreduces the absorber surface area relative to the
collectoraperture area and thus significantly reduces thermal
loss-es. This optical concentration requires the collector torotate
about a tracking axis, following the daily movementof the sun.
Since only direct solar radiation is optically
concentrated, diffuse solar radiation is lost. One advantageof
parabolic trough collectors is the low pressure drop ofthese
systems as the working fluid passes through a single,straight
absorber tube.
This article describes the DISS facility, followed by
anexplanation of the main control problem. The controlscheme for
one of the three operating modes, the once-through mode, is then
discussed, including system models,feedforward blocks, and PI
blocks. Finally, experimentalresults and conclusions are
presented.
Parabolic TroughsWithin the range of 200400 C, present-day
parabolictrough technology uses oil as a working fluid in
theabsorber tubes, whereas a mixture of water and ethyleneglycol
can be used for lower temperatures. The workingfluid is heated as
it passes through the absorber tube of the
0272-1708/04/$20.002004IEEEIEEE Control Systems Magazine
By Loreto Valenzuela, Eduardo Zarza, Manuel Berenguel, and
Eduardo F. Camacho
EYE
WIR
E
-
April 200416 IEEE Control Systems Magazine
solar collectors, thus converting the direct solar radiationinto
thermal energy. The hot working fluid is then sent to aheat
exchanger, where its thermal energy is transferred.
For the last 15 years a considerable effort has beenmade to
develop efficient control systems for solar ther-mal power plants
with parabolic trough collector fields[1]. Currently, the best
commercial examples of state-of-the-art parabolic trough collectors
are the eight solar ther-mal power plants operating in California.
These plants,called SEGSs, use oil as the heat transfer medium
betweenthe solar field and the power block. The outlet of the
solarfield is connected to a heat exchanger that generatessteam to
feed a Rankine cycle [2].
Solar radiation, the primary energy source in theseplants, is
not controllable. The temperature of the oil atthe outlet of the
solar field is controlled by modulatingthe oil mass flow at the
field inlet. These systems pos-sess nonlinear dynamics and are
affected by severaltypes of disturbances, mainly in the energy
input vari-able, which suggests the need for advanced
controlstrategies [1].
A parabolic-trough collector system prototype wasimplemented in
19961998 at the PSA in southeast Spain toinvestigate using water as
the working fluid in the solarfield of a thermal power plant using
a DSG process [3].From 1999 to 2001, different operating strategies
and con-figurations were evaluated, taking efficiency, cost,
andcontrollability into consideration, and promising resultswere
obtained for the commercial implementation of thisnew system, which
at present constitutes the mostadvanced plant of this type.
The main industrial application of DSG is expected tobe
Rankine-cycle electricity generation, in which steam isdelivered by
the parabolic-trough collectors. The DSGprocess increases overall
system efficiency while reducinginvestment costs, since it
eliminates the oil used at theSEGS plants as a heat transfer medium
between the solarfield and the power block, and, consequently, the
heatexchanger is also eliminated. The electricity generationcost
will be reduced by 26%, according to current avail-able data.
Furthermore, a DSG solar field can be used tofeed other industrial
processes requiring thermal energyin the form of saturated or
superheated steam at tempera-tures below 400 C and pressures below
100 bar. Desalina-tion is a good example of an industrial process
suitable forbeing fed by a DSG solar field [4].
The main task of the control system is to provide asteady supply
of live steam at the outlet of the solar fieldunder all operating
conditions. In the first stage of the DISSproject, financed by
CEE-Joule contract JOR3-CT98-0277,simple control structures and
pragmatic approaches for thecontroller design were chosen [3], [5],
[6]. The advantage ofthis approach is that the plant operators are
already famil-iar with the PI or PID controllers used and are able
to modi-fy controller parameters to achieve secure operation,
thusallowing them to concentrate on the process itself. Afterthis
training stage devoted to identifying plant dynamics
Acronyms
DISS Direct solar steam (the R + TD project)DSG Direct steam
generation (the process)FFFV Feedforward function feed valve FFIV
Feedforward function injector valve MISO Multiple inputs, single
output PI Proportional integralPID Proportional integral
derivativePSA Plataforma Solar de Almera (Spain)SEGS Solar
electricity generating systemSISO Single input, single output
Figure 1. A parabolic trough collector: working principleand
components. Solar radiation collected by a parabolicshape reflector
is concentrated on an absorber pipe locatedin the focal line of the
parabola and through which a heattransfer fluid is pumped.
Steel Structure Parabolic TroughAbsorber
Figure 2. The DISS plant. The collectors are in the
trackingposition following the sun when the plant is running.
-
and operating modes, advanced control strategies,
mainlymodel-based predictive control schemes, will be investigat-ed
in an effort to optimize plant production.
The PSA DISS FacilityThe PSA DISS facility is a solar system
that serves as a test-bed for investigating the DSG process in
parabolic troughsolar collectors. Figures 2 and 3 show two views of
thefacility; its main characteristics are listed in Table 1.
Although the solar field can operate over a wide rangeof
temperatures and pressures, the three main operatingpoints
investigated in the DISS project are listed in Table 2.The
thermohydraulic behavior and system performance ofthree basic
operating modes, namely, once-through, recir-culation, and
injection modes (see Figure 4), were investi-gated under actual
conditions to identify the specificadvantages and disadvantages of
each mode.
In the once-through mode, feedwater is preheated,evaporated, and
converted into superheated steam as itcirculates from the inlet to
the outlet of the collector loop.The main disadvantage of this
concept, which is the sim-plest of the three, is the
controllability of the superheatedsteam parameters at the collector
field outlet. A waterinjector is placed in front of the last
collector to controlthe outlet steam temperature. The water
injector is placedhere since, by design, superheated steam is
available atthis point. The injector is not placed at the end of
the col-lector row, which would be advantageous for control,because
thermal losses in the last collector would signifi-cantly increase,
thus reducing the energy gain. Moreover,the selective coating of
the absorber pipes would bedegraded if the metal piping reached
temperatures ofaround 450 C, which is possible when the system is
work-ing in operating mode 3 (see Table 2).
