Accepted Manuscript

Integer programming to optimize Micro-Hydro Power Plants for generic river profiles

A. Tapia, P. Millán, F. Gómez-Estern

PII: S0960-1481(18)30410-5
DOI: 10.1016/j.renene.2018.04.003
Reference: RENE 9961

To appear in: Renewable Energy

Received Date: 13 October 2017

Revised Date: 30 March 2018
Accepted Date: 2 April 2018

Please cite this article as: Tapia A, Millán P, Gómez-Estern F, Integer programming to optimize
Micro-Hydro Power Plants for generic river profiles, Renewable Energy (2018), doi: 10.1016/
j.renene.2018.04.003.

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please
note that during the production process errors may be discovered which could affect the content, and all
legal disclaimers that apply to the journal pertain.
 ACCEPTED MANUSCRIPT

Integer programming to optimize Micro-Hydro Power Plants for generic river

profiles

A. Tapiaa , P. Millána , F. Gómez-Esterna

a Universidad Loyola Andalucía, Departamento de Ingeniería, Sevilla, Spain

PT
RI
Abstract
This paper addresses the problem of designing an optimal micro-hydro power installation in rivers with
generic profiles, when micro-hydro schemes are studied. This is geared towards the application of Micro-

SC
Hydro Power Plants to supply marginal isolated areas using small Pelton wheels, where both technology
and resources are limited. For this purpose, a model of a Pelton micro-hydro plant is first developed. Sub-
sequently, a discretization of the river profile is made, on the basis of which a set of integer variables are
proposed, being the model transformed then into an integer optimization problem. Finally, the effectiveness

U
of the proposed method is showed through a specific design problem. The application of the developed
method is especially interesting when designing micro-hydro plants to provide electricity to isolated popu-
lations, where both technology and resources are limited.
AN
Keywords: integer programming, micro-hydro power plants, penstock optimization
M

1 1. Introduction
2 Hydropower is one of the oldest renewable energy sources used to generate electrical power, whereby
D

3 it has established itself as the most frequently used one around the world, providing 19% of the planet’s
4 electricity [1], and also the most efficient one [2]. By virtue of the versatility of the modern water turbines,
TE

5 this energy source is suitable not just for mass energy supply, but also for the supply of small isolated areas,
6 where due to geographic or economical issues, electricity grid access is not available [3] [4]. In this context,
7 renewable energy technologies can make an important contribution to solving the basic energy needs with a
8 minimal impact. Although several technologies, such as photovoltaic [5], wind [6] and bio-mass [7] systems,
EP

9 are succesfully used within this purpose, Micro-hydro Power Plants (MHPP) have been proved as the most
10 affordable method of dealing with energy poverty [8] [9], being considered the cheapest option for off-grid
11 pico generation [10]. Although mass manufactured propeller systems, like Pelton and Turgo wheels [8] [11]
12 and Francis turbines, represent the usual equipment in MHPP [12] [13], pumps working as turbines (PAT)
C

[14] [15], and locally made systems are also a frequent option.
13
14 Notwithstanding the above, there are some challenges posed by the typical context of these micro hydro
AC

15 installations. The limited resources, together with the lack of qualified manpower, can cause that the
16 available resources are not used in the most efficient way, in particular, due to the decisions related to the
17 design of the plant and its most relevant parameters, that are mostly based on experience and know-how. For
18 all this, the development of automatic design tools is particulary important, much more so when important
factors are usually neglected.
19
20 The emplacement of the turbine and the dam, together with the penstock layout is particularly relevant,
21 since its influence on the plant performance and cost is clear [16]. However, this problem has received little
attention in the related literature [17] [18] [19].

Email addresses: atapia@uloyola.es (A. Tapia), pmillan@uloyola.es (P. Millán), fgomezestern@uloyola.es (F.
Gómez-Estern)

Preprint submitted to Renewable Energy April 3, 2018

 ACCEPTED MANUSCRIPT

22
23 The studies aiming at solving this problem include [16], where the author presents an analysis for
24 penstock optimization. The goal of this work is to minimize the cost of the energy production, as part of
25 a further study that aims to develop a modular set of cost-effective microhydro schemes for isolated areas
26 where economics are not appropiate, further explained in [20]. In a similar way, [21] presents a dimensional
27 analysis for determining optimal flow discharge and optimal penstock diameter for impulse and reaction
28 turbines, with the aim of providing general insights for minimizing water consumption. In these studies, a

PT
29 general simplification of the river profile is made, consisting on assuming a straight line, characterised by a
30 constant slope. Although this approximation has so far given reasonable results when it concerns to decide
31 the general parameters of the penstock, it overlooks the clearly irregular topography of the terrain, thus
neglecting an important decision in the penstock layout design.

RI
32
33 This work aims to develop a tool for finding the optimal location of the main parts of the installation,
34 namely: dam, turbine and penstock. A general, irregular river profile is taking into account for determining
35 the optimal solution. It is assumed that this profile is characterized by a set of topographic references

SC
36 measured in-situ (drone-based mapping is becoming an affordable trend), so it is being taked into account
37 for determining the most efficient solution. To fulfill this objective, both a model of the plant and a
38 characterization of the problem geometry are developed. After that, a set of new variables are defined
39 in order to transform the optimization problem in an integer programming problem, for which efficient

U
algorithms are available in the literature.
40

This paper is organized as follows. In section 2, a general overview of the problem set-up and its main
41
AN
42 variables is given. In section 3, a nonlinear model of the MHPP is developed. This model is expressed in
43 terms of decision variables in section 4, where a discretization of the river profile is assumed. In section 5 an
44 integer optimization problem is defined. The proposed problem is tested for an example MHPP installation
45 in section 6, being the results and its computational complexity discussed. Finally, conclusions are resumed
M

46 in section 7.

