You are on page 1of 13

International Journal of River Basin Management

ISSN: 1571-5124 (Print) 1814-2060 (Online) Journal homepage: http://www.tandfonline.com/loi/trbm20

Optimization of reservoir operation using linear


program, case study of Riam Jerawi Reservoir,
Indonesia

Bobby Minola Ginting, Dhemi Harlan, Ahmad Taufik & Herli Ginting

To cite this article: Bobby Minola Ginting, Dhemi Harlan, Ahmad Taufik & Herli Ginting
(2017): Optimization of reservoir operation using linear program, case study of Riam
Jerawi Reservoir, Indonesia, International Journal of River Basin Management, DOI:
10.1080/15715124.2017.1298604

To link to this article: http://dx.doi.org/10.1080/15715124.2017.1298604

Accepted author version posted online: 22


Feb 2017.
Published online: 10 Mar 2017.

Submit your article to this journal

Article views: 18

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=trbm20

Download by: [182.253.7.85] Date: 21 March 2017, At: 18:17


INTL. J. RIVER BASIN MANAGEMENT, 2017
http://dx.doi.org/10.1080/15715124.2017.1298604

RESEARCH PAPER

Optimization of reservoir operation using linear program, case study of Riam Jerawi
Reservoir, Indonesia
Bobby Minola Gintinga, Dhemi Harlanb, Ahmad Taufikc and Herli Gintingd
a
Department of Civil Engineering, Parahyangan Catholic University, Bandung, Indonesia; bDepartment of Civil Engineering, Bandung Institute of
Technology, Bandung, Indonesia; cCenter for Research on New and Renewable Energy, Bandung Institute of Technology, Bandung, Indonesia;
d
Department of Physics, University of Sumatera Utara, Medan, Indonesia

ABSTRACT ARTICLE HISTORY


The pattern of water use operation is important to ensure the continuity of water supply system of a Received 1 December 2015
reservoir. This pattern, which is usually called power rule curve in a hydropower system, should be Accepted 12 February 2017
optimally obtained, so that the yielded electrical power also becomes optimal while the continuity
KEYWORDS
of reservoir storage can also be tenable. In this study, in order to satisfy the main objective of Riam Objective functions;
Jerawi Reservoir, where the energy of 6 MW should be provided every month, a linear optimization optimization model; power
model is applied for three objective functions such as maximizing total energy, maximizing rule curve; linear program
minimum energy and minimizing energy shortage. In addition, a simulation model is also
presented for comparison purpose particularly to emphasize some disadvantages of using this
model. The results show that the optimum energy can be achieved by applying this optimization
model where the continuity of reservoir volume is also satisfied. Meanwhile, the simulation model
produces the rule curve, which cannot satisfy the continuity criterion. This optimization model is
expected to be applied for other similar cases and give the optimum power rule curve for the
related stakeholders.

1 Introduction
electrical generator. Katingan regency has a long river net-
Supply scarcity and expensive price of energy are potential work located in Katingan catchment area, where the length
issues that still occur in Indonesia nowadays. Generally, vil- of main river is approximately 650 km.
lage regions particularly on isolated areas have to face these A reservoir uses its capacity (volume) and head of water to
issues. Logistic problems and limitation of trade make deliv- produce the electrical power. One of many technical factors
ery cost and also expansion of electrical services very expens- that must be concerned in designing a reservoir is the oper-
ive. Meanwhile, energy supply is a social investment which ation pattern of water where in hydropower system’s point
cannot be avoided to increase the economy growth and pros- of view it is usually called power rule curve. It is a very impor-
perity. Of course, these issues then encourage a decentraliza- tant issue because the continuity of reservoir capacity must be
tion approach to rely on local resources in satisfying domestic tenable. If the reservoir rule curve is not designed properly,
needs, for which the energy supply is an important require- there will be lack of capacity during dry season or conversely
ment to support economy and social growth. Also, low-cost too much water spilling out of the reservoir. Therefore, a good
energy for domestic needs can increase the quality of life. management of a reservoir operation is required to produce
The uses of renewable energy and sustainable resources optimum results. In a good management of a reservoir oper-
(e.g. water, sun, wind, biomass and geothermal) are therefore ation, an effective method should be applied which can opti-
expected. mize the use of river inflow for the capacity of a reservoir in
The continuity of electric supply in Indonesia is faced with order to supply various purposes, minimize water losses and
costive price due to high consumption of fuel whose cost risks such as flood problem and also reduce the environ-
increases rapidly. This fuel dependency is hard to avoid par- mental negative impacts. Achieving this effective method is
ticularly for isolated area where diesel is a main generator for a very complicated task since the roles of all stakeholders
electricity. Therefore, it is prominent to find other energy are required. Nevertheless, with regard to the roles of engin-
sources. Hydropower system is an alternative solution to eers and practitioners, an effort can be undertaken such as
solve the energy crisis of electricity particularly in isolated performing a study for optimizing a reservoir operation.
area. Katingan is a regency located in Central Kalimantan Further, the results of the study could be used by other related
province which connects the southern part of Central Kali- stakeholders such as decision-makers.
mantan province with the centre of growth in Pangkalan There are two common techniques which have mostly been
Bun and Sampit and also the northern part with the centre used to describe the analysis of a reservoir operation. A simu-
of growth in Palangkaraya, Pulang Pisau and Muarateweh. lation technique is a representation of a system under a given
Kasongan as a capital city of Katingan regency is located set of conditions (Wurbs 1993). This technique will only pro-
exactly across the road of Central Kalimantan. The industrial duce a pattern depicting a reservoir operation which is limited
development in Katingan regency is faced with energy crisis to the user-specified set of variable values. Meanwhile, an
where until now the energy supply is limited to the existing optimization technique can represent a reservoir operation

CONTACT Bobby Minola Ginting bobbyminola.g@unpar.ac.id, bobbyminola.g@gmail.com


© 2017 International Association for Hydro-Environment Engineering and Research
2 B. M. GINTING ET AL.