April 2004 17IEEE Control Systems Magazine
Nomenclature
Acol Collector aperture [m]E Solar irradiance [W/m2]K
Proportional gain (PID)Lcol Collector length [m]Lloop Collector
loop length [m]P Pressure [bar]T Temperature [C]Tamb Ambient
temperature [C]Tav Average fluid temperature in the collector
loop [C]Ti Integral time constant (PID) [s]Tin Water temperature
at collector loop inlet [C]Tin c Water/steam temperature at
collector inlet [C]Tinj Injection water temperature [C]Tout Steam
temperature at collector row outlet
outlet [C]Tref Steam temperature reference [C]Sabs Absorber pipe
area of the collector loop [m2]afv Feed valve aperture demanded
[%]aiv Injector valve aperture demanded [%]apv Steam pressure valve
aperture [%]em in Inlet mass flow error [kg/s]eT Temperature error
[C]hin Specific enthalpy of water at collector row
inlet [kJ/kg]hin c Specific enthalpy of water at collector
inlet [kJ/kg]hinj Specific enthalpy of injection water at
collector inlet [kJ/kg]hout Specific enthalpy of steam at
collector
row outlet [kJ/kg]href Specific enthalpy reference of steam
at
collector row outlet [kJ/kg]min Mass water flow at collector row
inlet [kg/s]min c Mass water/steam flow at collector inlet
[kg/s]minj dem Injection water flow demanded at collector
inlet [kg/s]minj set Injection water flow reference at
collector
inlet [kg/s]minj Injection water flow at collector inlet [kg/s]w
Feed pump power [%] Incrementpfv Pressure drop across feed valve
[bar]loop Global collector loop efficiency []col Global collector
efficiency []
Figure 3. The Plataforma Solar de Almera. This aerialview of the
research center where the DISS plant is mountedshows the CESA solar
system with central receiver on theleft-hand side, the DISS
collector loop in the center, and theSSPS area on the right-hand
side. Located in the SSPS areaare the CRS solar system with central
receiver, the ACUREXparabolic-trough collector field, a solar
chemistry area, sixparabolic dishes with Stirling engines, a solar
furnace, andother solar facilities for research.
-
In the injection mode, water is injected at severalplaces along
the row of collectors. The measurement sys-tem needed for control
in this mode did not work proper-ly during experiments [3], [7].
The complexity and cost ofthis operating mode made it advisable to
undertake newdevelopments.
In the recirculation mode, the most conservative ofthe three
modes, a water-steam separator is placed at theend of the
evaporation section of the row of solar collec-tors. The amount of
water fed at the inlet of the evapora-tor is greater than the
amount that can be evaporated. Inthe intermediate separator, the
excess water is recirculat-
ed to the collector loop inlet,where it is mixed with the
pre-heated water. The excess water inthe evaporation section
guaran-tees good wetting of the absorbertubes and prevents
stratification.The steam produced is removedfrom the water by the
separatorand is fed into the inlet of thesuperheater section. This
type ofDSG system is highly controllable[6], but the excess
recirculatedwater, the middle water-steamseparator, and the water
recircula-tion pump all increase the para-sitic load of the
system.
The preheating, evapo-ration, and superheatingsections are not
preciselydefined in the once-through and injectionmodes. The length
ofthese zones depends onthe inlet water flow rateand temperature,
the pres-sure in the solar field, and
the radiation available. In the recirculation mode,
thesuperheating process starts in the next-to-last collector,
butthe length of the preheating section and, consequently,
thestarting point of the evaporation section are not
preciselydefined, depending in part on the operating
conditions.
All three modes have advantages and disadvantages.The investment
costs and complexity of the once-throughmode are lowest, and this
mode has the best perfor-mance. On the other hand, the once-through
mode is noteasy to control, requiring a more complex control
system.One of the objectives of the DISS project has been
todemonstrate that it is possible to operate the plant in
theonce-through mode by guaranteeing flow stability andacceptable
controllability [3].
Control ProblemSolar radiation cannot be manipulated and is
subject toslow changes due to the daily cycle and in mirror
reflectivi-ty caused by dust, as well as fast changes, such as
passingclouds. The inlet energy of this plant is affected by
thesedisturbances, as well as by changes in the temperature
orpressure of the inlet water. While these effects are true of
April 200418 IEEE Control Systems Magazine
Parameter Value
Collector row length 500 mCollector type Modified LS-3Collector
aperture 5.76 mNumber of collectors 9 50-m-long collectors
2 25-m-long collectorsOrientation of the solar collectors
North-SouthAbsorber pipe outer diameter 70 mm Absorber pipe inner
diameter 50 mmOptical efficiency of solar collectors 73%Total
mirror surface 2760 m2
Maximum pressure at the field outlet 100 barMaximum outlet
temperature 400 CMaximum steam production 0.85 kg/s
Table 2. Operating points studied in the DISS solar field.
Solar Field Inlet Conditions Temperature Outlet Conditions
TemperatureConditions Pressure [bar] [C] Pressure [bar] [C]Mode 1
40 210 30 300Mode 2 68 270 60 350Mode 3 108 300 100 375
Figure 4. Basic concepts for direct solar steam generation
inparabolic trough collectors. In the once-through mode, wateris
directly converted into superheated steam in the collectorrow. In
the injection mode, water is injected at several placesalong the
collector row. In the recirculation mode, there is awater-steam
separator placed in the middle of the row.
Once-Through Concept
Injection Concept
Recirculation Concept
Table 1. Technical data for the PSA DISS test loop.
-
all solar plants with collector fields, in the particular case
inhand, the control task is still more complex than in Califor-nia
solar power plants [2] or others at PSA [1] because ofthe two-phase
flow, complicating the engineering of the sys-tem as well as the
control system that must be designed forthe solar field. As
mentioned earlier, the thermal fluid cur-rently employed in other
parabolic-trough solar fields is synthetic oil, theoutlet
temperature of which is controlledby mass flow manipulation at the
fieldinlet. In the DISS test loop, both the tem-perature and
pressure of the fluid mustbe controlled to maintain the
desiredsteam conditions at the outlet as deter-mined by turbine
specifications.