47 2. Problem set-up
D

48 The elements that make up a typical MHPP scheme can be classified in two main groups: the piping
(where potential energy is transformed into kinetic energy) and the powerhouse (where kinetic energy is
TE

49

50 converted into electrical energy). Even though water storing is not usual in most micro-hydro schemes, a
51 small dam is frequently installed in order to provide a smooth transition to the penstock, so friction losses
52 and air entry are inadmissible. The location of these elements defines the maximum attainable energy,
according to the gross height, Hg , as shown in figure 1.
EP

53

dam
piping
powerhouse
C

L
Hg
AC

Figure 1: River profile scheme

54 2.1. Piping
55 Although it is common to install more than one pipeline for large power power plants, this paper focuses
56 on MHPP, where the installations are usually restricted to a single pipeline, due to the low power and eco-
57 nomical needs. This piping consists on a single pipeline. This is composed of a certain number of connected
2
 ACCEPTED MANUSCRIPT

58 straight lengths that obtain the water flow from the river and drive it to the powerhouse on a lower level.
59 The water intake is located at the dam, and is placed in the river upstream.
60

61 As water flows through the penstock, its potential energy is gradually transformed into pressure energy.
62 The efficiency of this conversion depends on the friction losses (length and roughness of the inner pipe wall),
63 and the local losses (pipe elbows and intake). Thus, given a certain gross head Hg between the dam and
the powerhouse, it is always convenient to minimize both the penstock length L and the number of elbows nc .

PT
64

65

66 Due to the piecewised nature of the pipe profile, a certain gap, ǫ, appears between this and the terrain
67 in the points between elbows. A positive value of ǫ indicates that a support is needed, while a negative one
indicates that an excavation is needed to deploy the pipe.

RI
68

69 2.2. Powerhouse
The term powerhouse refers to the building, normally roofed, that contains both the turbine and the

SC
70

71 generator.
72

rotor generator

U
nozzle
AN
valve
M

jet bucket

Figure 2: Pelton turbine working scheme

D

73 At the end of the penstock (see figure 2), water goes through a nozzle with section Snoz , forming the water
TE

74 jet and getting its pressure energy transformed into kinetic energy. This jet impacts on the Pelton turbine
75 buckets, making it rotate around its axis and transforming its kinetic energy into mechanical energy. The
76 efficiency of this transformation, ηt , depends on the nozzle and rotor geometry and their material. Lastly,
77 the mechanical energy of the rotor is converted into electrical energy in the generator, with an efficiency ηg .
EP

78

79 The purpose of this paper is to address the optimization of the MHPP installation, by finding the most
80 suitable location of dam and turbine, together with the pipe connections or elbows, that satisfies the following
81 constraints:
C

82 • The obtained power, P , is above a certain prescribed minimum supply value, Pmin .
AC

83 • The water flow rate to be used in the MHPP, Q, is below a given bound that depends on the flow rate
84 of the river, Qmax .
85 • Supports needed to bridge the gap, ǫ ≥ 0, where penstock is above the terrain, can not exceed a
86 maximum value, ǫsup , that represents the maximum support that ensures stability.
87 • The depth of the excavations needed, ǫ ≤ 0, where penstock rests below the ground level, can not
88 exceed a minimum value ǫexc , that represents the maximum excavation that the ground allows.

3
 ACCEPTED MANUSCRIPT

89 3. Plant model

90 In this section, a nonlinear model of the micro-hydro power plant is presented. First, the obtainable
91 power of the MHPP is characterized by means of the corresponding hydraulic model and the efficiencies of
92 the different parts of the installation [22]. After that, the expression that describres the cost of the piping
93 installation is introduced.

PT
94 3.1. Water flow and obtainable power
95 The theoretical electric power available for the MHPP, P , can be expressed in terms of the water flow Q
96 and the net head at the end of the penstock h [17], as follows

RI
97 P = ηρgQh, (1)
98 being ρ the water density, g the gravitational constant, and being η defined as a global efficiency of the
MHPP, defined by

SC
99

100 η = ηg ηt . (2)
101 Assuming an impulse turbine, the net head at the entrance of the turbine, the water head at the nozzle,

U
102 h, is entirely transformed into kinetic energy, so it can be expressed in terms of the flow, Q, as
1
h= Q2 . (3)
AN
103
2
2gSnoz
104 The difference between the gross head, Hg , and the net head h represents the friction losses in the
105 penstock, hL , so it can be written
M

106 Hg = h + h L . (4)
These losses, hL , can be expressed [21] as the sum of distributed losses in the pipes (due to the friction
through the pipe) and the concentrated losses (due to friction in the nc pipe elbows and the nozzle), this is
D

" 2 #
Q2

L Sp
hL = f + nc kc + knoz ,
TE

Dp Snoz 2gSp2

being Sp , Dp , and L, respectively, the cross sectional area, the diameter and the length of the penstock. The
constants f , kc , and knoz represent, respectively, the friction losses coefficients due to: the pipe roughness,
the pipe elbows, and the nozzle. It can be proved [23] that
EP

 2
L Sp
f ≫ kc nc + knoz ,
Dp Snoz
C

and thus, the following aproximation is made,

L Q2
AC

hL ≈ f .
Dp 2gSp2

In this expression, the value of the cross sectional area of the penstock, Sp = πDp4 /4, can be introduced,
resulting the simplified expression
L 2
hL ≈ kp Q , (5)
Dp5
where constants have been properly grouped using
8f
kp = .
π2 g

4
 ACCEPTED MANUSCRIPT

107 Expressions (5) and (3) can be now introduced in (6). The resultant expression can be written in a
108 simplified way as
1 L
109 Hg = 2
Q2 + kp 5 Q2 , (6)
2gSnoz Dp
110 where the flow Q can be cleared, resulting an expression in terms of the gross height, Hg , and the geometry
of the MHPP:

PT
111

" # 12
Hg
112 Q= . (7)
1
2gS 2 + kp DL5

RI
noz p

113 Finally, an expression for the obtainable power, P , in terms of the height (Hg ) and length of the penstock
114 (L), can be obtained subtituting (7) and (3) and in (1), resulting

SC
" # 32
ηρ Hg
115 P = . (8)
2
2Snoz 1
2
2gSnoz+ kp DL5
p

U
116 3.2. Cost of the installation
117 In this work, a realistic cost function will be used to represent the cost of the penstock. The price of the
AN
118 pipes [16] tipically varies with the length of the pipe, L, and the square of the diameter, Dp . To assume
119 the cost of the nc connections between consecutive pipe lengths in the cost function, a new variable λ is
120 introduced to represent this cost in terms of a virtual additional length. This is
C = CL Dp2 (L + λnc ), (9)
M

121

122 where CL is a cost constant. Although excavations and supports of the penstock layout are considered in
123 this paper, these are not included in the cost function. The main reason is that these installations, typically
funded by develop cooperation agencies, involve restrictive budget to design, devices and equipment, while
D

124

125 the workforce to carry out the installation is assumed to be a contribution from the community involved.
126 Note that both the diameter and the wall width affect the final cost and its mechanical resistance, being
TE

127 possible that these parameters vary along the penstock layout. However, in this paper a fixed pipe section
128 is considered, so including diameter and width as optimization variables is left for a future extension of this
129 work.
EP

130 4. Geometry of the problem

131 The goal of the opimization problem defined in the last section is to design a MHPP that satisfies a set
of requirements. In order to do that, it is necessary to determine the location of the dam, the powerhouse
C

132

133 and the piping installation. This section is devoted to model the geometry of the problem, which will allow
134 us to define an appropriate set of decision variables. The equations developed in the previous section are
AC

135 valid for each possible pair {Hg , L} in a river profile. However, in order to address the optimization of the
136 installation, it is necesary to characterize the geometry of the problem and then find an appropiate set of
137 decision variables.

138 4.1. Discretization of the river profile

139 The topographic profile of a river is determined by means of an expression in the form ζ = (x(k), y(k), z(k))
140 that describes a curve in space, where x(k), y(t) and z(k) are the spatial coordinates, and k is a parameter.
141 Due to the arbitrariness of the rivers studied, an analytical expression for ζ is not usually possible to be
142 found, so an easy way to accurately determine its profile is by means of a discretization.
143 Using traditional or automated cartography, a set of N position vectors, summarized in r ∈ RN ×3 ,
144 with ri = (xi , yi , zi ), can be easily obtained (see figure 3). Each of these points ri , will be considered as a
5
 ACCEPTED MANUSCRIPT

145 candidate to locate the different elements of the installation, namely: the turbine, the dam, or elbows of the
146 penstock installation.

(xi , yi , zi )

PT
y

RI
Figure 3: Cartesian topographic survey

SC
147 Assuming that the river layout follows a prevailing direction, a new vector s ∈ RN can be defined as the
148 arc-length of the projection of the river onto the x-y plane. This projection is indexed by i, being si the
149 i-th element of s, and being each segment in the horizontal plane added to the length variable as follows:

U
150 si+1 = si + k[1 1 0](ri+1 − ri )k, (10)
151 where s1 =0, as represented in figure 4. This way, it is possible to develope the river profile in the s-z plane,
AN
152 and therefore, to analyze it in the 2D space. Analogously, the vector z ∈ RN is defined as the height of the
153 N points of discretization.
154 With some abuse of notation, in the rest of the paper the components si , ∀i ∈ 1, 2, . . . , N and zi , ∀i ∈
155 1, 2, . . . , N will be referred as the coordinate of the i-th point of the river, the i-th point of the river, or
M

156 simply the i-th point.

D

(si , zi )

z
TE

s
EP

Figure 4: sz-parametrized topographic survey

For a better understanding, an arbitrary river profile is represented in figure 1.

C

z10
AC
...

z6
...

z1

s1 s2 s3 s4 s5 s6 s7 s8 s9 s10

Figure 5: Discrete river profile

157

6
 ACCEPTED MANUSCRIPT

158 4.2. Penstock layout

Since a real penstock is composed of an amount of rectilinear connected pipe lengths, a linear piecewised
function can be used to describe its layout. This function can be defined using a set of N variables δi ,
defined as
δi ∈ {0, 1}; i = 1...N, (11)

PT
159 that indicates if a connection between two pipe lengths is placed at point i, as it is shown in figure 6, where
160 zP i represents the height of the penstock at each point si . Also, a set of N (N − 1)/2 binary variables γij ,
161 defined as

RI
162 γij ∈ {0, 1}; i = 1...N − 1, j = i + 1...N, (12)
163 are introduced to indicate if a straight pipe length is installed between points i and j. For a better under-
164 standing of the variables δi and γij , these can be arranged in a matrix ∆, where δi are the diagonal terms

SC
165 and γij are placed in the upper diagonal terms, being irrelevant the lower diagonal of ∆.
 
δ1 γ12 · · · γ1,N −1 γ1N
 δ2 · · · γ2,N −1 γ2N 

U
.. ..
 
∆=
 .. 
(13)
166
 . . . 