system by finding an optimum solution. One of the advantages In reality of course, we recognize that the optimization of a
in using the optimization technique is that all optimization reservoir operation cannot be separated from human judg-
models also ‘simulate’ the system (Wurbs 1993). While finding ment. However, in this study, we show that from another per-
the optimum solution for a set of constraints defined by users, spective, this simple technique could be used as a useful
an optimization technique will automatically perform the consideration in order to obtain the optimum power rule
simulation procedures. The advanced computer models for curve and electric power. The structure of this paper is
both simulation and optimization techniques have been estab- described as follows: In Section 2, we describe the technical
lished since many years ago where Yeh (1985) and Wurbs data of Riam Jerawi Reservoir. In Sections 3 and 4,
(1996) are some of the pioneers. we show, respectively, the optimization and simulation
Based on our experience, most practitioners in Indonesia, techniques as well as the mathematical formulation we used
who are involved in hydropower projects, prefer performing a in this study. In Section 4, we present that the use of simu-
simulation technique to an optimization one, since it is lation technique may cause some new problems particularly
admittedly that a simulation technique is easier. They only in determining some unknown variables. We show our results
need to simulate the water balance based on the values they in Section 5. In Section 6, we present our opinion to be dis-
have specified. For example, they are given a task to provide cussed about the recent optimization techniques which
the energy for a certain value, let us say × MW. After obtain- were proposed by Jordan et al. (2012) and Mower and Mir-
ing the inflow discharge (or calculating with synthetic for- anda (2013) and also we present our plans for future work.
mula typically for 10 years due to the unavailability of the Finally, we give the summary and conclusions in Section 7.
observed inflow data) and some reservoir characteristics
data such as topography and sedimentation volume which
2 Technical data of Riam Jerawi Reservoir
are used as inputs, they compute the water balance within
that period. In many cases, the maximum elevation of the The technical data of Riam Jerawi Reservoir are obtained
reservoir is limited to a certain value due to several reasons based on the design report (Laporan Akhir Proyek 2012).
either technically or not. It means that one cannot obtain The analysis in this report was mainly focused on the feasi-
more benefits from the capacity of the reservoir nor from bility study of Riam Jerawi Reservoir. The report was con-
the head elevation in order to yield more energy, even though ducted by a consultant company. The field data, which
it is possible theoretically. Therefore, if they get the negative were measured by the consultant, are topography around
water balance for the targeted energy × MW, they will just the proposed location of the dam, the inflow discharge of
say that the possible yielded energy should be less than × the river for short period and some sediment samples both
MW. Even for a multi-purpose reservoir, for example, for suspended and bed loads. Meanwhile, the others are second-
hydropower and irrigation purposes, they would allocate ary data which are not directly measured by the consultant,
less amount of water for the irrigation since they focus such as the 90 m Digital Elevation Model, rainfall and clima-
more on the hydropower purpose, although actually they tology data for long period (10 years), land use maps, etc. We
could obtain better results. It is because, in simulation tech- do not perform any computations, for example, inflow dis-
nique, they can only adjust the targeted value in order to charge or sediment calculations since we only focus on the
achieve the positive water balance. In general, it is like an optimization technique for obtaining the optimum power
iterative procedure. In the end, it could not be ensured rule curve. Therefore, we just use these data in our compu-
whether the results have been optimum or not, even though tation. However, in this section, we explain them briefly
the positive water balance is achieved. In a few cases, after such as catchment area, inflow discharge, relationship curve
the reservoirs have been operating for 2 or 3 years, we often between elevation-inundation area-storage capacity and sedi-
face some problems such as lack of water during dry season mentation volume.
or too much water spilling out of the reservoirs. Our main
intention emerged from this problem. We would like to
2.1 Catchment area
show that an optimum result for a reservoir operation
could be achieved in a straightforward way by using an The Riam Jerawi Reservoir will be located at coordinate
optimization technique. One does not require performing 731,286, 9,912,314 with the catchment area approximately
an iterative procedure for a simulation technique anymore of 694.26 km2. The 90 m Digital Elevation Model is used as
in order to obtain a rule curve of a reservoir. By performing topographical data and the delineation of catchment area is
an optimization technique, it can also be ensured in the end performed using Watershed Modeling System 8.1 as shown
that no other possibilities one can do to achieve better results, in Figure 1. The steps of delineating the catchment area are
unless by changing the objective function. not discussed here. Therefore, interested readers are referred
There are three types of optimization technique which are to Aquaveo (2008).
most commonly used such as linear, non-linear and dynamic
programming. In this study, a linear program is used. Instead
2.2 Inflow discharge
of performing more complex technique like non-linear or
dynamic programming, we would like to emphasize more Based on the design report (Laporan Akhir Proyek 2012), due
on linear programming step by step and reveal how the linear to the unavailability of the measured discharge data, the
programming can reach an optimum solution. In this study, inflow discharge was computed using synthetic formula
we write the codes in FORTRAN. We would like also to (rainfall-runoff procedure). The further explanation of
describe some advantages of using an optimization model some rainfall-runoff models could be read in Beven (2012).
in reservoir operation rather than a simulation model. We In the design report, NRECA model was used. NRECA is a
take the case study of Riam Jerawi Reservoir which is used water balance model which was developed by Norman Craw-
mainly for hydropower use. ford and Steven Thurin in 1981 (Crawford and Thurin 1981).
INTERNATIONAL JOURNAL OF RIVER BASIN MANAGEMENT 3

Figure 1. Catchment area of Riam Jerawi Reservoir (total area = 694.26 km2).

It is a simple rainfall-runoff model based on monthly rainfall 33 m3/s. In this study, the optimization technique will be
data. This model can be divided into two parts such as direct focused only on the discharge with probability of 50%.
runoff and base flow computations. The total values of both
parts are the inflow discharges. The scheme is depicted in
Figure 2. 2.3 Storage characteristics
The main input parameters of this model are rainfall, cli- As previously mentioned, the storage characteristics such as
matology and land use coefficients. The rainfall and climatol- storage capacity and inundation area were obtained from a
ogy data were collected from some gauging stations which are field survey. These data are required to compute the evapor-
located inside the catchment area, for example, monthly rain- ation losses on the reservoir which is the function of water
fall and climatology data such as temperature, sun radiation, surface area. In the design report (Laporan Akhir Proyek
humidity and average wind speed were collected from 1998 to 2012), the evaporation was set to 4 mm/day. As shown in
2007. The land use maps show that almost 70% of the catch- Figure 4, the lowest and highest contours range respectively
ment areas are forests. Since the values of the monthly evapo- from +105 m to +225 m with the maximum volume is
transpiration are required in the NRECA model, they were approximately 816 MCM. In reality, the bottom elevation
computed using the Penman Modified Method, which was of the reservoir will always change due to the sedimentation
developed by Howard Penman in 1948 (Penman 1948, Oliver problem. The sediment will be trapped at the toe of the
2012). These values range from 5.14 to 6.38 mm/day. After dam and accumulate to a certain level during a certain period.
obtaining the monthly evapotranspiration, the inflow dis- In the report, the elevation of the crest of the spillway was set
charges for every month during 10 years were obtained. to +185 m which has the volume of 260 MCM. The dead sto-
Based on these values, the monthly inflow discharges for rage volume for 100 years was predicted to 1.27 MCM which
probabilities of 20%, 50% and 80% could be obtained as is located at the elevation of +108 m. The values of +185 m
shown in Table 1 and Figure 3. The values range from 7 to and +108 m are used respectively for the maximum and mini-
mum boundary conditions of the elevation in both optimiz-
ation and simulation models.

3 Optimization model
In order to ensure the continuity of the volume of a reservoir,
an optimum power rule curve should be obtained by using an
optimization model. This has some boundary conditions,
which has to be satisfied. In some cases, the boundary con-
dition is usually called constraint. The continuity aspect of
a reservoir is based on the equilibrium of inflow and outflow.
The inflow consists of river inflow and the outflow consists of
water release, evaporation, seepage, infiltration and other
Figure 2. Scheme of NRECA model (after Laporan Akhir Proyek 2012). hydrologic processes which decrease the volume of water.
4 B. M. GINTING ET AL.