Independent of the operating mode,the main objective of the
control system is to obtainsteam at a constant temperature and
pressure at the out-let of the solar field in such a way that
changes producedin the inlet water conditions and in the solar
radiationaffect only the amount of steam produced by the system,and
not its quality.
Control Scheme for the Once-Through ModeThe control scheme for
the once-through mode presentedhere has been designed, implemented,
and tested withinthe framework of the DISS project [3]. As the
first step, themain dynamics were approximated by linear models.
Afterstudying the control scheme and analyzing possible loop
interactions, the SISO transfer functions of all relevant
con-trol loops (Table 3) were experimentally investigated forthe
three different operating points defined in Table 2. Sys-tem
identification was used to estimate open-loop processparameters
such as gains, deadtimes, and time constantsthat experimentally fit
step response data. Based on the
identified low-order models, PI controller parameters werechosen
by means of the process reaction method by simu-lating the
closed-loop responses and by modifying theparameters when necessary
to provide safe stability mar-gins [8], [9]. A final optimization
of the parameter valueswas made in subsequent tests at the plant.
Additionally,the control scheme designed for the once-through
modeincludes mixed feedforward-cascade control schemes tocontrol
the outlet steam temperature.
Some characteristics of the implemented PI functionsare as
follows:
The PI output is calculated using a classical interac-tive
controller. The transfer function of the con-troller has the form
output = K(1 + 1/Tis)error.
April 2004 19IEEE Control Systems Magazine
Table 3. Once-through mode: Models and PI control loop
parameters.
PI Parameters
Control Loop Model Kp Ti [s]
Feed pump G(s) = 0.1375s2 + 1.322s + 0.3329 1.1%/bar 10
Outlet steam pressure G(s) = 4.543103s 5.05 105
s2 + 4.976 102s + 9.693 105 5.3 %/bar 184
Outlet steam G1(s) =a
s + becs 0.0015 kg/s/C 600
temperature control viainjector valve (G1 master where a [3.12,
8.13], b [5 103, 6.6 103]loop, G2 slave loop) and c [70, 100]
G2(s) =3.2 104
s + 0.2 500%/kg/s 12
Outlet steam G1(s) =a
s + becs 8 105 kg/s/C 250
temperature control viafeed valve (G1 master where a [1.365,
2.526], b [1.8 103, 9.6 104]loop, G2 slave loop) and c [395,
750]
G2(s) =3 102s + 0.1 20% kg/s 12
The main task of the control system is toprovide a steady supply
of live steam at
the outlet of the solar field under alloperating conditions.
-
An antireset (integral) wind-up function to compen-sate for
output saturation.
A set-point modifier to define the action to be takenon a
set-point change. The two options implemented
are PI and integral only on set-point change. The firstoption
causes a jump in the control output due tothe proportional
contribution from the error created
by a set-point change, whereas, in the second option,the
proportional contribution of the error is sub-tracted from the
integral contribution, thus eliminat-ing the jump in control output
and resulting in
integral action only on a change in set-point. Bumpless
manual-to-auto transfer.
The process diagram, including themost important feedback loops
for theonce-through operating mode, isshown in Figure 5. The main
controlloops for the solar field in the once-through operating mode
are as fol-lows: Feed pump control loop: The rota-
tional speed of the feed pump isadjusted by a PI controller to
maintain a specificpressure drop in the feed valve, creating a flow
insteady state that is directly proportional to the valve
April 200420 IEEE Control Systems Magazine
Figure 5. Diagram of the DISS test loop configured in the
once-through mode. Four control loops comprise the implemented
control scheme. Although outlet steam pressure and temperature
loops are the main control loops, there is anadditional controller
for maintaining a constant pressure drop in the feed valve.
POT
FT
FC TC
Solar Collectors Row
PDC
FeedPreheater
Feed PumpFeedwater Tank Air Condenser L.P.
EquivalentTurbineLoad
H.P
. Steam
Superheated Steam
LC
FinalSteam Separator
FC TC
FTLT
TT
PTPC
Vapor Feed
Injection LineFlashTank
LegendTT - Temperature TransmitterFT - Flow TransmitterPT -
Pressure TransmitterPDT - Pressure Drop TransmitterLT - Level
Transmitter
TC - Temperture Control LoopFC - Flow Control LoopPC - Pressure
Control LoopPDC - Pressure Drop Control LoopLC - Level Control
Loop
In the once-through mode, feedwateris preheated, evaporated, and
convertedinto superheated steam as it circulatesfrom the inlet to
the outlet of thecollector loop.
-
opening. The feed pump control loop, therefore, pro-vides a
linearized flow relationship between valveposition and flow for PI
control.
Outlet steam pressure control: The steam produced bythe
collector row feeds a steam separator. The outletsteam pressure is
kept constant by adjusting a steamcontrol valve with a PI
controller.
Outlet steam temperature control loops: The outlettemperature
control is achieved by inlet feed flowcontrol and water injection
in the superheater. Theformer control ensures that the steady-state
inletflow matches radiation conditions, whereas the lat-ter control
is based on PI-feedforward control to pro-vide rapid response to
sudden disturbances bymeans of water injection at the inlet of the
last col-lector of the solar field.
PI functions for the first two control loops are imple-mented
based on the process reaction method, and stabili-ty margins are
determined by simplified linearized models
(see Table 3). Outlet steam temperature control requires amore
detailed design because the process is strongly affect-ed by
disturbances at the inlet and by disturbance vari-ables, and
acceptable control cannot be achieved withconventional PI or PID
schemes. Contrary to the recircula-tion mode illustrated in Figure
6, in the once-through modethere is no intermediate separator in
the field that sup-presses disturbances occurring in the preheating
and evap-oration sections, and the starting point of the
superheatingsection is not precisely defined. This condition
reduces thecontrollability of the once-through operating mode
whencompared with the recirculation mode, in which the con-trol
loops are based on the following simple PI
controllers:recirculation pump control loop (recirculation flow
con-trolled by PI control of the rotational speed of the
recircula-tion pump), feed pump control loop (the rotational
speedof the feed pump adjusted by a PI controller to maintain
aspecific pressure drop across the feed valve), middle
steamseparator liquid level control loop (to maintain the level
April 2004 21IEEE Control Systems Magazine
Figure 6. Diagram of the DISS test loop configured in the
recirculation mode. Five control loops comprise the implemented
control scheme. Although outlet steam pressure and temperature
loops are the main control loops, there arethree additional
controllers for settling the recirculation water flow, the pressure
drop in the feed valve, and the liquid level inthe middle
separator.
