CHAPTER 10
OPTIMIZATION OF PLANT OUTPUT
For initial site screening the volume of a proposed barrage basin may be
related to approximate annual energy estimates. However, accurate estimates
of energy output depend upon a summation of the continuously varying pro
duction of energy since the operating heads change with the varying sea
and basin levels, as well as upon the specific operating characteristics of the
turbogenerating equipment. The major portion of this chapter is devoted
primarily to analysis of single-effect, ebb generation for maximum energy pro
duction, without pumping. However, the same basic approach would apply,
although somewhat more complex, to an analysis of the double-effect operat
ing mode.
10.1
PRINCIPAL VARIABLES
Basically, the objectives of generation optimization are, first, to select the size
and characteristics of turbogenerating machinery to suit site conditions and
output requirements, and second, to determine the number of sluices and
generating units that should be installed for maximum energy production to
yield the lowest overall cost per unit of output for each location. The princi
pal variables in the analysis, for any specific site, can be separated into two
classes-those directly related to costs and those related to operating condi
tions and largely independent of cost, as follows:
Elements of Tidal-Electric Engineering. By Robert H. Clark
Copyright 2007 the Institute of Electrical and Electronics Engineers, Inc.
1 17
1 18
OPTIMIZATION OF PLANT OUTPUT
1. Cost-related variables
a. Dike and sluiceway designs
b. Turbogenerator design
c. Numbers of sluices and turbines
2. Operating variables
a. Time of commencing generation
b. Time of concluding generation
c. Timing of reversal and pumping cycles for other operating modes
d. Runner blade and/or distributor gate settings during generation
The method outlined in this chapter for determining the optimum com
bination of sluice and turbine capacity requires: (1) mathematical modeling
to find the maximum output attainable with a given set of sluice and turbine
capacities; (2) using the models to determine outputs for a wide range of
such capacities; and (3) comparing cost versus output over the whole range.
A conceptual treament of these steps follows.
10.2
ENERGY PRODUCTION AND OPTIMIZATION
The variations in water level during a cycle of ebb generation are illustrated
in Fig. 10.1. As described in Chapter 5, the generation phase commences
at point A when the gross generating head is Hs and continues until point
B when the head is equal to HF At this point, the turbine openings are
closed off, the basin level remains constant, and the sea level continues rising.
At point C, the levels become equal and the sluices are opened to allow
the tidal basin to refill. If double regulated, the runner blades, as well as the
Turbining
IBL
1
FBL
Water Elevation on
Seaward Side of
Barrier
c
o
Water Elevation in
the Tidal Basin
>
Q)
iIi
Refilling
Time
Figure 10.1
Ebb generation parameters for mathematical model.
ENERGY PRODUCTION AND OPTIMIZATION
1 19
distributor gates, can be adjusted to facilitate inflow to the basin through the
turbine waterways. Basin filling continues until the levels again become equal
at point D when the sluices are closed and the basin level held at its top eleva
tion until the generation cycle commences again. For any selected turbine
design, there will be a minimum value of Hs and HF below which unit synchro
nization will be impractical. Moreover, to obtain maximum energy production,
continuing operation to these minimum values is not necessarily required.
There are, in fact, an infinite number of operational timings for each level of
installation, only one of which will yield maximum energy production.
In simple terms, it may be imagined that if a small number of turbines were
used for ebb generation, only a small amount of water would be released from
the basin and the basin water level would remain relatively high. As illustrated
by Fig. 10.2, if the number of turbines were increased, more water would be
released, in a given time and the basin level would fall off more rapidly.
The optimum operating condition can be effectively determined by a
computerized mathematical model. The objective of the model would be to
produce estimates of energy production at a selected site for several combi
nations of turbine numbers and sluiceway capacities, and, by obtaining a cost
estimate of each combination, it would be possible to select the installation
that produces the most economical energy at a site, that is, the optimization
of the scale of development. A characteristic site optimization curve is illus
trated in Fig. 10.3.
Filling
Generating
Notes:
I I I
I II
I II
I II
I II
I II
I II
" 0111
'(} I
01
If the optimum number
of turbines causes basin
level variation CD, fewer
turbines produce basin
level CD, and more
turbines produce GJ.
Similarly, if the optimum
number of sluices causes
the basin level to rise
as in , fewer sluices
will cause slower filling l2l
and increasing the
number of sluices
produces faster filling
Time -----
Figure 10.2
Effect of varying numbers of sluices and turbines.
120
OPTIMIZATION OF PLANT OUTPUT
, "ea"Og N,m"", of SI"",
YI
\
\/
_/
Turbines
Optimum number of
turbines and sluices
is found at this point.
Energy (GWh)
Figure 10.3
10.3
-----
Characteristic optimization curve.