 ∗ δN −1 γN −1,N 
AN
δN

167 For instance, a possible penstock layout for the example of the river profile in figure 5 is represented in
figure 6.
M

z10
z7
D

z6
z3
TE

δi
1
EP

0
s1 s2 s3 s4 s5 s6 s7 s8 s9 s10

Figure 6: Penstock parametrization

C

168

169 Following the structure shown in (13), for this penstock installation, the variables δi and γij take the
170 values shown in (14), where only non-zero elements (circled for δi and squared for γij ) are represented.
AC

· · · · · · · · · ·
 
 · · · · · · · · · 
 

 1 · · 1 · · · ·  

 · · · · · · ·  
 · · · · · · 
171 ∆=   (14)
 1 1 · · ·  

 1 · · 1  

 ∗ · · ·  
 · · 
1

7
 ACCEPTED MANUSCRIPT

172 4.3. Number of pipe lengths and elbows

173 It is evident that the summation of variables δi represents the number of pipe connections needed in the
174 penstock installation. The term nc is introduced to describe this variable, being expressed as
N
X
175 nc = δi . (15)
i=1

PT
176 Analogously to nc , it can be clearly checked that the summation of variables γij represents the number
177 of pipe lengths in the penstock installation, nl , as expressed in (16), where nl has been introduced to denote
178 this variable.

RI
N
X −1 N
X
179 nl = γij . (16)
i=1 j=i+1

SC
180 4.4. Gross head and length expressions
181 In order to formulate the optimization problem as an integer programming, it is convenient to express
182 both the gross head, Hg , and the total length of the penstock, L, in terms of the binary variables, γij , defined
in the last section. It is easy to see that, using variables γij in (12), the gross head Hg can be expressed as

U
183

184 the sum of the net height of the pipe lengths as

AN
N
X −1 N
X
185 Hg = γij (zj − zi ). (17)
i=1 j=i+1

186 Equivalently, it is possible to compute the total length of the penstock as the sum of the individual
M

187 length of each of the intervals that compose the penstock as

N
X −1 N
X
188 L= γij lij , (18)
D

i=1 j=i+1

where the constants lij represent the length of the possible straight pipe length between each pair of points
TE

i and j, which can be computed as

q
lij = (sj − si )2 + (zj − zi )2 ; i = 1...N, j = i + 1...N.
EP

189 4.5. Penstock height expression

190 Finally, this section provides a method to represent the height profile of the penstock through the
191 definition of a new set of variables zpi ∈ Rz×N . Analogously to the terrain height profile variable zi , this
new variable will be also associated to the river points si , so that pairs (si , zpi ) are defined. However, for
C

192

193 each coordinate of the river profile si , zpi will indicate the height of the penstock, if there is a pipe in this
194 particular point, or the height of the terrain, elsewhere. At the river points where there is a pipe elbow
AC

195 connecting to pipe lengths, this is δi = 1, it is clear that both variables zi and zpi take the same value.
196 However, this is not necessarily true in the rest of the penstock installation (notice figure 6).
197 To develop a general expression of the penstock height profile, the penstock height, zpi , at each point of
198 the river, si , is computed as the sum of the penstock height at the previous point, zi−1 , and the height gain
199 of the pipe between si−1 and si , as
200 zP i+1 = zP i + (si+1 − si )mi , i = 1, . . . , N − 1, (19)
201 where the variable mi represents either the slope of the penstock, in case that there is a pipe length between
202 the river points si+1 and si , or the slope of the terrain, otherwise. Observing the structure of the matrix ∆
203 in (13), it i not difficult to see that, for each point j when a straight pipe length installed passes through
204 a given point of the river si , the slope of this segment mi is equal to the slope corresponding to the only
8
 ACCEPTED MANUSCRIPT

205 non-zero element in the submatrix defined by the elements (j, j + 1) and (1, n), so mj at every point of the
206 river can be calculated as the slope of the terrain in j, mj,j+1 , and the summation of the elements of this
207 submatrix multiplied by the corresponding penstock length slope minus terrain slope, as
N
X q
X
208 mj = mj,j+1 + γpq (mpq − mj,j+1 ), (20)
q=j+1 p=1

PT
209 where mpq represents the slope of a penstock interval that connects points p and q, as defined as
zq − zp
210 mpq = , (21)
sq − sp

RI
211 p = 1 . . . N − 1, q = p + 1 . . . N.

212 4.6. Gap expression

SC
213 Once an expression for the penstock profile, zP , is defined, the gap ǫ between the terrain and the pipe
214 can be determined as the difference between these:
215 ǫi = zP i − zi , i = 1...N (22)

U
216 where zP is defined in expression (19).
AN
217 5. Integer programming

218 In this section, based on the plant model developed in section 3, and on the geometry of the problem and
219 the new variables introduced in section 4, the optimization problem introduced in section 2 is formulated as
M

220 an integer programming problem, for wich efficient algorithm exist in the literature.

221 5.1. Optimization problem

D

As introduced in section 2, the aim of the optimization problem is to find a MHPP such that, for
prescribed values of generated electric power, water flow, excavations and supports, the piping cost is
minimized. In order to solve it, this section transforms the problem to a linear integer programming problem,
TE

for which efficient algorithm can be implemented [24]. This problem has the structure shown in (23).
minimize cx
x
s.t. Ax ≤ b (23)
EP

Aeq x = beq
222 where x ∈ Z+n is a vector stacking the n positive integer decision variables, and A ∈ Rm×n , b ∈ Rm and
Aeq ∈ Rmeq ×n , b ∈ Rmeq are the matrix and the independent terms of the m, meq inequality and equality
C

223

224 constraints, respectively.

AC

225 5.2. Decision variables

226 In this problem, the vector x is composed of n = N (N + 1)/2 decision variables; being N of them δi ,
227 defined in (11), and N (N − 1)/2 of them γij , defined in (12). In view of the structure in (14) , it is clear
228 that, for a continuos single penstock, certain constrains have to be imposed to variables γij and δi so that
229 they are consistent with the problem.
230

231 Firstly, the number of pipe elbows, nc , is equal to the number of pipe lengths, nl , plus one. Using (15)
232 and (16), this restriction can be expressed as
N
X N
X −1 N
X
233 δi = 1 + γij . (24)
i=1 i=1 j=i+1

9
 ACCEPTED MANUSCRIPT

Secondly, γij can be equal to one if and only both δi and δj are non-zero, and every δk with i ≤ k ≤ j
is zero. This is analogous to force each δi in matrix (13) to be greater or equal than its file and column
summations of γij , separately, which can be expressed as
N
X
δi ≥ γij ; i = 1 . . . N − 1, (25)
j=i+1

PT
j
X
δj ≥ γij ; j = 2 . . . N. (26)
i=1

RI
234 5.3. Cost function
235 The main objective of the problem studied in this paper is satisfying the power need without exceeding the
236 available resources and minimizing the cost, so a cost minimization objective function g(δi , γij ) is presented.

SC
237 Using (8) and (9), and assuming that the diameter and the material of the penstock are fixed, the proposed
238 cost funcion can be written as
239 g(δi , γij ) = λnc (δi ) + L(γij ). (27)

U
240 Introducing (17) and (18) in this expression, the cost function can be expressed as
N N −1 N
AN
X X X
241 g(δi , γij ) = λ δi + γij lij . (28)
i=1 i=1 j=i+1

242 5.4. Power constraint

M

243 As mentioned, one of the main restrictions of the problem is the satisfaction of basic needs, which in this
244 context can be expressed as generating at least a minimal amount of power. This restriction can be written
245 as
D

246 P ≥ Pmin (29)

where P is defined in (8) and Pmin is a constant parameter that represents the minimal power that needs
TE

247

248 to be generated. Substituting (8) in (29), this expression can be written in terms of Hg and L as
" # 32
ηρ Hg
249 ≥ Pmin . (30)
EP

2
2Snoz 1
2
2gSnoz + kp DL5
p

250 Reordering terms in this expression, it can be expressed as

a 1 Hg + a 2 L ≤ a 3 , (31)
C

251

where constants a1 , a2 and a3 are defined by

AC

 2
ηρ 3 kp 23 1 2
a1 = − 2
, a 2 = Pmin , a 3 = − P 3
min .
2Snoz Dp5 2
2Snoz

252 Lastly, introducing (17) and (18), the constraint in (31) can be written as a linear function of the decision
253 variables, as shown in (32).
N
X −1 N
X N
X −1 N
X
254 a1 γij (zj − zi ) + a2 γij lij ≤ a3 (32)
i=1 j=i+1 i=1 j=i+1

10
 ACCEPTED MANUSCRIPT

255 5.5. Flow constraint

256 Taking into account that the river flow is limited, the following restriction has to be introduced.
257 Q ≤ κQav (33)
258 where Qav si the average river flow (m3 /s), and κ represents the maximum portion (p.u.) of flow allowed
259 to be used in the MHPP, and it is introduced to prevent the drying up of the river. Introducing (7) in (33)

PT
260 and reordering terms, the restriction can be expressed as
" # 12
Hg
261 ≤ κQav . (34)
1
2gSnoz 2
+ kp DL5

RI
p

262 Reordering terms, (34) can be transformed into

263 b 1 Hg + b 2 L ≤ b 3 , (35)

SC
where constants b1 , b2 and b3 are defined by

2 2 kp 2 2
b1 = 2gSnoz , b2 = −2gSnoz [κQav ] , b3 = [κQav ] .
Dp5

264

U
Finally, using (17) and (18), the constraint in (35) can be written as a linear function of the decision
AN
265 variables. The resultant expression is
N
X −1 N
X N
X −1 N
X
266 b1 γij (zj − zi ) + b2 γij lij ≤ b3 (36)
i=1 j=i+1 i=1 j=i+1
M

267 5.6. Gap constraint

To represent the limitations in height and depth, respectively, of supports and excavations, the following
restrictions are introduced:
D

ǫi ≤ ǫsup ∀i = 2 . . . N − 1,
TE

ǫi ≥ ǫexc ∀i = 2 . . . N − 1.
268 Introducing (22) and (19) in these constraints, they are transformed into
i
EP

X
269 (sj+1 − sj )mj ≤ ǫsup + zi − z1 , (37)
j=1

270 ∀i = 2 . . . N − 1.
C

i
X
271 (sj+1 − sj )mj ≥ ǫexc + zi − z1 , (38)
AC

j=1

272 ∀i = 2 . . . N − 1.

with mj defined in (20). Introducing (20), these restrictions can be written in terms of the decision variables
as
i
X N
X q
X
(sj+1 − sj )(mpq − mj,j+1 )γpq
j=1 q=j+1 p=1
i
X
≤ ǫsup + zi − z1 − (sj+1 − sj )mi,j+1 (39)
j=1

11
 ACCEPTED MANUSCRIPT

273 ∀i = 2 . . . N − 1

i
X N
X q
X
(sj − sj+1 )(mpq − mj,j+1 )γpq
j=1 q=j+1 p=1

PT
i
X
≤ −ǫexc + z1 − zi + (sj+1 − sj )mi,j+1 (40)
j=1

∀i = 2 . . . N − 1

RI
274

275 Note that, due to the definition of zP i in (19), the restrictions from (39) and (40) are, in a trivial way,
276 satisfied in the points outside the penstock domain.

SC
277 5.7. General formulation
278 The complete formulation of the integer optimization problem can be written combining the cost func-
279 tion and the constraints related to the minimal power obtainable, the maximum flow extractable, and the

U
280 maximum available excavation and supporting as follows. AN
N
X N
X −1 N
X
minimize λ δi + γij lij
δi ,γij
i=1 i=1 j=i+1
s.t. (24), (25), (26), (32), (36),
M

(39), (40), (11), (12).

D

281 To facilitate the implementation of the problem, a graphic user interface (see figure 7) has been created
282 using the Matlab GUI enviroment, which is available for download at https://github.com/atapiaco/
TE

283 PenstockOptimizer.

284 6. Practical example

EP

285 In this section, the proposed optimization problem is tested for a random river profile. The river profile
286 studied has a total height of 250 m with a horizontal domain of 1.50 km, and water flow Qav = 70L/s. A
287 set of N =150 points is taken to discretize the river. Accordingly to this, a total of 150 variables δi and
288 n =11175 variables γij are defined in the problem. The proposed pipe is made in PVC, with an internal
C

289 diameter D = 0.20m and a friction coefficient of f = 9.96 × 10−4 [23]. The nozzle diameter is Dnoz = 22mm,
290 and the global efficiency of the MHPP is assumed to be η = 0.85. A minimal power supply of 8 kW and a
maximum flow extraction rate of 50% are imposed as main constraints. The excavations and supports are
AC

291

292 both limited to 1,5 m.

293

294 To evaluate the influence of the elbow equivalent cost λ, introduced in section 3.2, the problem is first
295 solved using λ=0 (problem a), and then solved using λ=50 m (problem b). As an ilustrative example of
296 the influence of the supports and excavations constraints, the assumption of a high bearing capacity of the
297 terrain can be taken into account in terms of these limits, which would be a low excavation allowance due
298 to the difficuly of digging, and a higher supports allowance due to the high tip-resistance of the terrain.
299 Following this, a third problem is studied (problem c) setting these limits, respectivley, to 0.50 m and 3
300 m, maintaining the rest of previous constraints. Another case is studied (problem d) by setting an infinite
301 value of these limits, so the constraints related to the supports and excavations are no longer relevant in the

12
 ACCEPTED MANUSCRIPT

PT
RI
U SC
AN
M

Figure 7: Graphic user interface for the optimization problem

optimization. The parameters of these problems and their optimal solutions are summarized in table 1. Fi-
D

302

303 nally, a last problem is proposed (problem e) using the same parameters presented in problem b, but setting
304 a higher value of the minimal power, Pmin =20kW. The optimal penstocks associated to these solutions are
TE

305 represented in figure 8 together with the river profile.

306

307 After observing the results, a few comments can be made. For the first problem (problem a), it can be
308 seen that neglecting the cost of the elbows does not imply an excessive number of them, due to the fact that,
EP

309 for each couple of tubine and dam locations, the fewer elbows implies the longer pipe lengths, and so the
310 lower total length of the penstock. Notwithstanding this, when a positive equivalent cost of the elbows is set,
311 the algorithm finds an optimal penstock with a lower number of elbows, requiring a longer penstock and a
312 resultant higher cost. Regarding the effects of a high bearing capacity of the terrain, it can be seen that the
C

313 optimal placement is far different from the previous cases, which make sense, since longer pipe lengths can
AC

Table 1: Practical example results

Problem Solution
Pmin κ ce ǫsup ǫexc P Hg L nc C
Case (kW) (%) (m) (m) (m) (kW) (m) (m) (m)
a 8,0 50 0 1,5 1,5 8,146 69,7 357,6 6 357,6
b 8,0 50 50 1,5 1,5 8,330 70,8 358,0 5 608,0
c 8,0 50 50 3 0,5 8,043 69,1 416,5 4 616,5
d 8,0 50 50 ∞ ∞ 8,146 69,7 355,4 2 455,4
e 20,0 50 50 1,5 1,5 20,339 128,42 697,6 9 1147,6

13
 ACCEPTED MANUSCRIPT

Elbow location

300

200
zi (m)

100

PT
300

200
zi (m)

RI
100

SC
300

200
zi (m)

100

300

U
AN
200
zi (m)

100
M

300

200
zi (m)

100
TE

0
0 500 1000 1500

si (m)
EP

Figure 8: Optimal penstock for example problems (a) to (e).

314 be installed in hollow areas, due to the higher available supports, while convex areas need a higher number
C

315 of elbows due to the severe excavation limit. Notice that this terrain implies the need of a longer and more
316 expensive penstock. For problem d, where supports and excavationes are no longer restricted, the algorithm
finds a new optimal penstock with a single length, as expected. In this particular case, the cost is obviously
AC

317

318 inferior than in cases b and c, and the optimal penstock is the shortest one. Lastly, the optimal penstock
319 for problem b is the longest and most expensive, requiring a higher number of elbows, as expected due to
320 the high power needed.
321

322 Regarding the computational cost, it is easy to see that the complexity class of the problem is N P .
323 Although the number of variables is O(N 2 ) the linear structure allows a significant reduction of the search
324 space, so branch and bound methods are ideally suitable. For an empirical evaluation, using a random river
325 profile with a discretization of N =300 points, the optimization problem has been solved using N =300,
326 N =150, N =75, N =38 and N =17. The optimal cost of the installation and the computation time for each
327 case are showed in figure 9. It can be seen not only that the problem is feasible for a reasonable amount of

14
 ACCEPTED MANUSCRIPT

328 geographical points N , but also that improvements in the discretization leads to lower optimal costs.

80 450
Solving time
Optimal cost 400
60

cost (m)
time (s)

350

PT
40
300

20 250

RI
50 100 150 200 250 300
discretization, N

Figure 9: Computation times and optimal cost for different number of points N

SC
329 7. Conclusion

U
330 In this paper an efficient, robust penstock optimization problem is developed. This problem helps to
331 determine the optimal MHPP installation, ensuring a minimum power supply, a maximum use of water flow
AN
332 and assuming terrain excavation and supports limitations. A generic river profile has been proposed, and
333 three different problems has been solved, showing evidence that this method produces good quality solutions
334 with a low computational cost.
M

335 Acknowledgement

336 The research project on MHPP of which this work is part is supported by grant 2014/ACDE/006016
(AECID) and grant DPI-75294-CS-2-R (AEI / FEDER, UE).
D

337

References
TE

338

339 [1] Hira Singh Sachdev, Ashok Kumar Akella, and Niranjan Kumar. Analysis and evaluation of small hydropower plants: A
340 bibliographical survey. Renewable and Sustainable Energy Reviews, 51:1013 – 1022, 2015.
341 [2] J Kaldellis and K Kavadias. Laboratory applications of renewable energy sources. Stamoulis, Athens, 2000.
342 [3] Nima Izadyar, Hwai Chyuan Ong, W.T. Chong, and K.Y. Leong. Resource assessment of the renewable energy potential
EP

343 for a remote area: A review. Renewable and Sustainable Energy Reviews, 62:908 – 923, 2016.
344 [4] A.H. Elbatran, O.B. Yaakob, Yasser M. Ahmed, and H.M. Shabara. Operation, performance and economic analysis of
345 low head micro-hydropower turbines for rural and remote areas: A review. Renewable and Sustainable Energy Reviews,
346 43:40 – 50, 2015.
347 [5] Sonali Goel and Renu Sharma. Performance evaluation of stand alone, grid connected and hybrid renewable energy systems
C

348 for rural application: A comparative review. Renewable and Sustainable Energy Reviews, 78:1378 – 1389, 2017.
349 [6] Sanjoy Kumar Nandi and Himangshu Ranjan Ghosh. A wind–pv-battery hybrid power system at sitakunda in bangladesh.
Energy Policy, 37(9):3659 – 3664, 2009. New Zealand Energy Strategy.
AC

350
351 [7] M. Kashif Shahzad, Adeem Zahid, Tanzeel ur Rashid, Mirza Abdullah Rehan, Muzaffar Ali, and Mueen Ahmad. Techno-
352 economic feasibility analysis of a solar-biomass off grid system for the electrification of remote rural areas in pakistan using
353 homer software. Renewable Energy, 106:264 – 273, 2017.
354 [8] C.P. Jawahar and Prawin Angel Michael. A review on turbines for micro hydro power plant. Renewable and Sustainable
355 Energy Reviews, 72:882 – 887, 2017.
356 [9] Oliver Paish. Small hydro power: technology and current status. Renewable and sustainable energy reviews, 6(6):537–556,
357 2002.
358 [10] ESMAP. Technical and economic assessment of off-grid. Technical report, 12 2017.
359 [11] Bryan R. Cobb and Kendra V. Sharp. Impulse (turgo and pelton) turbine performance characteristics and their impact
360 on pico-hydro installations. Renewable Energy, 50:959 – 964, 2013.
361 [12] A.V. Braga, A.J.J. Rezek, V.F. Silva, A.N.C. Viana, E.C. Bortoni, W.D.C. Sanchez, and P.F. Ribeiro. Isolated induction
362 generator in a rural brazilian area: Field performance tests. Renewable Energy, 83:1352 – 1361, 2015.
363 [13] SJ Williamson, BH Stark, and JD Booker. Low head pico hydro turbine selection using a multi-criteria analysis. Renewable
364 Energy, 61:43–50, 2014.

15
 ACCEPTED MANUSCRIPT

365 [14] Maxime Binama, Wen-Tao Su, Xiao-Bin Li, Feng-Chen Li, Xian-Zhu Wei, and Shi An. Investigation on pump as turbine
366 (pat) technical aspects for micro hydropower schemes: A state-of-the-art review. Renewable and Sustainable Energy
367 Reviews, 79:148 – 179, 2017.
368 [15] D.R. Giosio, A.D. Henderson, J.M. Walker, P.A. Brandner, J.E. Sargison, and P. Gautam. Design and performance
369 evaluation of a pump-as-turbine micro-hydro test facility with incorporated inlet flow control. Renewable Energy, 78:1 –
370 6, 2015.
371 [16] K. V. Alexander and E. P. Giddens. Optimum penstocks for low head microhydro schemes. Renewable Energy, 33(3):507–
372 519, 2008.

PT
373 [17] ESHA. Guide on How to Develop a Small Hydropower Plant. 2004.
374 [18] Thake J. Micro-hydro Pelton Turbine Manual: Design, manufacture and installation for small-scale hydropower. 2000.
375 [19] Japan International Cooperation Agency (JICA). Manuals and Guidelines for Micro-hydropower Development in Rural
376 Electrification Volume I, volume I. 2009.
377 [20] K. V. Alexander and E. P. Giddens. Microhydro: Cost-effective, modular systems for low heads. Renewable Energy,

RI
378 33(6):1379–1391, 2008.
379 [21] Arturo S Leon and Ling Zhu. A dimensional analysis for determining optimal discharge and penstock diameter in impulse
380 and reaction water turbines. Renewable Energy, 71:609–615, 2014.
381 [22] Massey B. Mechanics of fluids. 7th ed. Cheltenham: Stanley Thornes Ltd. 1998.

SC
382 [23] Perry R. Green D. Perry’s Chemical Engineers’ Handbook, Eighth Edition. 2008.
383 [24] Sierksma G. Linear and Integer Programming: Theory and Practice, Second Edition. 2001.

U
AN
M
D
TE
C EP
AC

16
 ACCEPTED MANUSCRIPT

384 Appendix: Problem formulation

385 In this appendix, the general problem formulation for a N -discretized river profile is developed, in the
386 form of

minimize cx,
x

PT
s.t. Ax ≤ b
Aeq x = beq ,
x ≥ 0,

RI
387 T

SC

x= δ1 δ2 · · · δN γ12 γ13 . . . γ1N γ23 γ24 . . . γ2,N . . . γN−1,N γN−2,N−1 γN−2,N ,
   
Aeq = 11×N −11×N−1 −11×N−2 −11×N−3 . . . −1 −1 −1 , beq = 1 ,
   
Arows brows

U

 Acolumns 


 bcolumns 

 Apower   bpower 
A= , b= ,
Af low bf low

AN   
   
 Asup   bsup 
Aexc bexc

11×N−1 01×N−2 01×N−3 0 0 0

 
M


 01×N−1 11×N−2 01×N−3 0 0 0 

Arows =  −IN−1×N−1
 0N−1×1 01×N−1 01×N−2 11×N−3 ... 0 0 0 ,

brows =

0N−1×1

,
 .. .. .. .. .. .. 
 . . . . . . 
D

01×N−1 01×N−2 01×N−3 0 0 1

 
TE


 01×N−1 02×N−2 0N−1×1 
  
Acolumns = 
 0N−1×1 −IN−1×N−1 IN×N ··· ,
 bcolumns = 0N−1×1 ,
 IN−1×N−1 IN−2×N−2 1 
EP

Agap_1 + . . . + ATgap_N−1
 
Asup = Agap_1 Agap_1 + Agap_2 Agap_1 + Agap_2 + Agap_3 ... ,

−Agap_1 − Agap_2 − Agap_3 . . . −Agap_1 − . . . − ATgap_N−1 ,

 
Aexc = −Agap_1 −Agap_1 − Agap_2
C

   
bpower = a3 , bf low = b3 ,
ǫsup − z1 + z2 − m12 ∆s12 ǫexc + z1 − z2 − m12 ∆s12
   
AC


 ǫsup − z1 + z3 − m23 ∆s23 


 ǫexc + z1 − z3 − m23 ∆s23 

bsup =
 ǫsup − z1 + z4 − m34 ∆s34 ,

bexc = 
 ǫ exc + z 1 − z 4 − m 34 ∆s 34
.

 ..   .. 
 .   . 
ǫsup − z1 + zN−1 − mN−2,N−1 ∆sN−2,N−1 ǫexc + z1 − zN−1 − mN−2,N−1 ∆sN−2,N−1

17
 ACCEPTED MANUSCRIPT

 T  T
0N×1 0N×1
   
   

 a1 (z2 − z1 ) + a2 l12 


 b1 (z2 − z1 ) + b2 l12 

a 1 (z3 − z1 ) + a2 l13 b1 (z3 − z1 ) + b2 l13
   
   
 ..   .. 

 . 


 . 


 a1 (zn − z1 ) + a2 l1n 


 b1 (zn − z1 ) + b2 l1n 

   

 
a1 (z3 − z2 ) + a2 l23
 
 
a1 (z4 − z2 ) + a2 l23
 
 
..
 
 
 .  
PT
 
b1 (z3 − z2 ) + b2 l23
 
 
b1 (z4 − z2 ) + b2 l24
 
 
..
 
 
. 

RI
   

 a 1 (z n − z 2 ) + a2 l2n 


 b 1 (z n − z 2 ) + b2 l2n 

Apower =

 ,
 Af low =

 ,


 a1 (z4 − z3 ) + a2 l34 


 b1 (z4 − z3 ) + b2 l34 

a 1 (z5 − z3 ) + a2 l35 b1 (z5 − z3 ) + b2 l35
   

SC
   
 ..   .. 
. .
   
   
a1 (zn − z3 ) + a2 l3n b1 (zn − z3 ) + b2 l3n
   
   
   
   
.. ..
   

U
   

 . 


 . 

   
   
 a1 (zn−1 − zn−2 ) + a2 ln−2,n−1  b1 (zn−1 − zn−2 ) + b2 ln−2,n−1
AN
 
   

 a1 (zn − zn−2 ) + a2 ln−2,n 


 b1 (zn − zn−2 ) + b2 ln−2,n 

   
a1 (zn − zn−1 ) + a2 ln−1,n b1 (zn − zn−1 ) + b2 ln−1,n
M

 
0
 0 
0
   
 (m14 − m34 )∆s34 
(m13 − m23 )∆s23
 
..
D

     
(m12 − m12 )∆s12 
 (m14 − m23 )∆s23



 . 

 (m13 − m12 )∆s12  
 ..



 (m1,N − m34 )∆s34 

.
 
TE

 (m14 − m12 )∆s12  








(m1,N − m23 )∆s23
 
 ..  




 0



 . 
 




 (m24 − m34 )∆s34



 (m1,N − m12 )∆s12 
 
 (m23 − m23 )∆s23 


 .. 




  (m24 − m23 )∆s23   . 
EP

   

 0N−2×1 
  ..   (m2,N − m34 )∆s34 
.
     
     
(m2,N − m23 )∆s23
   
0N−3×1 (m23 − m34 )∆s34
     
Agap_1 = , Agap_2 = , Agap_3 = ,
 
     (m24 − m34 )∆s34 
0N−3×1
   
0N−4×1 ..
C

     
 









 . 

(m2,N − m34 )∆s34
   

..
  0N−4×1   
AC

 
.
     
     
0N−4×1
   



  ..   
 0  
 . 





 
0 ..
     
   
.
 
  
 0 





0

 0 





  
 0 

0 
 0 

 
0

18
 ACCEPTED MANUSCRIPT
• The paper proposes a design methodology to optimize Micro-Hydro Power Plants
• Models of the obtainable power, the problem geometry, and the installation cost are
developed
• The resulting optimization problem is casted as an integer optimization problem
• A simulated case study shows the main characteristics of the design method
• A friendly graphic user interface has been programmed

PT
RI
U SC
AN
M
D
TE
C EP
AC