Solar Collectors Row
PDC
FeedPreheater
Feed PumpFeedwater Tank Air Condenser L.P.
EquivalentTurbine
Load
H.P
. Steam
Superheated Steam
LC
FinalSteam Separator
TC
LT
TT
PTPC
Vapor Feed
Injection Line
LegendTT - Temperature TransmitterFT - Flow TransmitterPT -
Pressure TransmitterPDT - Pressure Drop TransmitterLT - Level
Transmitter
TC - Temperture Control LoopFC - Flow Control LoopPC - Pressure
Control LoopPDC - Pressure Drop Control LoopLC - Level Control
Loop
FlashTank
PDT
LC LTSeparator
MiddleSteam
FT FC
RecirculationPump
FeedValve
-
around a nominal value, the feed flow is adjusted to controlthe
aperture of the feed valve whose pressure drop is beingcontrolled
by the feed pump with P-PI control), outletsteam pressure control
loop (by adjusting the aperture ofthe steam control valve in the
steam separator with PI con-trol), and outlet steam temperature
control loop (by waterinjection in the inlet of the last collector
using PI control ofthe injector valve). The last two control loops
are the maincontrollers of the system in recirculation mode used
toguarantee the steam quality at each instant. The
remainingcontrollers are needed to improve the behavior of the
over-all control system and for operational feasibility. Themethod
for obtaining the PI parameters is the same asexplained for the
case of the PI controllers for the once-through mode.
The solution adopted in the DISS project for the once-through
mode is to control the outlet steam temperaturewith forward action.
The parameters of the PI functionsappearing in these schemes are
chosen using the processreaction method and by studying the
closed-loop systemstability margins. Parameters a, b, and c in
Table 3 are
related to the uncertainty of the models obtained. Depend-ing on
the operating conditions (outlet steam flow produc-tion,
temperature, pressure, and solar radiation available),the gain,
time constants, and time delays vary. The differ-ent model
parameters influence the PI control design.Therefore, once a set of
PI parameters is chosen, closed-loop simulations are performed by
varying the model para-meters to guarantee wide stability margins
for the wholerange of model parameters. Consequently, the selection
ofthe PI parameters is conservative. The detailed schemesare
discussed in the following subsections.
Although interactions between loops exist, they aresmall because
the two slave loops, which are fast com-pared to the other loops,
are able to reject slow distur-bances due to the interactions
caused by other loops.The outlet steam pressure loop is also faster
due to asmaller time constant and no dead time and is thus ableto
reject the disturbances coming from the slower tem-perature loops.
Also, the temperature loops have aninherent interaction reduction
mechanism. As a result,the interactions are canceled by the control
strategy
April 200422 IEEE Control Systems Magazine
Figure 7. Outlet steam temperature control based on forward
action by means of a feed valve. The outer loop of the
cascadestructure consists of a feedforward controller, which
dictates the nominal feedwater flow, in parallel with a PI
controller. Theinner control loop is a PI controller with
antiwindup action.
ETin
TambTinj
minj_set
FeedforwardController
FFFV
PITrefmeTeT
mff
min_dem em_in PIAnti-Windup Plant
Tout+ afv
min
Table 4. Specific enthalpy of the fluid: Parameters of the
linear regressions.
StandardPressure [bar] Phase Temperature [C] a1 a2 R-square
Deviation [kJ/kg]30 Water 100 < Tfluid < 234 21 4.38 0.99989
3
Steam 234 < Tfluid < 400 +2225 2.54 0.99868 760 Water 100
< Tfluid < 276 34 4.48 0.99965 7
Steam 276 < Tfluid < 400 +1940 3.12 0.99686 11100 Water
100 < Tfluid < 312 54 4.61 0.99905 12
Steam 312 < Tfluid < 400 +1480 4.10 0.99415 15
-
designed. These points willbe explained more fully inthe
following sections.
FeedforwardControl of OutletSteam Temperature
Feed Valve AdjustmentThe great variations in solarradiation and
the long resi-dence time of the fluid in thefield call for the use
of feedfor-ward action to anticipate theeffect of load changes on
theoutputs to be controlled; thatis, the control system
shouldcalculate the adequate value of the inlet mass flow in
advanceso that the outlet steam temperature remains within therange
of desirable reference values. The performance of thesystem in the
once-through mode depends on the inlet flowcontrol. Changes in
solar radiation and inlet fluid temperaturerequire the flow rate to
change to maintain the desirable out-put. If changes involve wide
oscillations, the solar field perfor-mance is strongly affected.
Not only do thermal and pumpinglosses increase, but the relatively
narrow margin between thedesign maximum outlet temperature and the
actual tempera-ture, which triggers the alarm signal, may be
bridged by wideoscillations.
To manage these instabilities, the outlet steam tempera-ture
control loop is a mixed cascade-feedforward controlloop (Figure 7)
aimed at guaranteeing a desired flow in theface of valve
nonlinearities and changes in disturbancesaffecting the loop (see
Figure 7 and Nomenclature). Thefeedforward term uses a process
model to effect changesin the controller output in response to
measured changesin the load before errors occur. The outer loop is
com-posed of a feedforward function FFFV in parallel with a
PIcontroller with fixed parameters. Block FFFV calculates anominal
flow mff , and the parallel PI controller correctsthis value
according to the current output Tout. In a set-point change, this
PI controller uses only integral actionbecause the new temperature
reference also passes to theFFFV block that calculates the nominal
flow mff . The flowmin dem calculated by this master loop is the
input to theinner slave PI control loop, which calculates a new
aper-ture afv of the feed valve. The saturation included in frontof
the PI inner control loop limits the inlet water flow to aminimum
value of 0.3 kg/s; this limitation guarantees tur-bulent flow in
the absorber pipes and consequently con-strains the temperature
gradients in the cross-sectionalarea of the pipes that should be
less than 50 C, which is atemperature gradient limit from the point
of view of thepipe thermal stress.