SIMULATION OF OPERATION
In order to simulate with reasonable accuracy the operation of a tidal
electric plant, the mathematical model should account for the following
principal features:
1. The tidal basin shape in terms of related water elevations and surface
areas from which volume changes can be deduced
2. A correction for the "backwater effect" in the tidal basin during plant
operation in either direction
3. The shape of a typical tidal cycle expressed as a time series of water
surface elevations
4. The use of either fixed-blade or variable-pitch runners accounting for
efficiency variations
5. The operation of the sluices and of the turbines as orifices, with dis
charge expressed as a function of head
6. Variations in startup and finishing conditions with respect to turbining
and sluicing
When such characteristics are included in the model, appropriate variations
in tidal amplitude, turbine characteristics, and numbers of turbines and sluices
DEVELOPMENT OF MODELS
121
can be introduced in each run of the model program to facilitate the estima
tion of optimum output for comparison with the estimated project cost. One
consideration in development and use of simulation and optimization models
is that programming cost increases and transparency and robustness may be
adversely affected by increasing detail and sophistication of such programs.
Fairly sophisticated models can be developed to allow greater freedom for
the determination of optimum conditions, irrespective of operation mode.
However, the degree of sophistication of the model used should be commen
surate with the objective, purposes, and scope of the investigation.
10.4
DEVELOPMENT OF MODELS
The models used for the Bay of Fundy (Canada) reassessment study (Bay of
Fundy Tidal Power Review Board and Management Committee, 1977) are
discussed here only for the purpose of illustrating the basic mathematics and
logic of the simulation and optimization. The development of these optimiza
tion models is presented in Appendix A for information. Other models may
be more appropriate under tidal characteristics different from those of the
Atlantic Ocean, particularly where the diurnal inequality is more pronounced.
However, the objective would be the same: to estimate the energy generated
at a tidal-electric plant of given characteristics by simulating its operation
under the tidal characteristics at the site.
10.4.1
Single-Tide Optimization Model
The single-tide optimization model (STOM) is based on the Swales-Wilson
model and incorporated many computationally efficient improvements. The
model is relatively simple and cost efficient for optimization purposes but
subject to certain limitations.
Its optimization logic is based upon the operation of a tidal-electric plant
during an infinite series of identical tides. Under such a regime, the operation
of the plant would be identical from one cycle to the next when operating
against a constant objective function, that is, the objective function is not time
dependent and the tidal basin levels at the beginning and end of the cycle are
equal. Hence, a single tide cycle may be considered and the operation during
that cycle optimized with the constraint that the basin level at the beginning of
the cycle is identical to the level at the end. By examining a number of single
tidal cycles of different ranges and optimizing the output for each separately, a
close estimate of annual energy production can be derived using a tidal range
histogram for the site. STOM also provides for internal storage of the step
by-step operational details that may be used to prepare hour-by-hour energy
output estimates for use in studies of system integration. The model is very
fast computationally. The model developed for the Mersey Barrage energy
studies is similarly based (Potts and Wilson, 1992).
122
10.4.2
OPTIMIZATION OF PLANT OUTPUT
One-Cycle Dynamic Programming Model
Since in reality consecutive tides are not equal in amplitude, and in order
to account for more complex operating objectives than maximum energy
production such as, for example, the variation in the value of energy to the
electrical system, a more general model, one-cycle dynamic programming
model (OCDPM), was developed for the Fundy tidal power reassessment
studies to simulate operation during the natural series of tides so as to define
an optimum operation.
Dynamic programming models produce an optimum solution over a finite
series of tides. Actual operation of a tidal plant must consider an infinite
number of tides. However, use can be made of the quasi-periodicity of natural
tides and the fact that an operation during a certain tide influences conditions
of operation only during the next 2 or 3 cycles. Similar tide elevations are,
as a close approximation, reproduced after a lunar month of 54 tides, and a
computation over about 60 tides is therefore sufficient so long as the last few
cycles are subsequently discarded to account for the fact that, for those cycles,
optimization does not properly account for the future operation.
10.4.3
Operation for Maximum Energy
Since the OCDPM is significantly more complex to use than STOM, the
former can be used to adjust or provide correction factors to the results from
STOM. It is worthwhile to note that, since the Bay of Fundy tides are quite
regular with only a small diurnal inequality, STOM produced energy estimates
over a sequence of Fundy tides that were well within 1 percent of the total
predicted on the basis of OCDPM. On the other hand, OCDPM can be used
to optimize energy production according to an energy value function that
varies with the time of day.