Table 1. Inflow discharge in m3/s


Prob. (%) January February March April May June July August September October November December
20 29.62 31.53 25.74 26.85 17.43 21.21 14.59 12.31 11.96 16.39 33.47 18.92
50 14.71 14.48 23.76 23.87 14.44 11.81 10.25 9.76 9.90 11.87 12.60 13.16
80 11.63 8.90 12.97 14.63 10.88 10.03 8.91 8.30 8.07 7.67 7.31 7.75

An optimization technique is usually related to some math- efficiency when using annual flow duration curve, adjustment
ematical expressions which represent an objective function and of efficiency due to tailwater fluctuations and adjustment in
some constraints as a function of decision variables (Wurbs considering the unscheduled down time, Q is inlet discharge
1996). The constraints are mass balance, storage characteristics (m3/s) and Hnet is net head (m). In this study, the average tail-
including the maximum and minimum volume, water release water elevation and total head loss are taken based on Laporan
based on the objective, water losses due to evaporation, seepage Akhir Proyek (2012). The average tailwater elevation was pre-
and also other criteria such as maintenance flow, the maxi- dicted approximately to +115 m, even though in fact it will
mum and minimum capacity of turbines, etc. The objective always vary. The head losses due to inlet, friction, expansion
function is a mathematical formulation which defines a main and outlet were taken into account based on the design layout.
objective. Based on Wurbs (1993) in general, some objective Also, in fact the total value of head loss will always vary since it
functions with regard to the reservoir operation study can be depends on the water surface elevation in the reservoir and the
categorized into three groups as follows: tailwater elevation. However, for simplification, the total head
loss was estimated approximately to 5 m after considering all
a. Economic benefits and costs head losses. The values of ηN, η1, η2 and η3 are set, respectively,
b. Water availability and reliability to 0.80, 0.95, 0.97 and 0.97, ρ and g are set, respectively, to
c. Hydroelectric power generation 1000 kg/m3 and 9.81 m/s2.
The algorithm used in this optimization technique is
and with regard to the hydroelectric power generation, there sketched in Figure 5. Subscript t in Figure 5 defines the
are some common objective functions such as: month where in this case t ranges from 1 to 12 (January to
December). River inflow (It) is given in Table 1. Reservoir sto-
a. maximizing firm energy rage (St) is the unknown variable that should optimally be
b. maximizing average annual energy obtained where the continuity criterion is only satisfied when
c. minimizing energy shortages S12 ≥ S1 . Reservoir release (rt) includes all amount of water
d. maximizing the potential energy of water stored in the which are released out of the reservoir such as the water
system released from the reservoir to the hydropower system (ht)
and the water spilling out of the reservoir. In a hydropower sys-
In this study, since the Riam Jerawi Reservoir will be used tem, the water is only used to rotate the turbine. Therefore, if
mainly for hydropower purpose, the objective function is the water is not used for other purposes or the evaporation is
then taken with regard to the hydroelectric power generation. not taken into account, it can be stated that theoretically there
Therefore, we take three objective functions such as maximiz- is no loss of water, since all amount of water will flow again
ing total annual energy, maximizing the minimum energy back into the river. The diversion release (dmt) usually includes
and minimizing the energy shortage. the irrigation and drinking water requirement, where in this
case the value is zero. The maintenance flow (mft) is the mini-
mum amount of water which is required for the river mainten-
3.1 Mathematical formulation
ance. It is obvious that now there are three unknown variables
As previously mentioned, in this section, we will explain the which must be optimized such as reservoir storage (St), reser-
formulations of the three objective functions that we have cho- voir release (rt) and hydropower release (ht). In the Section
sen. The energy produced by a hydropower is computed as: 3.1.1–3.1.3, the objective functions will be described.
E = hN h1 h2 h3 Q r g Hnet , (1)
3.1.1 Objective function: maximizing total energy
where E is energy (Watt), ηN, η1, η2 and η3 are, respectively, This objective function aims at maximizing the total annual
water to wire efficiency on energy output, adjustment of energy. It is written mathematically as:
 

12
f (Qt=1,...,12 , Hnet , St=1,...,12 ) = max Et , (2)
t=1

where Et is the energy.

3.1.2 Objective function: maximizing minimum energy


This objective function aims at maximizing the minimum
energy which occurs in a year. The maximum energy could
be obtained during the wet season and the minimum energy
might occur during the dry season. Sometimes it would be
hard to predict exactly when the dry season occurs. Therefore,
in this objective function, the minimum energy will be
Figure 3. Inflow discharge. searched within the period and then optimized with an
INTERNATIONAL JOURNAL OF RIVER BASIN MANAGEMENT 5

Figure 4. Elevation versus storage capacity and inundation area.

expectation that the maximum energy will also be optimal. (5) is slightly changed into Equation (6).
This objective function is written as:
Min[c1 x1 + c2 x2 + c3 x3 + · · · + cn xn ]
f (Qt=1,...,12 , Hnet , St=1,...,12 ) = max[min(Et=1 , . . . , E12 )]. (3)
= Max − [c1 x1 + c2 x2 + c3 x3 + · · · + cn xn ]. (6)

Based on Figure 5, now we formulate the constraints for


3.1.3 Objective function: minimizing energy shortage our case as:
This objective function aims at minimizing the difference
⎡ ⎤
between the yielded energy and the energy demand, which St − St−1 + rt = It − et
is written mathematically as: ⎢ St ≤ Vmax ⎥
⎢ ⎥
⎢ S ≥ V ⎥
f (Qt=1,...,12 , Hnet , St=1,...,12 ) = max[min(Edemand − Et=1 ,..., 12 )], ⎢ t min ⎥
⎢ r − h ≥ 0 ⎥ for t = 1, 2, 3, . . . , 12. (7)
⎢ t t ⎥
(4) ⎢ rt + LIt ≥ dmt + mft ⎥
⎢ ⎥
⎣ ht ≤ qmax ⎦
where Edemand is the specified energy demand. Based on
Laporan Akhir Proyek (2012), the Riam Jerawi Reservoir is ht ≥ qmin
expected to produce the monthly energy in the range of 5–
The linear program shown in Equation (5) is only for
6 MW. Therefore, in this study, Edemand is set to 6 MW.
equality constraints. Therefore, it cannot directly be used to
arrange the constraints in Equation (7) since the non-equality
appears. Equation (7) must be converted by adding some
3.2 Linear program model slack variables zi. These variables can be either non-negative
A linear program usually consists of an objective function and or non-negative surplus variables which are shown in
some constrains. The standard form of a linear program is Equations (8) and (9), respectively.
written as follows: ⎡ ⎤
ai1 x1 + ai2 x2 + ai3 x3 + . . . + ain xn ≤ bi
Max[c1 x1 + c2 x2 + c3 x3 + · · · + cn xn ] ⎣ convert into ⎦,
⎡ ⎤ ai1 x1 + ai2 x2 + ai3 x3 + . . . + ain xn + zi = bi where zi ≥ 0
a11 x1 + a12 x2 + a13 x3 + · · · + a1n xn = b1
⎢ a x + a x + a x + ··· + a x = b ⎥ (8)
⎢ 21 1 22 2 23 3 2n n 2 ⎥
⎢ ⎥ ⎡ ⎤

Subject to ⎢ ··· ⎥.
⎥ ai1 x1 + ai2 x2 + ai3 x3 + . . . + ain xn ≥ bi
⎢ ⎥ ⎣ convert into ⎦.
⎣ am1 x1 + am2 x2 + am3 x3 + · · · + amn xn = bm ⎦
x1 ≥ 0, x2 ≥ 0, · · · , xn ≥ 0 ai1 x1 + ai2 x2 + ai3 x3 + . . . + ain xn − zi = bi where zi ≥ 0
(5) (9)

It is shown that the objective function in Equation (5) is to Sometimes, it is usually found that some variables are not
maximize the summation of some variables. All constraints restricted in sign or in other words the value can be either
and variables are equalities and non-negative, respectively. positive or negative. For this condition, a different manner
If the objective function is to minimize the variable, Equation is applied into Equation (5) by replacing x1 with (x1′ −x1′′ ).
6 B. M. GINTING ET AL.

Figure 5. Schematic of reservoir operation and the parameters.

This is shown in Equation (10). 3.3 Solution of linear program model


Max [c1 (x′ 1 − x′′ 1 ) + c2 x2 + c3 x3 + . . . + cn xn ] In this study, the Simplex Method is used to solve the linear
Subject to program. In this method, there are some criteria that must be
⎡ ⎤ satisfied in order to obtain a feasible and convergent solution.
a11 (x′ 1 − x′′ 1 ) + a12 x2 + a13 x3 + . . . + a1n xn = b1 It is obvious that the well-known procedure such as the
⎢ a (x′ − x′′ ) + a x + a x + . . . + a x = b ⎥ Gauss–Jordan method is a very powerful technique in solving
⎢ 21 1 1 22 2 23 3 2n n 2 ⎥
⎢ ⎥ some sets of matrix, but since some unknown variables exist
⎢ . . . ⎥.
⎢ ⎥ in Equation (12), it is difficult to choose a pivot variable.
⎢ ′ ′′ ⎥
⎣ am1 (x 1 − x 1 ) + am2 x2 + am3 x3 + . . . + amn xn = bm ⎦ Choosing the right pivot variable is really important in
x1′ ≥ 0, x1 ≥ 0, x2 ≥ 0, . . . , xn ≥ 0 order to perform an optimal computation. Therefore, in
(10) this case, the Gauss–Jordan procedure cannot directly be
applied to solve Equation (12). With regard to the Simplex
Applying equations (5), (6), (8) and (9) into Equation (7), Method, Equation (5) is changed into another form as
the Equation (11) is yielded. shown in Equation (13).
⎡ ⎤
St − St−1 + rt = It − et
⎢ ⎥ Row 0 F − c1 x1 − c2 x2 − c3 x3 − · · · − cn xn = 0
⎢ St + z1t = Vmax ⎥
⎢ ⎥ Row 1 a11 x1 + a12 x2 + a13 x3 + · · · + a1n xn = b1
⎢ St − z2t = Vmin ⎥
⎢ ⎥
⎢ ⎥ Row 2 a21 x1 + a22 x2 + a23 x3 + · · · + a2n xn = b2
⎢ rt − ht − z3t = 0 ⎥ for t = 1, 2, 3, . . . , 12 (13)
⎢ ⎥ ···
⎢ rt + LIt − z4t = dmt + mft ⎥
⎢ ⎥
⎢ ⎥ Row m am1 x1 + am2 x2 + am3 x3 + · · · + amn xn = bm
⎣ ht + z5t = qmax ⎦
Row (m + 1) x1 ≥ 0, x2 ≥ 0, . . . , xn ≥ 0.
ht − z6t = qmin
St ≥ 0, rt ≥ 0, ht ≥ 0, z1t ≥ 0, z2t ≥ 0, The objective function is now changed into an equivalent
z3t ≥ 0, z4t ≥ 0, z5t ≥ 0, z6t ≥ 0. (11) form and it is called Row 0. With regard to Equations (8) and
(9), all constraints should also be changed into other equival-
The last part of Equation (11) shows that all variables for ent forms. For some conditions, where all variables have the
the optimization model must be non-negative. Of course it is non-negative coefficients, it could be stated that the current
not required to apply Equation (10), since all variables such as basic solution is optimal. Otherwise, a variable xi with a nega-
storage, hydropower release and reservoir release are tive coefficient in Row 0 should be chosen. Let us now con-
restricted in sign where the values cannot be negative. For sider that all coefficients c1, c2, c3, … , cn in Row 0 have the
simplification, the maintenance flow (mft) is set to 5% of negative values and choose the variable x1 as a basis. This
monthly inflow. Therefore, Equation (11) is slightly changed variable is then called entering variable. After choosing the
into Equation (12). variable as a basis, a pivot Row should be determined. In
⎡ ⎤ this case, we choose Row 1 as the pivot. It should be noted
St − St−1 + rt = It − et
⎢ ⎥ that all variables in Row 0 could freely be chosen as the
⎢ St + z1t = Vmax ⎥ basis since they have negative values. Since now Row 1 is cho-
⎢ ⎥
⎢ S − z = V ⎥ sen as the pivot, Equation (13) changes into Equation (14).
⎢ t 2t min ⎥
⎢ ⎥
⎢ rt − ht − z3t = 0 ⎥ for t = 1, 2, 3, . . . , 12
⎢ ⎥ a12 c2 a13 c3
⎢ rt + LIt − z4t = dmt + 0.05 It ⎥ Row 0 F + − x2 + −
⎢ ⎥ x3
⎢ ⎥ a11 c1 a11 c1
⎣ ht + z5t = qmax ⎦
ht − z6t = qmin a1n cn b1
+ ... + − xn =
a11 c1 a11
St ≥ 0, rt ≥ 0, ht ≥ 0, z1t ≥ 0, z2t ≥ 0, z3t ≥ 0, z4t ≥ 0,
a12 a13 a1n b1
z5t ≥ 0, z6t ≥ 0. (12) Row 1 x1 + x2 + x3 + . . . + xn =
a11 a11 a11 a11
INTERNATIONAL JOURNAL OF RIVER BASIN MANAGEMENT 7

Table 2. Value of variables


a22 a12 a23 a13
Row 2 − x2 + − x3 Inflow prob. of Maintenance Local Diversion
a21 a11 a21 a11 Month 50% flow inflow release
T It MFt Lit dmt
a2n a1n b2 b1 January 39.39 1.97 0.00 0.00
+ ... + − xn = − February 35.03 1.75 0.00 0.00
a21 a11 a21 a11
March 63.63 3.18 0.00 0.00
... April 61.88 3.09 0.00 0.00
May 38.68 1.93 0.00 0.00
(14)
June 30.60 1.53 0.00 0.00
am2 a12 am3 a13
Row m − x2 + − x3 July 27.46 1.37 0.00 0.00
am1 a11 am1 a11 August 26.15 1.31 0.00 0.00
September 25.66 1.28 0.00 0.00
amn a1n bm b1 October 31.80 1.59 0.00 0.00
+ ... + − xn = −
am1 a11 am1 a11 November 32.67 1.63 0.00 0.00
December 35.25 1.76 0.00 0.00
Row (m + 1) x1 ≥ 0, x2 ≥ 0, . . . , xn ≥ 0. Vmax 260
Vmin 1.27
Note: All units in MCM (million cubic metre).

Now, a new problem arises due to the difficulty in choos-


ing the effective variable for pivot. The wrong choice may lead written as:
to an infeasible basic solution, for example, if Row 2 is ⎡ ⎤
12
selected as a pivot and the solution is infeasible, the value
⎢ y − h N h 1 h 2 h3 r g (ht Hnet t ) = 0 ⎥
of x1, x2, x3 or xn could be negative which does not satisfy ⎢ i=1 ⎥
⎢ St − St−1 + rt = It − et ⎥
the criteria in Row (m + 1). Therefore, we apply a simple ⎢ ⎥
⎢ St + z1t = Vmax ⎥
method to determine the proper variable for pivot by com- ⎢ ⎥
⎢ St − z2t = Vmin ⎥ for t
paring the ratio of Right Hand Side (RHS) with Entering Vari- ⎢ ⎥
⎢ rt − ht − z3t = 0 ⎥
able Coefficient (EVC). The value should be a minimum one. ⎢ ⎥
⎢ rt + LIt − z4t = dmt + 0.05 It ⎥
Neither a non-minimum ratio nor a negative pivot element ⎢ ⎥
⎣ ht + z5t = qmax ⎦
will produce a feasible solution (Oliver 2012). With regard
to Equation (13) and choosing variable x1 as a basis, the ht − z6t = qmin
ratio of RHS/EVC for Row i = 1, … , m are b1 /a11 , b2 /a21 , = 1, 2, 3, . . . , 12 (16)
… , bm /am1 respectively. After knowing the minimum ratio,
the Gauss–Jordan procedure is performed. This will yield a Equation (16) shows that only hydropower release ht
new basic solution, where a looping procedure is performed (not It) is used for rotating the turbine. The known values
in Equation (14) until the optimal solution is reached. It of the variables in Equation (16) are written now in
should also be noted, that some special conditions in a linear Table 2.
program such as alternate optimal solutions, degeneracy, The unknown variables qmax and qmin are the maxi-
unboudedness and infeasibility may occur. However, we do mum and minimum turbine capacity, respectively. In
not discuss them further in this study. Interested readers this study, qmax is set to have unrestricted value for not
are referred to Reeb and Leavengood (1998), Albright et al. detaining the turbine capacity to reach the maximum
(2011) and Source Material (2014). value. For the sake of simplicity, qmax is set to be a func-
Let us now combine the objective function in Equation (2) tion of qmin written in Equation (17), where qmin is set to
and Equation (12) into another form as Equation (15). 14 m3/s.
qmax = 1.5 qmin . (17)
⎡ ⎤

12

⎢ y− Et = 0 ⎥
⎢ i=1 ⎥
⎢ S −S +r =I −e ⎥
⎢ t t−1 t t t ⎥
⎢ St + z1t = Vmax ⎥
⎢ ⎥ 4 Simulation model
⎢ St − z2t = Vmin ⎥ for t
⎢ ⎥
⎢ rt − ht − z3t = 0 ⎥ A simulation model has a relatively simpler procedure than
⎢ ⎥
⎢ r + LI − z = dm + 0.05 I ⎥ an optimization one, where the key point of performing a
⎢ t t 4t t t⎥
⎣ h +z =q ⎦ simulation model is to ensure the continuity of the reser-
t 5t max
ht − z6t = qmin voir storage. As previously mentioned, in a simulation
model, the reliability of a reservoir could be investigated
= 1, 2, 3, . . . , 12. (15) only by observing the parameters such as storage level,
reservoir release and the yielded energy at each time step
whether satisfying the continuity criterion or not. In gen-
It is shown that now Equation (15) has a form similar eral, the mathematical formulation for a simulation
to Equation (13), where the coefficient of the objective model is similar to Equation (7), but the reservoir release
function in Row 0 has the negative value. With regard (rt) in the first line of Equation (7) now becomes hydro-
to the Simplex Method, the variable Et is called non- power release (ht), since the value of reservoir release (rt)
basic variable and the others are called basic variable cannot explicitly be determined in a simulation model.
(except for variable y in the objective function). By apply- Therefore, the constraint in the fourth line of Equation
ing Equation (1) into Equation (15), a new equation is (7) vanishes. The new equation for a simulation model is
8 B. M. GINTING ET AL.

written as: system. This is a main disadvantage of a simulation model.


⎡ ⎤ It is shown that in order to solve Equation (18), the variable
St − St−1 + ht = It − et − spt ht should explicitly be determined by users. Afterwards, the
⎢ St ≤ Vmax ⎥
⎢ ⎥ reliability of this value is reviewed whether satisfying all con-
⎢ St ≥ Vmin ⎥
⎢ ⎥ for t strains in Equation (18) or not. Therefore, a simulation model
⎢ rt + LIt ≥ dmt + 0.05It ⎥
⎢ ⎥ is simpler than an optimization one. However, in the end, we
⎣ ht ≤ qmax ⎦
do not know whether the value of ht, that we have deter-
ht ≥ qmin mined, has already been optimum for the energy or not.
= 1, 2, 3, . . . , 12. (18) Sometimes it is hard to determine the value of ht, since it
may vary every month. Therefore, for the simulation
It is shown in Equation (18) that as the consequence, there model, we specify a minimum value of the energy which
might be overflow water (spt) over the spillway if the amount should be satisfied by the hydropower system, in this case,
of water is much higher than the hydropower release, where 6 MW. Afterwards, the value of ht can be obtained from
the summation of the hydropower release and the overflow Equation (1), where Q in Equation (1) equals ht. The value
water is the reservoir release and written mathematically as: of 6 MW could be increased or decreased as long as all con-
if St−1 + It − (ht + et ) ≥ Vmax then straints in Equation (18) are satisfied. Also, the initial value of
spt = St−1 + It − (ht + et ) − Vmax St should be determined. The flow chart of the simulation
else model is given in Figure 6.
. (19)
spt = 0
end if
where rt = ht + spt 5 Results and analysis
It might not be a problem for the downstream part of a For both optimization and simulation models, the compu-
reservoir when the value of spt is high since there is no loss tation is performed for 13 months starting from January
of water in order to satisfy the first and fourth line of and ending in January for the next year. This aims at ensuring
Equation (18). However, with regard to the efficiency of the the continuity aspect where volume of the reservoir in Janu-
hydropower system, a problem may arise. The higher value ary of the next year should be greater than or equal to the
of spt causes the higher inefficiency for the hydropower volume in January of the previous year. Let us now define

Figure 6. Flow chart of simulation technique.


INTERNATIONAL JOURNAL OF RIVER BASIN MANAGEMENT 9

variables for each St, rt and ht. When the value of St for a cer-
tain month has been determined, the value of water surface
elevation and inundation area can be interpolated from
Figure 4. The value of water surface elevation is required to
compute both the value of Et and inundation area in order
to compute the value of et for the next month. Therefore, a
looping procedure, which requires a correct pivot, is required.
Otherwise, the feasible or the optimal objective function will
never be reached using the Simplex Method. In the simu-
lation model, the unknown variable is only St, since ht can
be defined as we previously mentioned and rt is computed
with Equation (19).
Figure 7 and Table 3 show, respectively, the rule curves
and the summary of the optimization model for the three
objective functions. It is shown that the lowest operated
water surface elevation for the objective function of maximiz-
ing minimum energy (OF-MME) is +177.52 m. This value
does not differ significantly with the objective function of
minimizing energy shortage (OF-MES) which is +177.53 m.
Both of these values occur in February. Meanwhile, the lowest
operated water surface elevation for the objective function of
maximizing total energy (OF-MTE) is +178.95 m which
occurs in December. All the objective functions reach the
maximum operated water surface elevation at +185 m in
May. The operated water surface elevations of both OF-
MME and OF-MES show a similar characteristic, whereas
the operated water surface elevations of OF-MTE are differ-
ent and never below the others.
With regard to hydropower release, from January to March
OF-MME gives the higher values than the others with the
maximum difference approximately of 2 MCM. Interestingly,
from March to June OF-MTE gives the higher values with
the maximum difference approximately of 11 MCM, where
both OF-MME and OF-MES keep producing constant values.
From June to December, the constant hydropower releases are
given by OF-MTE, whereas the values of the others keep
Figure 7. Summary of the results of optimization model for discharge with prob-
increasing. The maximum difference in this period is given
ability of 50% (a) the monthly operated water surface elevation at reservoir, (b) by OF-MES with the maximum difference approximately of
the monthly hydropower release (c) the monthly yielded energy. 6 MCM. A similar characteristic is shown for the yielded
energy. From January to March, OF-MME produces the rela-
the unknown variables for the optimization model to be St, rt tively constant and higher energy than the others with the
and ht. Since the computation is performed for 13 months, maximum difference approximately of 0.43 MW (289 MW-
the unknown variables now become 39, consisting of 13 hour). From March to June, OF-MTE produces the highest

Table 3. Summary of the results of optimization model for discharge with probability of 50%
Water elevation at reservoir Hydropower release
Storage (MCM) (+m) (MCM) Yielded energy (MW) Yielded energy (MW – Hour)
OF- OF- OF- OF- OF- OF- OF- OF- OF- OF- OF- OF- OF- OF- OF-
Month MTE MME MES MTE MME MES MTE MME MES MTE MME MES MTE MME MES
January 219.08 209.28 206.81 179.37 178.04 177.71 36.29 38.53 38.95 5.66 5.87 5.90 4,208 4,367 4,389
February 217.83 205.44 205.56 179.20 177.52 177.53 36.29 38.87 36.29 6.24 6.50 6.07 4,196 4,367 4,077
March 245.17 232.44 232.44 182.92 181.19 181.19 36.29 36.63 36.75 5.99 5.88 5.90 4,459 4,377 4,391
April 260.43 258.04 258.04 185.00 184.67 184.67 46.63 36.29 36.29 8.22 6.36 6.36 5,918 4,583 4,583
May 260.43 260.43 260.43 185.00 185.00 185.00 38.68 36.29 36.29 6.60 6.19 6.19 4,909 4,606 4,606
June 254.74 254.74 254.74 184.23 184.23 184.23 36.29 36.29 36.29 6.32 6.32 6.32 4,551 4,551 4,551
July 245.90 245.90 245.90 183.02 183.02 183.02 36.29 36.29 36.29 6.00 6.00 6.00 4,466 4,466 4,466
August 235.76 235.76 235.53 181.64 181.64 181.61 36.29 36.29 36.52 5.87 5.87 5.91 4,368 4,368 4,394
September 225.13 224.19 224.90 180.20 180.07 180.17 36.29 37.23 36.29 5.92 6.06 5.92 4,266 4,367 4,264
October 220.64 218.26 218.82 179.59 179.26 179.34 36.29 37.73 37.88 5.68 5.87 5.90 4,223 4,367 4,389
November 217.02 212.10 212.84 179.09 178.43 178.53 36.29 38.83 38.64 5.82 6.15 6.13 4,188 4,430 4,417
December 215.98 208.42 206.37 178.95 177.92 177.65 36.29 38.93 41.72 5.62 5.92 6.31 4,178 4,404 4,697
January 219.08 209.28 206.81 179.37 178.04 177.71 36.29 38.53 38.95 5.66 5.87 5.90 4,208 4,367 4,389
Min 215.98 205.44 205.56 178.95 177.52 177.53 36.29 36.29 36.29 5.62 5.87 5.90 4,178 4,367 4,077
Max 260.43 260.43 260.43 185.00 185.00 185.00 46.63 38.93 41.72 8.22 6.50 6.36 5,918 4,606 4,697
Total 53,927 53,251 53,223
Note: All units in MCM (million cubic metre).
10 B. M. GINTING ET AL.

Figure 8. The monthly yielded energy of the optimization model compared to the energy demand.

energy of 8.22 MW (5918 MW-hour) among the others. The model, the unknown variable St is computed based on the
maximum difference is 1.85 MW (1335 MW-hour). All objec- value of the previous step t-1. Therefore, the initial value of
tive functions produce relatively similar energy from June to St should be known first. Actually, this initial value could
August. However, from August to December, the energy of be obtained iteratively or from the actual reservoir release.
OF-MTE decreases. OF-MME produces the constant energy, For the sake of simplicity and since the actual reservoir release
while the energy of OF-MES increases. In this period, the is unavailable, we set the initial value of St for the simulation
maximum difference is 0.70 MW (519 MW-hour). model to 219.08 MCM (at the water surface elevation of
OF-MTE, OF-MME and OF-MES give the total energy in +179.37 m) which is similar to the largest storage value of
a year, respectively, of 53,927, 53,251 and 53,223 MW-hour. the three objective functions of the optimization model in
These values seem relatively similar, where for the total January. Variable ht in simulation model is also unknown,
energy OF-MTE should be the best option. However, back but this value can be determined either constantly or itera-
to the main objective that the monthly energy demand of 6 tively for the optimum result. It should be noted that it
MW should be satisfied, now OF-MTE is not the best option would be hard to determine this value iteratively, since
anymore. As shown in Figure 8, in January, OF-MTE cannot there will be too many possibilities. For the sake of simplicity
satisfy the energy demand with the energy shortage of 0.34 and with regard to the main objective, variable ht is set to the
MW. Also, from August to December, OF-MTE produces value which always produces the energy equal to 6 MW.
the energy lower than 6 MW, with the maximum shortage Another reason is that we can know whether the continuity
of 0.38 MW in December. Both OF-MME and OF-MES can- criterion will be satisfied or not for this value. We present
not satisfy the energy demand in January, where the energy the comparison between the result of simulation model and
shortages are, respectively, 0.13 and 0.10 MW, but these OF-MES in Figure 9.
values are lower than OF-MTE. From August to December, It is shown that by setting the value of ht, which always
both OF-MME and OF-MES can relatively satisfy the energy produces the energy equal to 6 MW, the operated water sur-
demand, where OF-MES shows a better performance than face elevation in January for the next year is lower than the
OF-MME. Therefore, in this case, OF-MES is the best option elevation in January of the previous year, which means that
in order to satisfy the main objective. the continuity criterion is not satisfied. If the computation
In optimization model, the unknown variables St, rt and ht is continued, one day the water in the reservoir would be
are determined automatically using the Simplex Method. empty, since the amount of water, which is released out of
However, as we previously mentioned, for simulation the reservoir, is higher than the inflow. This problem can
be anticipated either by reducing the value of ht or determin-
ing this value iteratively, but the new problem might arise.
The first problem is that by reducing the value of ht, the
main objective of 6 MW yielded energy is not achieved.
The second one, as we previously mentioned, is that there
are too many possibilities if the value of ht is determined
iteratively. Actually, the second problem can be solved
using optimization model. This is the essential part of optim-
ization model, where when solving an optimization model,
the simulation model is also being performed simultaneously.

6 Discussion
Figure 9. Comparison of monthly operated water surface elevation at reservoir Nowadays, significant developments have been shown for
between the simulation model and OF-MES. optimization model. For most readers, the optimization
INTERNATIONAL JOURNAL OF RIVER BASIN MANAGEMENT 11

Figure 10. Concept of our future study to include risk factor in optimization model.

model presented in this study might be quite old, since the Both of these objectives were coupled implicitly leading to
more advanced optimization models such as non-linear and one objective function of minimizing the damage cost during
dynamic programming have intensively been used by other the flood event. With regard to this method, we have a differ-
researchers. Some new techniques, which are not related to ent opinion. We hypothesize that for multi-purposes reser-
optimization programming, are even used. We note the voir, for example, economic and flood protection, it would
work of Mower and Miranda (2013), in which a new tech- be better to keep the main objective function of maximizing
nique was proposed to obtain the rule curve. Instead of the energy (for the economic purpose) and include the
using a complex model, they generated the rule curve by flood risk factor as the constraint. In this approach, the math-
using the historical water level data. They only required the ematical formulation of the main objective function could be
long-term data of water level for their model, which integrate applied similarly to OF-MTE, OF-MME or OF-MES, where
many variables required by those complex models (Mower for the constraints of flood risk, the formulation could be con-
and Miranda 2013). They processed the data with the structed as a function of the damage costs. The minimum
EXPAND procedure in the Statistical Analysis System result- damage cost is set to zero and the maximum one is specified.
ing the 97.5 quantile model fit to the distribution of 60-days As shown in Figure 10, now the damage cost is defined as a
summed changes in water volume. They analysed the reser- function of (Lit + rt – dmt + mft).
voir whose main purpose was to control the flood. This Our future study will be focused on an optimization model
method is very promising and the idea behind is great, for a river from segment A to B (as shown in Figure 10). A
since the result can describe the existing rule curve, even specific location, whose risk factors are simulated, is specified
though in fact there are always some differences between pre- within this segment. Therefore, in the future, the use of two
dicted and existing rule curve. In our opinion, with regard to models, such as hydrologic and hydraulic routing models,
flood control objective, this method might be better than will be investigated. The total discharge of (Lit + rt – dmt +
other optimization models, since the procedure is relatively mft) will be treated as an input for both of these models.
easier. Also, complex mathematical formulation is not The simple hydrologic model such as Muskingum-Cunge
required to be performed due to the use of real data, which or Kinematic Wave method could be used starting from
are sufficient to describe the real hydrologic phenomenon. point A to know the discharge at point B. The more advanced
The model of Mower and Miranda (2013) has also several model such as 1D or 2D hydraulic model based on the shal-
disadvantages. With regard to flood control objective, par- low water equations could also be used in order to simulate
ticularly for large dams which are designed with the Probable the inundation area for determining the risk factor. The out-
Maximum Flood (PMF), the absence of the PMF value in the put from both of these models is water elevation, which is
collected historical data would reduce the accuracy of result. related to inundation area. Therefore, the flood risk con-
The computed rule curve might underestimate the flood risk, straint can be determined based on the inundation area.
where the existing rule curve has been designed beforehand This proposed approach can be applied not only for opti-
with the consideration of PMF. Another disadvantage, mizing energy, but also for checking the vulnerability of a rule
which was also stated in their paper, is the requirement of curve to flood problems at the downstream area of the dam.
long recorded data with keeping in mind that the good data Let us now take an example of the Riam Jerawi’s rule curve,
should always be ensured without manipulation. In our which has previously been obtained based on OF-MES. As
opinion, with regard to the objective of maximizing energy, shown in Figure 7, there is a quite high difference of water
it would be hard to obtain a proper rule curve, due to the surface elevation of the rule curve in a year, which is approxi-
presence of flood events. Even though the flood is an event mately 7.47 m. Since the time of a flood event cannot exactly
with low probability of exceedance, its presence in a long be predicted, the risk factors for the specified location (within
data measurement might cause a significant error for the segment AB) vary depending on the water elevation at the
computation. For example, let us assume that we have 10 reservoir. If a flood occurs within January and February,
years historical data, where the first and fourth years are the risk factors are lower than in May or June, since there
the flood seasons with the return periods of 5 years and 100 is a space of volume of more than 7 m in January to February
years, respectively. In the eighth year, the PMF occurs. The as a flood detention. Meanwhile, there is almost no space for
use of all data will produce overestimated results for the flood detention in May or June, since the average water sur-
optimization model with OF-MTE. Particularly in developing face elevation at the reservoir in these months is +185 m. By
countries such as Indonesia, it is also difficult to obtain the investigating this problem, some new scenarios for rule curve
long-term historical data. could be obtained, as now the rule curve is also affected by the
We also note the works of Hydrologic Engineering Center risk factor. We realize that an advanced formulation is
(1991), Draper et al. (2003) and Jordan et al. (2012), that pre- required, since the computation of both hydrology and
sented the good explanation for combining the objectives of hydraulic models will be performed several times as much
economic and flood protection in an optimization model. as the number of simulation time. However, this is a
12 B. M. GINTING ET AL.

challenge, which is very interesting to be investigated in next Disclosure statement


study. Therefore, a highly stable numerical model is required. No potential conflict of interest was reported by the authors.
With regard to the 2D hydraulic model, interested readers are
referred to Ginting (2011) and Ginting et al. (2011, 2012,
2013).

References
7 Summary and conclusion
Albright, S.C., Winston, W.L., and Zappe, C.J., 2011. Data analysis and
The optimization model for Riam Jerawi Reservoir has been decision making. 4th ed. Mason: South-Western Cengage Learning.
presented. For comparison purpose, the simulation model Aquaveo. 2008. Watershed modeling system version 8.1. Provo, UT:
Aquaveo, LLC.
has also been presented. It is shown that for achieving the Beven, K.J., 2012. Rainfall-runoff modelling (The Primer). 2nd ed.
monthly energy demand of 6 MW, some difficulties arise in Chichester: Willey-Blackwell.
the simulation model particularly for determining the proper Crawford, N.H., and Thurin, S.M., 1981. Hydrologic estimates for small
value of hydropower release. As previously shown, the con- hydroelectric projects. Washington, DC: Small Decentralized
stant value of hydropower release for the energy of 6 MW Hydropower Program, International Programs Division, National
Rural Electric Cooperative Association.
produces a rule curve which cannot satisfy the continuity cri- Draper, A.J., et al., 2003. Economic-engineering optimization for
terion. Also, setting the value of hydropower release itera- California water management. Journal of Water Resources Planning
tively causes difficulties due to many possibilities. We have and Management, 129 (3), 155–164.
also explained that setting the value of hydropower release Ginting, B.M., 2011. Two dimensional flood propagation modeling gener-
iteratively in simulation model is actually similar to perform- ated by Dam Break using finite volume method. (Master Thesis).
Bandung Institute of Technology, Indonesia.
ing an optimization model but in a more effective way. In this Ginting, B.M., et al., 2011. Pemodelan 2 Dimensi Propagasi Aliran Banjir
study, we only use the inflow discharge with the exceedance Akibat Keruntuhan Bendungan dengan Metode Volume Hingga.
probability of 50%, since we only focus on the computation Konferensi Nasional Pasca Sarjana Teknik Sipil, Institut Teknologi
of optimization model. However, a similar way can also be Bandung. ISSN: 2098-3051. (in Indonesian).
applied to the inflow discharge with the exceedance prob- Ginting, B.M., et al., 2012. Application of finite volume cell center
method with wet and dry treatment in hydrodynamic flow modeling.
ability of 20% or 80%. Proceeding of the second international conference on port, coastal, and
In this study, the optimization model is a linear program, offshore engineering, Bandung Institute of Technology, 12th–13th
where the three unknown parameters, such as storage, reser- November 2012. ISBN:978-979-96161-2-8.
voir release and hydropower release are determined auto- Ginting, B.M., Riyanto, B.A., and Ginting, H. 2013. Numerical simu-
matically. The storage is required to compute the other lation of Dam Break using finite volume method case study of Situ
Gintung. 4th International seminar of HATHI, proceedings of
variables such as water surface elevation and inundation international seminar on water related disaster solutions,
area based on Figure 4. Once the value of inundation area Yogyakarta, 6th–8th September 2013, Vol 1, 209–220. ISBN: 978-
is obtained, the value of evaporation for the subsequent 979-988-5-5-0.
time step can be computed. This requires a looping procedure Hydrologic Engineering Center. 1991. Optimization of multiple-purpose
until the objective function is solved and the feasibility is also reservoir system operations: a review of modeling and analysis
approaches. Davis, CA: U.S. Army Corps of Engineers.
satisfied. As also published in Source Material (2014), we con- Hydrologic Engineering Center. 2009. HEC-DSSVue HEC data storage
clude that in order to achieve a feasible solution, the compu- system visual utility engine, user’s manual. Davis, CA: U.S. Army
tation of Simplex Method and Gauss–Jordan procedure can Corps of Engineers.
simply be performed by following the two important steps. Jordan, F.M., Boillat, J.L., and Schleiss, A.J., 2012. Optimization of the
The first step is to choose a variable xi with a negative coeffi- flood protection effect of a hydropower multi-reservoir system.
International Journal of River Basin Management, 10 (1), 65–72.
cient in Row 0. The second one is to select the minimum Laporan Akhir Proyek. 2012. Pendalaman Kelayakan Pembangunan
value of the ratio of RHS with EVC. PLTA Riam Jerawi dan Potensi PLTA lain di Kabupaten Katingan.
The rule curve is selected based on OF-MES which gives (in Indonesian).
the total energy of 53,223 MW-hour in a year, even though Mower, E., and Miranda, L.E., 2013. Evaluating changes to reservoir rule
this value is lower than 53,927 MW-hour from OF-MTE. curves using historical water-level data. International Journal of River
Basin Management, 11 (3), 323–328.
The reason is that OF-MES produces the lowest energy short- Oliver, M., 2012. Practical guide to the simplex method of linear program-
age among the others, which is suitable for the main objective ming. Available from: http://math.jacobs-university.de/oliver/
of the 6 MW monthly energy demand. Although the optim- teaching/iub/spring2007/cps102/handouts/linear-programming.pdf
ization model presented in this study is a simple one, both [Accessed June 2015].
the advantage of using optimization model and disadvantage Penman, H.L., 1948. Natural evaporation from open water, bare soil and
grass. Proceedings of the Royal Society A: Mathematical, Physical and
of using simulation model from practical point of view have Engineering Sciences, 193, 120–145.
been well explained. Therefore, the formulation of the optim- Reeb, J., and Leavengood, S., 1998. Using the simplex method to solve lin-
ization model presented in this study could be useful for some ear programming maximization problems. Performance Excellence in
practitioners and related stakeholders in designing a reser- the Wood Products Industry, Corvallis, Oregon State University.
voir. The more advanced optimization models, such as Source Material. 2014. The simplex method. Avaliable from: http://mat.
gsia.cmu.edu/classes/QUANT/NOTES/chap7.pdf [Accessed October
non-linear and dynamic programming, are also interesting 2014].
to be investigated in next study. Wurbs, R.A., 1993. Reservoir-system simulation and optimization
models. Journal of Water Resources Planning and Management, 119
(4), 455–472.
Acknowledgment Wurbs, R.A., 1996. Modeling and analysis of reservoir system operation.
Upper Saddle River, NJ: Prentice-Hall.
The authors appreciate all anonymous reviewers, editor and Yeh, W.W-G., 1985. Reservoir managements and operation models: a
associate editor for providing many constructive comments state-of-the-art review. Water Resources Research, 21 (12), 1797–
and suggestions. 1818.

You might also like