As previously mentioned, the PI parameters of the outerand inner
loops are calculated from open-loop responsesusing the process
reaction method for feed pump control andpressure control loops.
Conservative parameters are chosenfor the PI controller of the
master loop, reducing interactionswith the temperature control loop
using the injector.
The feedforward action is obtained from the
simplifiedsteady-state energy balance formulation for the
collectorrow [13] given by
{Enthalpy
out
}
{Enthalpy
in
}=
{Energy
collected
} {Losses}.
The collected energy is corrected using an estimated effi-ciency
factor that implicitly considers the optical efficien-cy and,
consequently, the optical losses. The simplifiedenergy balance
equation can be written as
(min + minj)hout (minhin + minjhinj) =loop Acol LloopE Ul
Sabs(Tav Tamb).
In this equation, the specific enthalpy hout at the outlet
isreplaced by the outlet enthalpy reference href , and thewater
flow rate minj injected in the last collector isreplaced by the
nominal injection flow minj set establishedin the temperature
control loop by means of the injectorvalve to avoid feeding back
variations (that could be oscil-latory) dictated by the temperature
control loop by meansof the injector valve in the block FFFV. Such
feedback dete-riorates the temperature response. Considering these
sub-stitutions, the feedforward control equation used tocalculate
the nominal feedwater flow mff to achieve thedesired outlet
temperature Tref is given by
mff =loop Acol LloopEUl Sabs(Tav Tamb)minj set(hrefhinj)
href hin,
April 2004 23IEEE Control Systems Magazine
Table 5. Thermal loss factor Ul in LS-3 collectors: b1, b2, and
b3 values.
Fluid Average Temperature [C] b1 b2 b3Tav < 200 0.687257
0.00194 0.000026200 < Tav < 300 1.433242 0.00566 0.000046300
< Tav 2.895474 0.01640 0.000065
Table 6. Average temperature values used in the FFFV controller
for the three operating points of the DISS test loop.
Pressure [bar] Inlet temperature [C] Outlet Temperature [C] Tav
[C]
30 210 300 23760 240 350 277
100 280 400 316
-
where loop was estimated from experimental data as 0.53;Acol,
Lloop, Sabs are geometric parameters (Table 1); E is afiltered
value of the measured solar irradiance; href, hinj,and hin are
specific enthalpy values calculated from theoutlet pressure P and
the corresponding temperaturesTref, Tinj, and Tin using
hfluid|P = a1 + a2Tfluid [kJ/kg],
where a1 and a2 are coefficients estimated by linear regres-sion
using the enthalpies and temperature values in ther-modynamic
tables [10] (Table 4). Ul is a factor related tothe thermal losses,
which for an LS-3 type collector can beapproximated by [11]
Ul = b1 + b2(Tav Tamb) + b3(Tav Tamb)2,
where b1, b2, and b3 depend on the average temperature ofthe
fluid in the absorber pipes (Table 5). To simplify thecontrol-loop
structure, the average temperature Tav of thefluid in the field is
approximated by a constant value for thethree different operating
points. Values based on inlet andoutlet conditions and conditions
in the preheating, evapo-ration, and superheating sections are
listed in Table 6.
Injector Valve AdjustmentThe outlet steam temperature can also
be controlled byinjecting preheated water into the last collector,
providinganother degree of freedom to allow a fast reaction in
the
outlet temperature; however, there are regular changes inthe
outlet steam temperature of the previous collector, inthe steam
flow rate, and in the injection water temperatureinfluencing the
behavior of this loop. These changes aremore frequent and stronger
in the once-through modethan in the recirculation mode due to the
lack of an inter-mediate separator.
A controller based on forward action has also beendesigned for
the outlet steam loop. The feedforward blockcorrects the injection
water flow rate at the inlet of the lastcollector, taking into
account changes in the collector inlettemperature and mass flow,
the injection water tempera-ture, and the outlet temperature
reference.
The mixed cascade-feedforward control scheme isshown in Figure
8, where the cascade structure compen-sates for actuator
nonlinearities. The outer loop is com-posed of a feedforward
function FFIV in parallel with a PIcontroller with fixed
parameters. The output mff iv of theblock FFIV corrects the PI
controller output meT . Whenthe controller is set in automatic
mode, a nominal injectionwater flow minj set is established at
around 10% of theexpected steam mass production. The outer control
loopcorrects this nominal value, and a new injection water
flowvalue minj dem is calculated and dictated from the
injector.This new value is the input for the inner loop, a PI
controlloop, which determines a new aperture value aiv for
theinjection valve. Injection valve nonlinearity detected dur-ing
experiments is compensated by the cascade structure.The saturation
included in the master loop avoids zero
April 200424 IEEE Control Systems Magazine
Figure 8. Outlet steam temperature control based on feedforward
action by means of an injector. The nominal injection flowis
dictated manually when the controller is operating in automatic
mode. This nominal value is corrected by the output of theouter
loop of the cascade structure, which consists of a feedforward
controller in parallel with a PI controller. The inner con-trol
loop is based on a PI controller with antiwindup action.
Tinj
Tin_c
min_c
FeedforwardController
FFIV
PI
PIAntiwindup
Plant
Tref eT meT
mff_iv
minjminj
Tout
minj_set
minj_dem aiv
-
April 2004 25IEEE Control Systems Magazine
Table 8. Outlet steam temperature control with injector valve:
FFIV parameters.
Outlet Steam StandardPressure [bar] c1 c2 c3 c4 c5 R-square
Deviation [kg/s]
30 6.212 104 0.00313 1.7 106 3.0 106 0.00171 0.95035 0.0011260
8.457 104 0.00505 4.4 106 7.1 106 0.00279 0.95219 0.00146
100 8.942 104 0.00494 4.2 106 5.8 106 0.00261 0.95167
0.00117
Table 7. Design of the FFFV: Input data sets.