Single-effect operation (without pumping) for maximum energy requires:
(1) the maintenance of the tidal basin level as high as possible at the beginning
of the tidal cycle; (2) holding that level until the appropriate head becomes
available for turbine operation; (3) turbining until the head approaches the
minimum operating head for the turbines; and (4) refilling as rapidly as pos
sible to bring the reservoir back to the highest level. This operation is illus
trated in Fig. 10.1.
The single-effect optimization problem using STOM reduces to one of
manipulating the starting and finishing times of the generating period while
maintaining the same basin level at the beginning and end of the tidal cycle.
The model assumes, in the first instance, that the finish of the turbining
period occurs at the minimum operating head of the turbines. At three initial
basin levels, Yl the starting head, H, of the turbining period is examined
until the head for maximum energy production is found. These computations
are made without reference to the refilling period. A function relating the
HISTORICAL DEVELOPMENT
123
starting head for maximum energy, HsEm" , to the initial reservoir level is
derived as follows:
(10.1)
where AQ, All and A2 are computed coefficients. Using this relationsnip, Yl is
adjusted until the level Y2 at the end of the refilling period is equal to YI' The
energy production from this analysis is very close to the maximun energy
production from this single tide cycle, at least for those locations where the
diurnal inequality is small.
Double-effect operation to maximize energy production requires rapid
drawdown and refilling of the reservoir at the end of each generating period
to obtain maximum heads for turbining during the subsequent half cycle.
Consequently, although energy production continues until the minimum oper
ating head of the turbines is reached, the sluices are opened before that head
is reached. This operation is illustrated in Fig. lOA. Thus, with two-way or
double-effect generation, the turbines would operate somewhat less efficiently
at any given head than for single-effect or one-way generation since optimi
zation for the latter condition must be compromised to obtain an optimum
design for double-effect operation.
The double-effect optimization problem using STOM relates to the selec
tion of the time (or head) when the sluices should be operated in addition to
the time (or head) when turbining should start in order to maximize energy
production while keeping the initial and final tidal basin levels equal. The
search for the optimum becomes somewhat more complicated than that for
single-effect operation because a change of one variable disturbs all others.
A solution to the problem can be obtained by fixing the operating head for
the sluices, Hs'm in each half cycle and to adjust simultaneously the starting
heads, H" in each half cycle to obtain maximum energy production. Then, by
Turbining
Sluicing
I'
-I
Water Level on
Seaward Side of
c
.Q
co
the Barrier
>
OJ
LiJ
Water Level in
Tidal Basin
LBL
1 luicin1
Time
Figure 10.4
I-
Turbining
Double-effect operation parameters for single-tide optimization model.
124
OPTIMIZATION OF PLANT OUTPUT
maintaining both values of H, constant, Hs1u can be adjusted in each half cycle
to obtain a higher production. Finally, H, in each half cycle is adjusted while
maintaining H, constant in the other half cycle and holding Hs1u constant in
both half cycles. This approach gave consistent results for the Fundy tidal con
ditions and also gave results in close agreement with tne OCDPM for double
effect, which is based on a more rigorous approach.
It should be noted that the energy function for a tidal cycle subjected to
double-effect operation is not unimodal as in the case of single-effect opera
tion. There will be two local maxima-one in which the operation tends to
single effect and a second in which the operation stabilizes in the double
effect mode to produce an energy value in each half cycle that would be more
nearly equal in those locations where the diurnal inequality is low. The first
maximum would be found for low and medium tide ranges and the second
for high tides.
10.4.4
Complex Modes of Operation
Other modes of operation such as single effect with pumping, double-effect
operation with and without pumping, using variable-pitch turbine blades, and
the like may be optimized with the one-cycle dynamic programming model.
Suboptimization routines can be developed (Bay of Fundy Management
Committee, 1977) for such modes of operation.
10.5
PLANT OPTIMIZATION
Because of the number and complex interrelationships of the variables
involved (refer to Fig. 10.5) in satisfying an objective function, such as mini
mizing the cost per unit of output from a tidal-electric plant, an analytical
solution to site optimization is remote if the problem is to be treated
with rigor. Therefore, the approach to optimization must be through an
iterative process or search technique that should be made as cost-effective
as possible.
Using a model, such as STOM, the best operating strategy and the output
can be computed for one complete tidal cycle at one level of installed capac
ity, one level of sluice capacity and one set of turbine characteristics. This
procedure would be repeated for other values of installed sluice capacity to
provide data for one of the sloping lines in Fig. 10.6. Data for other values
of sluice capacity are similarly obtained. A capital cost estimate for each
combination would be prepared according to a costing formula developed
in the plant design studies. From this capital cost, an estimated annual cost
would be computed to cover amortization, interim replacement, operation
and maintenance.