Input 30 bar 60 bar 100 bar
Direct solar irradiance1 [650, 1000] W/m2 [650, 1000] W/m2 [650,
1000] W/m2
Global collector efficiency1 [0.40, 0.60] [0.40, 0.60] [0.40,
0.60] Collector inlet mass flow [0.35, 0.70] kg/s [0.350.70] kg/s
[0.35, 0.70] kg/sOutlet temperature reference [280, 320] C [320,
370] C [340, 400] CCollector inlet fluid temperature [250, 310] C
[290, 370] C [330, 390] CInjection water temperature [180, 215] C
[220, 260] C [260, 300] C1Changes in this parameter do not have
significant influence on the adjusted model.
Figure 9. Control loop responses during operation at 30 bar (22
April 2002): (a) includes the feed pump control loopresponse, and
(b) includes the temperature control loop response; (c) includes
the outlet pressure control loop response, while(d) includes the
available radiation and generated steam flow.
Outlet Steam Temperature
Flow
[kg/s
]
1.00.90.80.70.60.50.40.30.20.10.0
Flow
[kg/s
]
1.00.90.80.70.60.50.40.30.20.10.0
Pres
sure
[bar]
Pres
sure
[bar]
7
6
5
4
3
2
1
0
Pressure Drop Across Feed ValveSet PointFeed Pump Power (Control
Signal)
70
60
50
40
30
20
10
0
330300270240210180150120906030
Pow
er [%
]Te
mpe
ratu
re [
C]
09 10 11 12 13 14 15 16Local Time
(a)09 10 11 12 13 14 15 16
Local Time(b)
09 10 11 12 13 14 15 16Local Time
(d) 09 10 11 12 13 14 15 16
Local Time(c)
45
40353025
2015
105
0
72
645648403224
1680
1,000900800700600500400300200100
0
Aper
ture
[%]
Sola
r Irra
dian
ce [W
/m2]
Outlet Steam PressureSet PointFeed Tank PressureSteam Valve
Aperture (Control Signal)
Direct Solar IrradianceOutlet Steam Flow
Set PointInlet Water Flow - Feed Valve (Control Signal)Injection
Water Flow - Injector (Control Signal)
-
flow rate dictated from the injector valve, since zero flowrate
would deteriorate the control action due to the non-linearity of
the injector when this actuator is nearly closed,as was observed in
real tests.
The PI parameters of the outer and inner loops werealso
calculated from open-loop responses using theprocess reaction
method. Linearized models detailed inTable 3 were used to simulate
the closed-loop responsesand study the stability margins in the
worst cases for themodel uncertainty. The final selection of values
for the PIparameters was made in a conservative way to avoid
insta-bility in the system and to diminish interaction with therest
of controllers [8].
The feedforward action formulated for the feed valvecontrol loop
is obtained from a simplified steady-stateenergy balance
formulation for the collector given by
{Enthalpy
out
}
{Enthalpy
in
}=
{Energy
collected
} {Losses}.
The collected energy is corrected by the global efficiencycol of
the collector, which accounts for the thermal andoptical
efficiencies and thus the thermal and optical losses.The simplified
energy balance equation can be written as
(min c + minj)hout (min chin c + minjhinj) = col Acol LcolE,
where the right-hand side of the equation includes theenergy
collected and energy losses.
Substituting the specific enthalpy at the outlet hout bythe
outlet enthalpy reference href, which is directly calcu-lated from
the temperature and pressure references, the
April 200426 IEEE Control Systems Magazine
Figure 10. Control loop responses during operation at 30 bar (26
April 2002): (a) includes the feed pump control loopresponse, (b)
includes the temperature control loops response, (c) includes the
outlet pressure control loop response, and (d)includes the
available radiation and generated steam flow.
7
6
5
4
3
2
1
0
Pres
sure
[bar]
45
40
35
30
25
20
15
10
5
0
Pres
sure
[bar]
09 10 11 12 13 14 15Local Time
09 10 11 12 13 14 15Local Time
09 10 11 12 13 14 15Local Time
09 10 11 12 13 14 15Local Time
70
60
50
40
30
20
10
0 306090
120150180210240270300330
Pow
er
[%]
Tem
pera
ture
[C]
72
64
56
48
40
32
24
16
8
0
1,000900800700600500400300200100
0
Aper
ture
[%]
Sola
r Irra
dian
ce [W
/m2]
1.00.90.80.70.60.50.40.30.20.10.0
Flow
[kg/s
]
1.00.90.80.70.60.50.40.30.20.10.0
Flow
[kg/s
]
Outlet Steam TemperatureSet PointInlet Water Flow - Feed Valve
(Control Signal)Injection Water Flow - Injector (Control
Signal)
Direct Solar IrradianceEffective Radiance on AbsorbersOutlet
Steam Flow
Outlet Steam PressureSet PointFeed Tank PressureSteam Valve
Aperture (Control Signal)
Pressure Drop Across Feed ValveSet PointFeed Pump Power (Control
Signal)
(a)
(c) (d)
(b)
-
corresponding injection flow in steady state is given by
minj = col Acol LcolE min c(href hin c)href hinj.
Using this equation, thermodynamic tables [10] for calcu-lating
the enthalpies corresponding to each temperatureand pressure, and
data obtained using the data seriesdetailed in Table 7, a
regression analysis was performed toobtain the feedforward function
for calculating the injec-tion flow rate correction. The multiple
regression modelhas the form
mff iv = c1Tin c + c2min c + c3Tref + c4Tinj + c5.
The values obtained for the parameters c1, c2, c3, c4, and c5at
the various operating points are listed in Table 8. Theseparameters
depend on thermodynamic properties of thefluid as well as the
geometry and global efficiency of thecollector. Table 8 includes
the correlation coefficient andstandard deviation of the residual
errors. A good approxi-mation is obtained within the operating
ranges listed inTable 7, but outside these ranges the quality of
the model isnot guaranteed due to the nonlinear characteristics of
theprocess.
Both temperature control loops operate in parallel andare
necessary because the temperature control based onthe feed valve
adjustment calculates a nominal inlet flowrate for the field. Due
to the long time delay caused by thelength of the collector loop,
the temperature control loopscannot react rapidly to sudden
disturbances. The tempera-ture control based on the injector valve
provides a fastercontrol to compensate for sudden changes and
allows theoutlet temperature to more accurately follow the
refer-ence. Interactions between both controllers are avoided bythe
inclusion of parameter minj set in block FFFV and bychoosing
conservative PI parameters in the case of thetemperature control by
means of the feed valve.
Representative Experimental ResultsThe Symphony SCADA platform
[12] was used to imple-ment the controllers. Figure 9 shows the
results obtainedduring an experiment in once-through mode with 30
bar ofoutlet steam pressure. During startup, the water pressurein
the feed water tank varied, affecting the feed pump
response. The measured outlet steamflow oscillated due to the
circulation ofwater through the steam flow transmit-ter, providing
an incorrect signal. Outletsteam temperature was maintained at
aconstant temperature of 300 C. Themaximum steam temperature
overshootand undershoot occurred when thesuperheated steam
production started ataround 12 noon; their values were 8 C
(2.7% of the reference) and 16 C (5.3% of the
reference),respectively, far from the steam saturation
temperature,which is 234 C at 30 bar. The steam pressure was
main-tained close to the reference throughout operation
withoutsignificant deviations.
Figure 10 shows the results obtained during an experi-ment in
the once-through mode with 30 bar of outlet steampressure. The
experiments objective was to evaluate theresponse of the control
system to a defocusing of one col-lector of the evaporation
section, which is equivalent toproducing a 10% step decrement in
the inlet energy to thefield. Collector number 6 was defocused at
13:15 andstayed out of focus for 5 minutes. The resulting outlet
tem-perature deviation was 21 C (7% of the reference). Whenthe
temperature came close to the reference again, theirradiance
dropped 300 W/m2, causing the temperature toapproach the saturation
temperature value and changingthe outlet steam pressure around 0.8
bar (2.6% of the refer-ence). Subsequently, the nominal conditions
were recov-ered in 15 min. Prior to defocusing, the irradiance
haddropped around 150 W/m2 at 12:30, which mainly affectedsteam
flow. The outlet steam pressure was maintainedconstant, and the
maximum outlet steam temperaturedeviation was 4.5 C.
Figure 11 shows the results obtained during a once-through mode
experiment with 60 bar of outlet steam pres-sure. Outlet steam
temperature was maintained constantat 350 C, and the maximum
deviation from this referencewas 6 C (1.7% of the reference). The
temperature con-troller based on the injector valve adjustment was
put inautomatic mode around 10:00, but the superheatingprocess
started 45 min later. The dead time in response tothe steam
temperature is due to manual control of the inletflow rate during
the start-up (see also Figure 10). The oper-ator put the feed valve
temperature controller in automat-ic mode at around 11:00. The
experiment finished at 15:45when production was stopped.
The results show that all set points can be main-tained during
steady-state conditions, even with shorttransients in the solar
radiation. During longer solarradiation transients, it is more
difficult to maintain thesteam temperature, since a minimum flow
must be guar-anteed to avoid high temperature gradients in
pipecross-sections when solar radiation recovers (in any
April 2004 27IEEE Control Systems Magazine
The DISS project demonstrated that it ispossible to produce
high-pressure, high-temperature steam directly inparabolic trough
solar collectors.
-
case, null values of inlet mass flow led to zero produc-tion,
which is commercially undesirable). In this tran-sient situation, a
mixed steam/water flow feeds theseparator tank where it condenses,
returning to the feedwater through the separator drain valve,
increasing theparasitic load of the system as well as security. A
con-trol system configured to operate the plant with zeroinlet flow
would have to satisfy stringent specifications,mostly under
actuator saturation. Such operation wouldrequire that all the
collectors be defocused to avoiddangerous conditions in the solar
field.
ConclusionsThe DISS project demonstrated that it is possible to
pro-duce high-pressure, high-temperature steam directly inparabolic
trough solar collectors. A leading plant usingsolar technology has
been operated in two differentmodes. This article describes the
once-through mode,
which is the most difficult to control. Using a schemebased on
PI and feedforward controllers, the controllabili-ty of the plant
is guaranteed on clear days and duringshort transients in solar
radiation. Longer transients insolar radiation make it difficult to
maintain the steam tem-perature in favor of guaranteeing a minimum
flow to avoidhigh temperature gradients in the cross-sectional area
ofthe collectors absorber pipes when the solar radiationlevel is
recovered.
A structure partially based on classical controllers waschosen
because the plant operators are familiar with thistype of
controller and can adapt the controller parametersin situations
affecting plant dynamics and controller per-formance, such as
modifications in plant layout or systemchanges over time. In the
near future, a control strategybased on model predictive control
will be investigated toimprove system performance under
disturbances in thesystem inlet energy.
April 200428 IEEE Control Systems Magazine
Figure 11. Control loop responses during operation at 60 bar (17
July 2002). (a) includes the feed pump control loopresponse, (b)
includes the temperature control loops response, (c) includes the
outlet pressure control loop response, and (d)includes the
available radiation and generated steam flow.
8
7
6
5
4
3
2
1
0
Pres
sure
[Bar]
Pres
sure
[Bar]
9080706050403020100
80
70
60
50
40
30
20
10
0 6090
120150180210240270300330360
Pow
er [%
]Te
mpe
ratu
re [
C]
04
812
162024
283236 1,000
900800700600500400300200100
0
Aper
ture
[%]
Sola
r Irra
dian
ce (W
/m2)
1.00.90.80.70.60.50.40.30.20.10
Flow
[kg/s
]
1.00.90.80.70.60.50.40.30.20.10
Flow
[kg/s
]
09 10 11 12 13 14 15 16Local Time
09 10 11 12 13 14 15 16Local Time
09 10 12 13 14 15Local Time
09 10 11 12 13 14 15 16Local Time
Direct Solar RadianceOutlet Steam Flow
Outlet Steam TemperatureSet PointInlet Water Flow - Feed Valve
(Control Signal)Injection Water Flow - Injector (Control
Signal)Pressure Drop Across Feed ValveSet Point
Feed Pump Power (Control Signal)
Outlet Steam PressureSet PointFeed Tank PressureSteam Pump Power
(Control Signal)
(a)
(c) (d)
(b)
-
AcknowledgmentsWe thank the European Commission for support of
the sec-ond phase of the DISS project (contract
JOR3-CT98-0277)within the framework of the E.U. JOULE Program.