For each combination of turbines and sluices, a minimum of three tidal
ranges would be selected to encompass the extremes of the annual tidal
PLANT OPTIMIZATION
rInput
Headpond
Tidal
Data
Bathymetry
Range
125
Energy
- ---
Value
(Time Related)
I
-
Effective
Operating
Mode
rr=:
Head
Design
Sluice
Installed
Turbine
Variables
Capacity
Capacity
Characteristics
L-
Operating
Strategy
, d
Other
Costs
Output
Cost
r------
(Barrage, Etc.)
Value
1----
Objective
Function
Figure 10.5
Main interrelationships of variables for optimizing a tidal-electric plant.
histogram. The optimum energy output for each range could then be cal
culated using the STOM or a suitable model, and a relationship between
energy production and tidal range developed. The total annual output would
then be determined by reference to this relationship and the histogram. The
total annual output and annual cost will yield the mean annual unit cost
of energy. Curves may then be plotted for each level of installed capacity
with unit costs for varying levels of sluice capacity against the computed
annual output, and the optimum point of the curve located. Repetition of
this sequence for other levels of installed capacity would provide further
optimum points. A curve indicating the best combination of sluices and
turbines for a particular site can be drawn, as illustrated by the example in
Fig. 10.7. Adjustments to the tidal range would also have to be made for
each combination of installation to account for the effect, if any, of barrage
operation on tidal amplitude.
126
OPTIMIZATION OF PLANT OUTPUT
Turbine Data:
Alternative 15-Fixed Blades
13
Diameter
7.5 m
Speed
72 rpm
Generator Rating
38
Orifice Rating
379 m3/ s: 1 m
7.73 m
Rated Head
Blade Angle
(j)
o
Rated Discharge
MW
27S
644 m3/s
X
.L:
3:
12
c
o
t5
::J
"0
a..
>-
e>
Q)
c
W
c 11
c
-70 - Sluices
Number of Turbines
10
80
3040
3420
Installed Capacity
Figure 10.6
100
90
(MW)
3800
Annual energy production versus installation levels for ebb generation
at Site AS, Bay of Fundy with various number of sluices.
In determining the flows used for generation, turbine performance curves,
such as those illustrated in Fig. 9.8, would be utilized. As pointed out in
Chapter 9, such unitized characteristic curves must reflect all intake and draft
tube losses. Head losses through the intakes and draft tubes are not necessarily
accounted for in such unitized characteristic curves. For fixed-blade installa
tions, the units would be operated along the maximum power and limiting
power curves at all stages of the tide, resulting in a deterministic relationship
between head and discharge. Although this assumption would introduce a
slight error in the assessment of maximum production, it would amount to a
loss of only about one-half of one percent of the annual energy production
in comparison with the true optimum; an adjustment to the annual energy
production could, therefore, be made, if appropriate. Operation of variable
pitch blade turbines for maximum energy production can be simulated in
accordance with the more complex pathway through the turbine hill chart
PLANT OPTIMIZATION
127
26 ,---,---,---,
25 --------
Band
21
Acceptable Closure Criteria
20
______________________-L____________J-__________
2000
2500
3000
3500
4000
Annual Energy Generated (GWh)
pH - Powerhouse Unit
SL - Sluice Unit
Figure 10.7
Single-effect optimization for Site A8, Bay of Fundy.
determined by introducing the calculus of variations to define an optimal
(maximum energy or benefit) relationship between discharge, reservoir level,
and time during the generating period (Fig. lOA).
For basin refilling, all turbine openings should be assumed to perform as
orifices to augment the sluicing capacity. The sluices, if fully submerged ven
turis, would have a discharge coefficient of about 1.5, based on the results of
considerable hydraulic model testing. The discharge coefficients for both fixed
and variable-pitch turbine runners when operating as orifices are not known
precisely in the absence of available model test data but would probably be
of this order. In any event, the discharge through the sluiceways would be
several times larger than that through the turbines as orifices so that small
variations in the discharge coefficient for the latter would be acceptable for
prefeasibility studies.
Additional studies could be undertaken to evaluate alternative turbine
runner diameters, fixed-blade angles, and alternative generator ratings for
a given turbine. In the Bay of Fundy reassessment studies, it was found that
the cost of energy was relatively insensitive to these alternatives. Studies of
single- and double-effect operation, with and without pumping, to establish
128
OPTIMIZATION OF PLANT OUTPUT
the relative costs and energy production of such schemes and their value to
the system should be assessed (refer also to Chapter 5).
With reference to Fig. 10.7, it is interesting to note that the at-site
optimization curve tends to be relatively flat across a broad range of possible
energy outputs. Optimization curves for the Severn Estuary (Great Britain) and
Cook Inlet (Alaska, U.S.) studies showed a similar characteristic flatness.