Wewould also thank MCYT for funding this work under
grantsDPI2002-04375, DPI2001-2380, and QUI99-0663. We thankthe
anonymous reviewers for their helpful comments thatimproved the
article.
References[1] E.F. Camacho, M. Berenguel, and F.R. Rubio,
Advanced Control ofSolar Plants. London: Springer-Verlag, 1997.
[2] G. Cohen and D. Kearney, Current experiences with SEGS
parabol-ic trough plants, in Proc. 8th Int. Symp. Solar Thermal
ConcentratingTechnologies, Kln, Germany, 1996, pp. 217244.
[3] E. Zarza, L. Valenzuela, J. Len, D. Weyers, M. Eickhoff, M.
Eck, andK. Hennecke, The DISS project: Direct steam generation in
parabolictrough systems. Operation & maintenance experience and
update onproject status, J. Solar Energy Eng., vol. 124, May 2002,
pp. 126133.
[4] E. Zarza, Solar Thermal Desalination Project. Phase II
Results & FinalProject Report. Madrid, Spain: Editorial CIEMAT.
1995.
[5] University of Manchester (UMIST), Zentrum fr Sonnenergie-
undWasserstoff-Foruschung Baden- Wrttemberg (ZSW), PSA DISS
testfacility: Control scheme design studies for once-through and
recircula-tion concepts, Project DISS Int. Rep., Doc. ID:
DISS-EN-CD-02, Platafor-ma Solar de Almera, Aug. 1996.
[6] L. Valenzuela, M. Berenguel, E. Zarza, and E.F. Camacho,
Controlschemes for direct steam generation in parabolic solar
collectorsunder recirculation operation, submitted to Solar Energy
J., 2002.
[7] M. Eck and M. Eberl, Controller design for injection mode
drivendirect solar steam generating parabolic trough collectors, in
ISESSolar World Congress, Jerusalem, Israel, vol. I, 1999, pp.
247257.
[8] B.A. Ogunnaike and W.H. Ray, Process Dynamics, Modeling, and
Con-trol. New York: Oxford Univ. Press, 1994.
[9] R.N. Bateson, Introduction to Control System Technology. New
York:Prentice-Hall, 1996.
[10] W. Wagner and A. Kruse, Properties of Water and Steam.
Berlin-Hei-delberg-New York: Springer-Verlag, 1998.
[11] J.I. Ajona, Electricity generation with distributed
collector, inSolar Thermal Electricity Generation. Madrid, Spain:
Editorial CIEMAT,1999, pp. 777.
[12] Composer Series. Electronic Documentation Symphony.
ElsagBailey Process Automation, Ohio (U.S.A), 19971998.
[13] D.R. Coughanowr, Process Systems Analysis and Control. New
York:McGraw Hill, 1991.
Loreto Valenzuela ([email protected]) receivedthe B.S. in
physics and the electronics engineering degreefrom University of
Granada in 1994 and 1996, respectively.She is currently pursuing
her Ph.D. in the University ofAlmera. She joined the Plataforma
Solar de Almera in 1997to work as technical assistant in the
Department of Parabol-ic-trough Collectors, and in 2000 she took a
position withinthe research staff of the CIEMAT (the public
research cen-ter leading research on renewable energy in Spain).
Herresearch interests are in solar thermal concentrating tech-
nologies and control systems for solar energy systems. Shecan be
contacted at CIEMAT, Plataforma Solar de Almera,Ctra. Senes s/n,
E-04200 Tabernas, Almera, Spain.
Eduardo Zarza earned the industrial engineering degreefrom
University of Seville in 1985, where he is a member ofthe
Scientific Group of Thermodynamics and RenewableEnergies. Since
1985 he has participated in many interna-tional R&TD projects
regarding solar thermal energy appli-cations. From 1990 to 1994 he
was the project manager ofthe Solar Thermal Desalination Project.
In 1994 he under-took the coordination and management of the DISS
project,thus leading an international group of researchers
involvedin the development of the DSG technology. Since 1997, hehas
been the head of the Department of Parabolic-troughCollectors of
CIEMAT, where he conducts research on inno-vative solar energy
applications and systems.
Manuel Berenguel is an associate professor in the Depar-tamento
de Lenguajes y Computacin (rea de Ingenierade Sistemas y Automtica)
of the University of Almera,Spain. He earned the industrial
engineering degree and doc-torate from the Escuela Superior de
Ingenieros Industrialesof the University of Sevilla (Spain), where
he received thePremio Extraordinario de Doctorado award (given to
thebest engineering thesis of the year), and he was aresearcher and
associate professor in the Departamento deIngeniera de Sistemas y
Automtica for six years. Hisresearch interests are in the fields of
predictive, adaptive,and robust control, with applications to solar
energy sys-tems, agriculture, and biotechnology. He has been a
review-er for several journals including Control
EngineeringPractice; IEEE Transactions on System, Man and
Cybernetics;and IEEE Transactions on Fuzzy Systems. He has
authoredand coauthored more than 50 technical papers and is
co-author of Advanced Control of Solar (Springer, 1997).
Eduardo F. Camacho is a professor of system engineeringand
automatic control at the University of Seville. He haswritten the
books Model Predictive Control in the ProcessIndustry
(Springer-Verlag, 1995), Advanced Control of SolarPlants
(Springer-Verlag, 1997), Model Predictive Control(Springer-Verlag,
1999), and Control e Instrumentacin deProcesos Qumicos (Ed.
Sintesis). He has authored and co-authored more than a 150
technical papers, has served onvarious technical committees, and
was a member of the2001 Board of Governors of the IEEE Control
Systems Soci-ety. At present he is the chair of the IEEE Control
SystemsSociety International Affairs Committee, vice president
ofthe European Control Association, and chair of the
IFACPublication Committee. He is one of the editors of
ControlEngineering Practice and an associate editor of the
Euro-pean Journal of Control.
April 2004 29IEEE Control Systems Magazine
footer1: