This action might not be possible to undo. Are you sure you want to continue?
GEOTHERMAL TRAINING PROGRAMME Reports 2002
Orkustofnun, Grensásvegur 9, Number 13
IS108 Reykjavík, Iceland
MODELLING OF BALÇOVA GEOTHERMAL DISTRICT HEATING SYSTEM
Adil Caner Şener
İzmir Institute of Technology
Geothermal Energy Research Development Test and Education Centre
Gülbahçe, İzmir
TURKEY
acanersener@yahoo.com
ABSTRACT
In this study Balçova geothermal district heating system has been simulated to
determine the optimum working conditions of the system. Geothermal pipeline
system and city distribution system have been modelled in the Pipelab district
heating simulation program. To model the system close to the actual case,
database of Balçova geothermal company was used as an input, and the code of
Pipelab program adapted to system conditions. Moreover, to determine the
optimum operation strategy of the 8 well pumps according to the changing heat
energy demand, dynamic programming algorithm, selecting the best operation
strategy, was created. Then to model the time dependent behaviour of heat energy
demand of the system, time series analysis technique was used. Finally results
were compared with the actual situation of the system, and potential improvements
on the system were discussed.
1. INTRODUCTION
The term simulation refers to the process of reproducing the behaviour of one system through the
functions of another. In this project, the term simulation will refer to the process of using
mathematical representation of the real system, or model. Network simulations, which replicate the
behaviour of an existing or proposed system, are commonly performed when it is not practical for the
real system to be directly subjected to experimentation. (Walski et al., 2002)
During this study two goals were aimed at, 1. Simulating the system 2. From the results of the
simulation, determining the optimum operation strategy for the system.
Balçova geothermal district heating system is mainly composed of two subsystems, geothermal
pipeline system and city distribution loop. The geothermal pipeline system transfers the geothermal
fluid from wellheads to the pumping stations. In the pumping stations, the energy of the geothermal
fluid is transferred to the clean water, with the help of plate type heat exchangers, which is then
circulated in the city distribution loop.
The geothermal pipeline system is connected to production and injection wells. Geothermal fluid,
produced from the production wells, is pumped into the geothermal pipeline system and after giving
its energy; it is pumped into the injection wells from the geothermal pipeline system. Although the
1
Şener Report 13 2
system is generally called Balçova district heating system, the geothermal pipeline system does not
supply heat to only Balcova. It is also connected to the several facilities (Pool, spa, hotels, hospitals),
which take approximately 40% of the produced heat energy. Therefore the geothermal pipeline
system can also be treated as distribution system supplying geothermal water to different stations.
The Balçova city distribution loop delivers hot water to the buildings, which are connected to the
system. Each building has its own heat exchanger in which energy of city distribution water is
transferred to the building heating system. After giving its energy to building heating system, the city
circulation water returns to the pumping station.
There are two main criteria for the proper work of the geothermal heating systems; the required heat
energy should be supplied to system (thermodynamic balance) and this energy should be transmitted
to the location where it is required (hydraulic balance). Therefore as a first step, the geothermal
pipeline system has been modelled thermodynamically and hydraulically, assuming that the system
works properly. After modelling the system for various heat loads, the system has been simulated and
optimum working conditions for the system were determined according to varying heat loads.
During modelling, the actual system information taken from Balçova geothermal company was used
to get as correct as possible results from modelling. For the optimisation of well operations dynamic
programming algorithm was created in the Turbo Pascal language. The modelling of the network has
been done by using the Pipelab district heating simulation programme. The geothermal pipeline
system and the city distribution loop have been modelled separately. With the help of the programs,
pressure and heat losses, flow directions; temperatures and nodal pressures of the system have been
determined for the varying flow rates.
Finally, the results of the study were compared with the actual system data. Potential improvements
and possible control strategies were investigated to reduce the power consumption and provide the
proper operation of the system.
2. BALÇOVA GEOTHERMAL DISTRICT HEATING SYSTEM
2.1 Brief description and history of Balçova geothermal field
Balçova is one of the districts of İzmir City which is located at the western tip of Anatolia (Figure 1).
The history of Balçova geothermal field goes to ancient times. Balçova geothermal field or so called
Agamemnon Spas have been attractive place for settlers over the ages. Agamemnon Spas were known
in antiquity for the therapeutic qualities of the water. According to a legend, Agamemnon was advised
by an oracle to bring soldiers who had been wounded during campaign against Troy to the sulphur
rich waters of these natural hot springs. The periods that Ionians passed to Aegean coasts, a part of
Alexander the Great's army's injured soldiers were cured in these hot springs. It had a wide usage in
that period, constructions were brought and progressed. Today, the ancient ruins are not seen in the
area. Only information about the springs are available from the historical sources in 1763. After that
period, Agamemnon Spas are reconstructed by a Frenchman called Elfont Meil with adding the
staying units. (Gökçen, 2002)
In 1962 and 1963 the reconnaissance and exploration studies started with resistivity, thermal probing,
and self potential surveys in Balçova. It was the first time that the geothermal area received a
systematic, scientific delineation in Turkey. There was a single manifestation of hot water, a spring,
had a temperature of 72°C. At the begining three wells were drilled, including the first geothermal
exploratory well in Turkey. The first well drilled produced a mixture of hot water and steam at 124 °C
at a depth of 40 m. The survey revealed a fault zone delineated by low resistivity and huge
temperature closures under 3050 m thick alluvium. Because of high carbonate content and rapid
scaling the geothermal utilization did not start until 198182. From 1981 to 1983, 16 wells, including
7 thermal gradient and 9 production wells (100150 m) were drilled. Temperatures of 50°C to 126°C
Report 13 Şener 3
with a flow of 420 kg/s were encountered. Turkey's first downhole heat exchanger application was
used to heat the healt centre and hotel. In 1983, geothermal heating for Dokuz Eylül University,
Medical Facility Campus and Hospital Building (about 30,000 m
2
) began operation. (Battocletti,
1999). For the next 18 years, geothermal energy utilization was put into use for certain facilities like
swimming pools, health centres, hospital buildings, and district heating systems. At the present total
capacity of the geothermal complex is 50000 kW
th
.
FIGURE 1: Map of Turkey
2.2 Presentation of existing geothermal utilization at Balcova
2.2.1 Geothermal pipeline system
Balçova geothermal pipeline system carries geothermal fluid from 8 production wells to 8 different
heat exchanger stations. After giving its energy at the heat exchanger stations, geothermal fluid is
pumped to 8 reinjection wells. In Table 1 and Table 2 maximum flows and wellhead temperatures of
these wells are given. However, number of wells, and heat exchanger stations has been changing
since the system was first established. Number of connections to the system and from the system,
changing according to changing well characteristics, addition of new wells and new customers. Data
provided at Table 1 and Table 2 is refers to the 20012002 heating session.
TABLE 1:Production wells in Balçova
(2001 November)
Well Temperature (°C) Max flow (kg/s)
BD2 126 22.2
BD4 138 38.9
BD6 135 33.3
BD7 118 22.2
B4 86 13.9
B5 105 41.7
B10 92 30.5
B11 92 11.1
In addition to reinjection wells given in Table 2, geothermal water is used for the curing purposes in
the spa.
Şener Report 13 4
TABLE 2:Reinjection wells in Balcova
(2001 November)
Well Flow (kg/s)
BD3 13.88
BD5 16.67
B2 10
B9 60
B12 30
ND1 5
N1 3.33
K1 13.88
Facilities, which are connected, to geothermal pipeline system are given in Table 3,
TABLE 3 Heat exchanger stations directly connected to geothermal pipeline system (2001 November)
Facility Max. Heat demand (kW) Heat demand (%)
Balcova Geothermal DHS 50364 63%
Narlidere Geothermal DHS 5800 7%
9 Eylul Hospital  1 14000 17.5%
9 Eylul Hospital 2 1700 2%
Pools 1275 1.5%
Spa 2200 3%
T. Hotel 1700 2%
P. Hotel 3200 4%
Total 80239 100
Schematic presentation of the geothermal pipeline system is given in Figure 2. However, the real
system is much more complicated than the scheme, because heat exchanger stations and wells are not
located in the same order in the field. The system is actually composed of two parallel pipeline
systems. While the supply pipeline transmits the geothermal fluid to the stations, the return pipeline
collects the geothermal fluid at the heat exchanger outlets and transmits to the reinjection wells. The
biggest portion of the energy produced by wells is used by Balcova district heating system.
Each of the production wells have downhole pumps, these pump are controlled with the help of
frequency converters. Therefore, flow of each well can be controlled from zero to maximum value.
Due to practical reasons, a certain minimum flow does exist for each well.
FIGURE 2: Basic scheme of geothermal pipeline system in Balçova
Spa
BD2 BD4 BD6 BD7 B4 B5 B10 B11
BD3 BD5 B9 B2 N1 K1 ND1 B12
duction wells
ection wells
Heat
exchanger
stations
HEAT
Report 13 Şener 5
2.2.2 City distribution system
In Balcova, hot water at 85 °C is delivered to the customers by using city distribution loop. City
distribution loop is 42 km long pipeline system, which distributes hot water and collects warm return
water. Today approximately 850 buildings are connected to the system. Each building has its own
heat exchanger at the first floor. Heat is transferred to the building heating systems with the help of
these heat exchangers.
In Balcova, constant tariff is applied for the heating service. There is no flow meter at the customer
end of the system. Moreover types and sizes of buildings are very variable. For instance, five floor
buildings, mosque, schools, and single houses are connected to the distribution system. Therefore heat
load demands of the customers are variable. Diameters of building connection pipes, sizes of heat
exchangers and flow controller elements are set according the heat demand.
Basic control scheme of city distribution loop is shown in the Figure 3. Water is heated in the main
heat exchanger, and pumped to supply network. Supply network are branched at each building
connection. Diameter of pipes in main network varies from 350 mm to 40 mm, while building
connection pipes vary from 25 mm to 50 mm in diameter. Duty of flow regulators is vital for the
system. As can be seen from the Figure 3, flow regulator valves are installed to outlet of building heat
exchangers (city distribution site). Flow regulators are used to keep heat exchanger outlet temperature
of city circulation water at constant value. They measure the temperature with the help of sensors and
adjust the flow rate by changing the cross sectional area of flow.
MAIN
HEX
H
E
X
5
H
E
X
4
H
E
X
3
H
E
X
2
H
E
X
1
T T T T T
2
1
Regulator
Return
Supply
FIGURE 3 Basic control scheme of city distribution network
When, heat demand of the building decreases (outside temperature increases), the return temperature
of city circulation water, coming from building heat exchanger, increases. The temperature sensor of
flow regulator measures this temperature change. Then cross sectional area of flow is decreased by
the regulator. Also if heat demand increases, flow regulator increase the cross sectional area of flow.
These pressure changes are measured by system technicians at points 1 and 2 in the Figure 3 and
recorded at the pumping station where main heat exchanger and pumps are operated. According to
Şener Report 13 6
pressure changes of the system, flow rates of the city circulation pumps are changed by frequency
converters, which are adjusted manually. The main idea in using this kind of control is to decrease the
pumping cost by adjusting the flow rate of the hot water according to varying heat demand.
Expansion tank is used to compensate pressure fluctuations and supply stable suction head to the
circulation pumps. It is also used to add water to the system, when there is leakage. For proper
operation of expansion tank, minimum required pressure at point 2 is 20 m (2 bars).
3. MODELLING OF DISTRICT HEATING SYSTEMS
Models can be used to solve ongoing problems, analyse proposed operational changes, and prepare
for unusual events. By comparing model results with field operations, the operator can determine the
causes of formulate solutions that will work correctly the first time rather than a resorting to trialand
error changes in the actual system. (Walski et al., 2002).
In this study, Balçova geothermal pipeline system and the Balçova district heating system have been
modelled according to the graph theory approach. Use of graph theory in district heating system
modelling is rather new approach, giving fast convergence and reliable results at the end.
3.1 Theory behind the model
The treatment of the theoretical background in Section 3.1 is based on Valdimarsson (1995).
3.1.1 Graph Theory
A suitable method to describe a district heating pipe system is to use concepts from network theory. A
graph is commonly defined as a composite concept of:
a. a set of nodes
b. a set of branches
c. an incidence relation
The connectivity relation relates each branch to a pair of nodes, the node where the branch originates
and the node where it ends. A distribution system can be treated as a connected graph, where pipes
correspond to branches and nodes to points where the pipes divide or are united or convey the flow to
the consumer. Here the word "pipe" is used in a general sense, that is a conduit carrying fluid or heat
from one point in space to another, and can have many elements, pumps, valves, etc.
In network theory a connectivity matrix must be defined in order to describe the above mentioned
connectivity relation for a network with n
n
nodes and n
f
branches:
Matrix A is an n
n
· n
f
matrix, with entries a
ij
where:
a
ij
= 1 if pipe j starts at node i,
a
ij
= 1 if pipe j ends at node i,
a
ij
= 0 otherwise.
The connectivity matrix as defined above has one column for each flow stream in the system, and one
row for each node. Each column can only have two nonzero entries, 1 and 1, as the flow stream has
to originate somewhere and end at some other location. Therefore the column sums of the matrix will
always be zero. The rows can have any number of nonzero entries greater than one, as many flow
streams can be connected to a single node. A connectivity matrix for the pipes is normally not
sufficient to describe a district heating system completely. There are inflow and outflow points in the
system or in the part of the system to be studied. These points set boundary conditions for the system.
It is convenient to define the boundary conditions at the physical system boundaries in a similar
manner to that used in electrical circuit theory. A datum ("ground") point can be defined, where the
Report 13 Şener 7
driving potential is zero. It is not included in the A matrix, as the matrix will then become linearly
dependent. A boundary element can then be defined, connecting the physical boundary point and the
datum point. Following is a definition of some graph terms used in graph theory solution of networks:
Tree: A subgraph G
s
of the connected graph G
n
is a tree if it is connected and G
s
have no loops.
Spanning tree: A subgraph G
s
of the connected graph G
n
is a spanning tree if it is connected, G
s
contains all nodes of G
n
and G
s
has no loops.
Tree branch: The branches belonging to a tree T are called tree branches.
Cutset: A set of branches of a connected graph G
n
(not their endpoints) is a cutset if the removal of
these branches results in a graph that is not connected, and the restoration of any one of these
branches results in the graph being connected again. The cutset can be seen as a border going through
the graph. Associated with the cutset is a direction specified by the direction of a given datum branch
in the cutset. The separate graphs obtained by removing the branches of the cutset are called
components of the graph with respect to the cutset. The net flow over the cutset must be zero in order
to conserve the mass in each of the components.
Link: The branches not belonging to a tree T are called links.
Cotree: The set of links in a network with a tree T is named cotree L with respect to the tree T.
The connectivity matrix A can be rearranged with respect to a spanning tree T containing n
T
branches
by splitting it into two submatrices A
T
and A
L
in the following manner:
A =[A
T
A
L
] (1)
The submatrix A
T
is the n
n
· n
T
connectivity matrix for the branches of the spanning tree, and the
matrix A
L
is the n
n
· n
T
connectivity matrix for the links, where n
L
denotes the number of links. The
sum of n
T
and n
L
is n
f
, the total number of branches in the network. As the datum point is not included
in the connectivity matrix, and the submatrix A
T
is based on a spanning tree, n
n
= n
T
.
Therefore A
T
is a square invertible matrix.
The system elements consist of the following groups:
h: Head sources
s: Storage tanks
r: Nonlinear resistances
p: Pipes
m: Flow sources
x: Heat exchangers
q: Heat sources
t: Temperature h: sources
These group characters are used as indices in the following text, in addition to the tree and cotree
letters T and L.
3.1.2 Flow solution
The flow in the network is treated as a vector, where the entries are subvectors for each flow
transmitting element group, both in the tree and the cotree:
Şener Report 13 8
=
pL
mL
pT
rT
sT
hT
m
m
m
m
m
m
m (2)
The cutset matrix is calculated from the connectivity matrix by:
(3) A A D
1 −
=
T
As the flow sum for any cutset equals zero, the cutset matrix multiplied by the flow vector will equal
the zero vector. The cutset matrix is then also portioned into submatrices according to the element
groups.
Dm
I 0 0 0 F F
0 I 0 0 F F
0 0 I 0 F F
0 0 0 I F F
m
m
m
m
m
m
0 =
⋅
=
hT
sT
rT
pT
hT
sT
rT
pT
mL
pL
11 12
21 22
31 32
41 42
(4)
This allows all system flows to be calculated in terms of the flows in the flow source branches and the
flow in the branches in the cotree. The flow source flows are known, so a flow solution is obtained
by finding the flow in the cotree branches m
pL
.
(5)
m
m
m
m
F F
F F
F F
F F
m
m
hT
sT
rT
pT
mL
pL
= −
⋅
11 12
21 22
31 32
41 42
If the network does not contain pipes in the cotree, the flow can then be calculated by Equation (5), as
the vector m
pL
is empty and the vector m
mL
is a boundary condition vector of known values.
3.1.3 Heat solution
The system heat flow is found after the water flow has been found. A flow solution for this network
does exist, and is stored in the n
f
· 1 column vector m. The time dependent flow connectivity matrix
A
f
describes the real connectivity of the flow but not only the connectivity of the flow element. This
matrix does then change if any flowstream in the network does change direction. The flow
connectivity matrix is defined by:
( ) ( ) m A A
f
sign diag ⋅ = (6)
The element flow origin matrix E is defined in terms of the A
f
matrix, by:
( )
E A A
f f
= +
1
2
(7)
Report 13 Şener 9
The entries of the matrix are e
ij
= 1 if the flow stream j does originate at node i, else e
ij
= 0. The fluid
temperature does not change when the fluid goes through an insulated pipe element. The heat
transmitted through the element is only a function of the temperature at the input end of the element,
the heat capacity of the fluid, and the flow itself. The condition at the downstream end of the element
has no influence on the heat flow. The electrical analogy of voltage difference between element ends,
as a driving potential for the current does not apply at all for heat flow in pipe networks.
The heat transported with the flow in a pipe element is calculated by:
q c m T
j p j origin
= ⋅ ⋅ (8)
The origin temperatures for each element can be found from the nodal temperatures by:
T E
origin
T
T
n
= ⋅ (9)
The heat flow for the flow elements is then calculated by:
( )
q m E
f p
T
n
diag c = ⋅ ⋅ T ⋅ (10)
The heat flow through a heat exchanger is a function of four temperatures; the two fluid inlet
temperatures, and the heat transfer coefficient for the exchanger. The counterflow heat exchanger
element is shown on Figure 4.
FIGURE 4: Heat exchanger schematic (Valdimarsson, 1995)
No exchange of mass is between the hot and cold fluid, so this element does not appear at all in the
flow calculation model. The heat exchanger element is presented here as a linear resistance to heat
flow between its connection nodes. The heat flow in a heat exchanger element and the elements
connected to it is shown on Figure 5.
Şener Report 13 10
FIGURE 5: Graph representation of a heat exchanger (Valdimarsson, 1995)
The heat flow in the heat exchanger element is thus only based on the temperatures at the hot fluid
inflow end. The heat flow in each of the heat exchanger elements is calculated by:
( )
q U T T
x eq h in c out
= ⋅ −
, ,
(11)
Each heat exchanger elements is connected to two nodes in the network, and this connectivity is
described by the n
n
· n
x
. heat exchanger connectivity matrix A
x
. The equivalent heat transfer
coefficients are stored in the n
x
· n
x
diagonal matrix U
eq
in the same element order as the
connectivity matrix. The heat flow for the heat exchanger elements in the system is then calculated
by:
q U A T
x eq x
T
n
= ⋅ ⋅ (12)
The connection of the heat flow elements into the network is described by the n
n
· n
q
connectivity
matrix A
q
. The heat flow q
q
for these elements is known, and the connectivity matrix defines at
which nodes the heat is input.
The temperature element has a heat flow sufficient to maintain the desired temperature at the
connection node. This heat flow, q
t
, in the temperature element is unknown. The temperature
element is usually connected to the datum node. All boundary nodes for the flow have to have known
temperature. The connection of the temperature elements into the network is described by the n
n
· n
t
connectivity matrix A
t
.
The matrix notation for the current law of Kirchhoff has one row for each node in the network. A
flow vector multiplied by its connectivity matrix will contribute the correct flow to each node. The
current law for the heat flow is:
  0
q
q
q
q
A A A A =
⋅
q
t
x
f
q t x
(13)
Report 13 Şener 11
The heat flow in the heat sources is known, so the known factors are separated from the unknown:
  A A A
q
q
q
A q
x t
f
x
t
q
⋅
q
= − ⋅ (14)
The heat flow vector q
f
for the flow elements is substituted from Equation (10), and the heat flow
vector q
x
for the heat exchangers from Equation (12):
( )
 
n
T
x
eq
p
x
f
c diag
T A E
U 0
0 m
q
q
⋅
⋅
=
(15)
After this substitution, Kirchhoff's current law is written as:
 
( )
 
A A A
m 0
0 U
E A T
q
A q
x t
p
eq
x
T
n
t
q q
diag c
⋅
⋅
= − ⋅ (16)
The heat flow in the constant temperature elements is unknown, and has to be made a part of the
vector to be solved for:
 
( )
 
A A A
m 0
0 U
E A 0
0 I
T
q
A q
x t
p
eq
x
T
t
n
t
q q
diag c
⋅
⋅
= − ⋅ (17)
In order to get the correct temperature difference for the temperature elements, one line is added to
the set of equations for every temperature element. This new equation equates the temperature
difference between the element ends to the desired numerical value T
t
, according to the Modified
Nodal approach.
 
( )
 
⋅ −
=
⋅
t
q q
t
n
T
t
t
T
x
eq
p
x
c diag
T
q A
q
T
0 A
A A E
U 0
0 m
A A
(18)
The final solution is obtained by matrix inversion, and gives the nodal temperature and the heat flow
in the constant temperature elements as result. (Valdimarsson, 1993)
 
( )
 
T
q
A A
m 0
0 U
E A A
A 0
A q
T
n
t
x
p
eq
x
T
t
t
T
q q
t
diag c
=
⋅
− ⋅
−1
(19)
Şener Report 13 12
3.2 Pipelab district heating simulation programme
Pipelab is the district heating simulation software, which uses graph theory to solve network flow and
heat distribution problems. It is both used during design phase, and while analysing the networks.
Necessary data to model the system in Pipelab is:
1. Nodal coordinates of the system.
2. Start and end nodes of the network elements.
3. Roughness values of pipes (m).
4. Heat loss coefficients of pipes (W/°C).
5. Amount of head supply (m).
6. Required load (flow or heat) at the end points of the system (kg/s or W).
Once the necessary data entered, it is possible to determine the nodal heads, nodal temperatures, and
head and heat loss gradients on the screen as well as in the stored files. Figure 6 shows the
presentation of the Balcova city distribution system in the Pipelab user interface screen.
FIGURE 6: Presentation of Balcova city distribution system in the Pipelab screen all dimensions are in m
4.OPTIMISATION OF PUMP OPERATIONS
4.1 Well pumps
4.1.1 Statement of the problem
The geothermal district heating systems distribute water rather than energy because all cost is directly
related to water usage rather than energy usage (Valdimarsson, 1993). Therefore optimum usage of
pumping power is very important for economic usage of geothermal energy.
In Balcova, heat energy is produced from 8 production wells providing geothermal fluid in different
temperatures and flow rates. In addition to properties shown in Table 1, other properties like
drawdown, pump characteristics also vary well to well. These differences mean that cost of producing
heat energy is different for each well.
Report 13 Şener 13
As stated before, the well pumps are controlled by frequency converters, which provide continuous
control of flow from 0 to maximum flow rate for each well. Capability of controlling flow by using
frequency converters allow the operator to decrease production of geothermal fluid, when it is not
needed, therefore significant amount of pumping energy can be saved. This feature also brings the
problem of, “selecting the best well operating policy” to meet the energy demand of customers.
To provide the most economic operation of these wells there should be geothermal fluid production
strategy targeting minimum cost of production according to changing heat demand of customers.
Unfortunately today frequency converters are adjusted manually to control the pump operations and
there is no production strategy for these 8 wells. Frequencies of the converters are adjusted according
to the experience of the technicians. Naturally, it is almost impossible to find optimum working
conditions without any serious study on this system.
However, finding the best production strategy providing minimum electricity consumption is not
enough, since effects of the pump operations on the hydraulics of the geothermal pipeline system
remain as an unanswered question. Therefore after obtaining the best production strategy, the result
should be simulated in Pipelab, and hydraulics of the geothermal pipeline system should be
investigated.
Each well and its pump have characteristics independent of each other. In order to find the optimum
operation strategy of the system computer programme, all possible combinations are calculated, and
the best option is then selected. Since there is no general solution for this kind of problem, programme
algorithm selecting the best possible option for the minimum power consumption was created in this
study. This programme is connected to Pipelab, therefore the best possible production option
simulated, and effects of pumps on geothermal pipeline system were tested.
4.1.2 Dynamic programming
The following definition of dynamic programming is based on Helmke (1994)
Dynamic programming is a method of optimisation that is applicable either staged processes or to
continuous functions that can be approximated by staged processes. The word “dynamic” has no
connection with the frequent use of the word in engineering technology, where dynamic implies
changes with respect to time.
As a method of optimisation, dynamic programming is not usually interchangeable with such other
forms of optimisation as Lagrange multipliers and linear and nonlinear programming. Instead, it is
related to the calculus of variations, whose result is an optimal function rather than an optimal state
point. An optimisation problem that can be subjected to dynamic programming or the calculus of
variations is usually different from those suitable for treatment by Lagrange multipliers and linear and
nonlinear programming. The calculus of variations is used, for example, to determine the trajectory
(thus, a function in spatial coordinates) that results in minimum fuel cost of spacecraft. Dynamic
programming can attack this same problem by dividing the total path into a number of segments and
considering the continuous function as a series of steps or stages. In such an application, the finite
step approach of dynamic programming is an approximation of the calculusofvariations method.
Dynamic programming can be applied if the problem has four features (Manoutchehr et al., 1971).
1. The problem must be one, which can be divided into stages with a decision required at each
stage.
2. Each stage of the problem must have a finite number of states associated with it. The states
describe the possible conditions in which the system might find itself at any stage of the
problem.
3. The effect of a decision at each stage of the problem is to transform the current state of the
system into a state associated with the next stage.
4. For a given current state and stage of the problem the optimal sequence of decisions is
independent of the decision made in previous stages. A policy is a set of decisions, which
Şener Report 13 14
contains one decision for each state variable for each stage. A policy may also be called a
decision trajectory. The set of states, which results from the application of a policy, is called a
state trajectory or simply trajectory. An optimal policy is the set of decisions that optimises
the objective function, which is a measure of effectiveness.
4.1.3 Formulation of the problem
By using frequency converters, it is possible get continuous flow between minimum to maximum
flow for each well.
n well
m m m m m
i
,..., , ,
2 1 0
= (20)
Heat energy obtained from well can be written as:
) (
return well p well well
T T c m Q − ⋅ ⋅ = (21)
In Balcova, average geothermal fluid temperature at heat exchanger outlets is 60 °C; therefore this
value can be kept constant in calculations.
Energy consumed by well pumps can be found from:
motor pump
pump well
pump
h g m
P
η η ⋅
⋅ ⋅
= (22)
By substituting equation 21 into 22, pump power can be obtained as a function of produced heat from
well.
) (
well pump
Q f P = (23)
For all system, total pump power can be written as;
) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (
11 11 10 10 5 5 4 4 7 7 6 6 4 4 2 2 B B B B B B B B BD BD BD BD BD BD BD BD total
Q f Q f Q f Q f Q f Q f Q f Q f P + + + + + + + =
(24)
At any time to meet the heat demand of the system, total heat production from wells must be equal or
bigger than heat demand of the customers.
demand B B B B BD BD BD BD
Q Q Q Q Q Q Q Q Q ≥ + + + + + + +
11 10 5 4 7 6 4 2
(25)
The performance criterion, which is to be minimised, is
+ + + + + + + =
∑
) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ( min
11 11 10 10 5 5 4 4 7 7 6 6 4 4 2 2 B B B B B B B B BD BD BD BD BD BD BD BD total
Q f Q f Q f Q f Q f Q f Q f Q f P
(26)
Then the programme should satisfy (25) according to the criterion stated in the (26). The results of the
programme were given in section 5 of the report.
4.2 Circulation pumps
In the Balçova city distribution system circulation of water is provided by 4 identical centrifugal
pumps, which are in the main pumping station. These pumps are connected in parallel. In heating
session 3 pumps work and one serves as a back up in the system. In Figure 7, performance curve of
single pump, used in the system is given for the pump test speed of 1450 rpm.
Report 13 Şener 15
0
20
40
60
80
100
120
140
0 50 100 150 200 250 300
Flow rate (l/s)
H
e
a
d
(
m
)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
E
f
f
i
c
i
e
n
c
y
FIGURE 7: Performance curve of circulation pump at 1450 rpm
The pump performance curve shown in Figure 7 is valid only if the pump is run at 1450 rpm. Pump
performance curves for different speeds can easily be derived from affinity laws, given below.
2
1
2
1
N
N
V
V
p
p
= (27)
2
2
1
2
1


.

\

=
N
N
h
h
p
p
(28)
Frequency converters can drive the pumps over a wide range of flows, however variablespeed pumps
do not run efficiently over a wide range of flows. Therefore, to minimise the power consumption of
circulation pumps it is important to know efficiency characteristics of pumps for varying flows.
Efficiency curve of pump at test speed is usually given by third degree polynomial;
V c V b V a
test pump
⋅ + ⋅ + ⋅ =
2 3
_
η (29)
If the speed of pump is changed,
n=(pump speed)/(pump test speed) (30)
then the new pump efficiency can be found from (Walski et al., 2002):
) / ( ) / ( ) / (
2 3
n V c n V b n V a
pump
⋅ + ⋅ + ⋅ = η (31)
If the pump station is to be operated such that different combinations of pumps will be run under
different demand conditions, it is important that the pumps be selected to work efficiently when
operating both alone and in parallel with the other pumps. Therefore, while developing pump
operation strategy operator should consider the different combinations of pumps, as well as varying
flows. In this study, all possible combinations of pump operations in Balçova were investigated and,
efficiency characteristics of different combinations at different speeds were compared. Finally, best
combination of parallel pumps was determined over a range of flow for minimum power
consumption.
Şener Report 13 16
5. MODELLING OF HEAT DEMAND BY USING TIME SERIES ANALYSIS
The term time series refers to an ordered sequence of values of a variable at equally spaced time
intervals. The usage of time series models is twofold:
• Obtain an understanding of the underlying forces and structure that produced the observed data
• Fit a model and proceed to forecasting, monitoring or even feedback and feedforward control.
In district heating systems main goal of using this model is to estimate heat load of the system
according to weather forecast, so that the heat demand of the customer can be met at right time.
In Figure 3, the points 1 and 2 indicate the locations where the temperature, pressure and flow rate of
city circulation water is measured hourly. By combining these system measurements with hourly
outdoor temperature measurements, data set, which is suitable for the time series analysis, can be
obtained.
Heat demand behavior of the system, is actually the reaction of the system to the outside temperature
changes. Therefore it is important to determine the most obvious relation between measured system
parameters and outside temperature.
In Figure 8, system heat load is plotted as a function of outside temperature by using 4500 hours
continuous data of 20012002 heating session. There is an obvious trend of the heat load according to
changing outside temperature.
FIGURE 8: Outside temperature vs. system heat load
Time series analysis of this data is done by using ready functions in Matlab. First, a linear prediction
model, AutoRegressive with exogenous inputs (ARX) technique, is estimated using a time series
recorded. Then, the residual error, which is the difference between the actual time measurement and
the prediction from the previously estimated ARX model, is defined. The results are presented in
Section 6.
Report 13 Şener 17
6. RESULTS
6.1 Distribution system
6.1.1 Results of the simulation
In the Pipelab district heating simulation program, the distribution system has been modelled between
the peak flow demand of 305 l/s total flow rate and minimum flow demand of 17 l/s. During
modelling, necessary information of the system was provided by the Balçova district heating
company. It is assumed that hot water is delivered to all buildings in a sufficient amount, to meet their
energy requirements. While modelling the system, it is divided into two parts, as supply and return
networks. In Figure 9, combined head loss versus distance from the head source diagram, which is
obtained from Pipelab, is shown for the supply and return network.
FIGURE 9: Length from source vs. head loss diagram for the Balcova distribution system
As can be seen from the Figure 9, the maximum head loss along the pipes of the distribution system is
43 m for the maximum flow rate. For the critical head loss path, heat exchanger and flow regulator
pressure drop values has been taken as 4 m.
The distribution system has been simulated in Pipelab for the range of flow demands and, system loss
curve for the critical path has been obtained. It should be stressed that, these head loss diagram does
not include the head losses in the pumping station.
Şener Report 13 18
0
5
10
15
20
25
30
35
40
45
0 50 100 150 200 250 300 350
Flow rate (l/s)
H
e
a
d
l
o
s
s
FIGURE 10:Pressure loss curve for the critical path of city distribution loop
FIGURE 11: Length from source vs. node temperature diagram for the Balcova distribution system
The temperature decrease of the distribution system for the supply network is shown on Figure 11. As
can be seen from the figure, temperatures of the nodes are in the range of 85 °C to 80 °C except for
one branch.
6.1.2 Problems of distribution system
During peak demand times certain parts of the system cannot get sufficient heat. These buildings are
shown by dots in the Figure 12. Although heating problems seem to occur mostly in one region (I and
II), it will be wise to investigate these parts according to their branching structures.
Report 13 Şener 19
FIGURE 12: Buildings with heating problem
As shown on Figure 12, there are three main regions where the heating problems occur. These three
regions constitute the 80% of the heating problems in the system. Among these three regions, region I
contain most of the buildings with a heating problem.
In Figure 13 these three branches, which are highlighted and taken into circle, are shown in the H vs.
L diagram of supply network. As can be seen from the diagram region I, which is the most
problematic part of the system, has the highest pressure drop in the network. Also, as shown in Figure
14, the largest temperature drop in the network belongs to one of the branches of region 1.
FIGURE 13: Presentation of regions with heating problem in hl diagram of supply network
Şener Report 13 20
FIGURE 14: Presentation of regions with heating problem in Tl diagram of supply network
Although region I and region II are located close to each other, they belong to different branches of
the system and different from region I, region II is not one of the lowpressure zones in the system.
While investigating this branch of the system, it should be noted that, it has the highest distance to the
pumping station
In Figure 13 there is steep pressure decrease for the region III. It is because of high speed of water at
that part of the network. Highspeed flow generally occurs at undersized distribution branches.
6.1.3 Possible sources of the problems
6.1.3.1 Region I
Region I, which has the highest pressure drop on the head loss diagram contains most of the buildings
with a heating problem. If outside temperature gets closer to the 0 °C which is the design temperature
for Izmir City, customers living in the region I start to complain about heating service.
From temperature and head loss diagrams, it is seen that this region has the highest temperature and
pressure drop in the system. Therefore, if there is lowpressure problem in the system, region I will be
affected by it at the first place.
One of the most frequently occurring operational problems associated with water distribution systems
is low or fluctuating pressure. While confirming that the problem exists is usually easy, discovering
the cause and finding a good solution can be much more difficult. According to Walski (2002),
customer complaints, modelling studies, and field measurements obtained through routine checks can
indicate that a portion of the system is experiencing low pressure. The pressure problem can be
verified by connecting a pressure gage equipped with data logging device or chart recorder to a
system continuously record pressure. Occasionally, a customer may report a lowpressure problem,
but the pressure at the main is fine. In such cases, the low pressure may be due to restriction in the
customer's plumbing, or a pointofuse/pointofentry device that is causing considerable head loss. If
measurements indicate that pressure in the main is low and a problem in the distribution system is
suspected, the next step is to examine the temporal nature of the problem. Pressure drops that occur
only during periods of high demand are usually due to insufficient pipe or pump capacity, or a closed
valve.
Report 13 Şener 21
As stated before, pressure measurements are done at two points in the city distribution loop. These
locations are points 1 and 2, which are shown on Figure 3. The water pressure at point 1 (Figure 3),
should be at sufficient amount to provide the circulation of water until point 2 (Figure 3). In Figure
15, variation of actual pressure difference between points 1 and 2 (Figure 3) is shown by dots. To
compare the actual data, with the simulation result system characteristic curve obtained from
simulation (Figure 10), is given by continuous curve in Figure 15. Values obtained from simulation
are accepted as a minimum required head difference (Between 1 and 2), to provide the circulation in
the system at any flow rate. Comparing actual measurements, with the required head value obtained
from simulation, the points under the curve indicates the existence of insufficient pressure difference
in the system. It can also be observed that the distribution of points under the curve are homogeneous,
which means that, the problem is not related with pump capacity. The possible sources of problem
are:
• Lack of coordination between operation of expansion tank and circulation pumps.
• Flow regulators cannot create enough resistance to increase the pressure difference between 1
and 2 (Figure 3).
• Speed of operator response to the system changes is slow.
0
1
2
3
4
5
6
0 200 400 600 800 1000 1200
Flow (m
3
/h)
H
e
a
d
l
o
s
s
(
b
a
r
s
)
FIGURE 15: Comparison of actual head difference and required head difference (Between 1 and 2 in Figure 3)
To overcome this problem;
• Operator should watch for the minimum head difference criterion between 1 and 2 (Figure 3).
• Circulation pumps should be operated according to future estimates of heating demand so that
system pressure loss can be kept above the allowable level.
• Set points of the flow regulators should be revised.
By investigating region I and its branches in Figure 13, it can be seen that there is no steep pressure
gradient for the pipes, which carry water to the region I. Therefore undersized piping is not one of the
sources of problems in this region.
Şener Report 13 22
In Figure 14, there is steep temperature decrease for the region I. The reason of high temperature
decrease is low speed of water at this area. Decreasing the size of the piping at this area can prevent
the temperature drop. However, decreasing size of piping increase the pressure loss of the system,
since this region already has the highest pressure drop, decreasing size of piping is not a cost effective
solution.
Once the cause of the pressure and temperature problem has been identified and confirmed, the
possible solutions are usually fairly straightforward, and include the following. (Walski et al., 2002)
• Changing pump control settings
• Locating and repairing any leaks
• Implementing capital improvement projects such as constructing new mains
• Installing pumps to set up a new pressure zone
• Installing storage tank
Each of these options affects the system in different ways and has different benefits, so the
comparison of alternatives should be performed based on a benefit/cost analysis as opposed to simply
minimizing costs. The modelling for this evaluation can usually be performed with a steady state
modelling.
Among these solutions installation of booster pump is usually the least costly method of correcting
lowpressure problems from an initial capital cost standpoint. Booster pumps can however
significantly increase operation and maintenance costs, and do not allow as much flexibility in terms
of future expansion as the other available options. Also, booster pumps can over pressurize portions
of the system and even cause water hammer, especially when there is no downstream storage or
pressure relief. It should be remembered that pumping increases the pressure at the pump location but
does not reduce the hydraulic gradient. (Walski et al., 2002)
Adding storage at the fringe of the system tends to be a costly alternative, but it provides the highest
level of benefit. Storage increases the reliability of the system in the event of pipe break or power
outage, and helps to dampen transients.
The first option is to focus on is pump and expansion tank settings, since the pumps have sufficient
capacity to meet maximum required pressure at the system, system control strategy should be re
considered under the light of simulation results. Although, lowpressure problem can be overcome by
changing pump and expansion tank settings, the low temperature problem is not that easy to solve.
Since the problem is related with low velocity of water at that area. In Figure 16, enlarged view of
region I is given. From Figure 16, it is seen that the lowest pressure and the lowest temperature drops
don’t occur at the same branch of the system. Branch with the lowest temperature is taken into circle
in Figure 16. Temperature drop of this branch from start to end is show in Figure 17. As can be seen
from Figure 17, only last 10 nodes have temperatures lower than 80 °C. Low temperature at these 10
nodes affect the last 6 buildings connected to the system.
Report 13 Şener 23
68
70
72
74
76
78
80
82
84
86
1400 1405 1410 1415 1420 1425
Node number
W
a
t
e
r
t
e
m
p
e
r
a
t
u
r
e
(
°
C
)
FIGURE 16: Enlarged view of region I
FIGURE 17: Temperature drop of indicated branch
To see the effect of changing piping size on system, the diameter of pipes, which are taken into circle
in Figure 16, were decreased from 50 mm to 32 mm and 25 mm to 20 mm. New temperature drop
curve is presented at Figure 18. It can be seen from figure that, although there is increase at the end
node of the branch, it is not in significant amount. Resizing this branch is not a cost effective method
to increase temperature drop problem in this branch.
70
72
74
76
78
80
82
84
86
1400 1405 1410 1415 1420 1425
Node number
W
a
t
e
r
t
e
m
p
e
r
a
t
u
r
e
(
°
C
)
79
80
81
82
83
84
85
1390 1400 1410 1420 1430
Node number
W
a
t
e
r
t
e
m
p
e
r
a
t
u
r
e
(
°
C
)
FIGURE 19: Temperature drop after increasing flow FIGURE 18: Temperature drop after decreasing the piping size
To outcome the problem of excess temperature drop, changing the settings of flow controllers at this
branch was also considered. Flow capacities of last 6 buildings in this branch were increased by 60%.
Results of simulation were given in Figure 19. By changing flow capacities of these 6 buildings,
temperature drop has been decreased to acceptable levels.
In Figure 20 head loss results of increasedflow simulation were presented, an as can be seen from
figure, head loss diagram changed drastically. Increase of flow at this branch, which is taken into
circle in Figure 20, made the branch have the highest pressure drop in the system. There is about 5 m
increase in the supply network maximum head loss value, which is the required pumping head for the
circulation pumps. Therefore, it is obvious that increasing the flow capacity of the branch is not a cost
effective solution, since it increases the pumping costs.
Şener Report 13 24
FIGURE 20: Pressure drop curve after increasing the flow capacity of the branch
As a result, to overcome the problems in this region first item to focus is pumping the necessary head
required by the system. It will drastically decrease the number of buildings with a heating problem.
Also changing the pipes with better insulated ones can prevent excess heat loss of water in the pipes.
Also while changing the pipes in that branch, new pipes diameters should be selected smaller than
existing ones.
It should be stressed that, the results of simulation changes according to roughness values and heat
loss coefficients of pipes. In Figure 21, variation of maximum temperature drop in the system
compared for the good insulation and bad insulation. It is obvious that temperature drop changes
significantly according to insulation quality. However, the trend of temperature decrease does not
change. Simulating the system by assuming pipes have good insulation, results in the highest
temperature drop in the same branch as for the less insulated case. Therefore, if there is excess
temperature drop problem in the system this branch is at the first place to be affected by this problem.
It should also be considered that customer complaints have been reported from buildings, which are
located at the end of highest temperature drop branch. Also close investigation of Figures 12,16 and
17 will give the result of obvious existence of high temperature drop in this branch. Therefore, with
the actual temperature measurements taken from system, heat loss coefficients used in the model
should be calibrated and the possible solution methods discussed in this study should be reconsidered.
0
5
10
15
20
25
30
35
40
45
0 50 100 150 200 250 300 350
Flow rate (l/s)
T
e
m
p
e
r
a
t
u
r
e
d
r
o
p
(
°
C
)
FIGURE 21: Maximum temperature drop in the system according to insulation quality
Report 13 Şener 25
6.1.3.2 Region II
The most significant characteristic of region II is, having the highest distance to the pump station.
From the head loss diagram (Figure 13) of this region, it can be said that, region II does not
experience lowpressure problem. Also even for the worst insulation of piping, simulation result for
maximum temperature drop at this region is only 5°C, which is sufficient for proper operation of
heating systems. However this region still contains buildings with a heating problem. Considering
distance between pump station and region II, it can be said that this region may experience low flow
during peak flow demands.
Low flow during peak demand times, are another common operational problem. Solving this problem
in an existing system is different than designing pipes for new construction, in that the utility cannot
pass the cost of improvements onto a new customer. Rather, the operator must find the weak link in
the system and correct it.
The possible reasons for poor flow during peak demand in an existing system are (Walski et al., 2002)
• Small mains
• Longterm loss of carrying capacity due to tuberculation and scaling
• Customers located far from the source
• Inadequate pumps
• Closed or partly closed valves
• Improper setting of flow regulators in that branch
• Some combination of above
As stated before, in Balçova district heating system flow regulator valves play very important role for
the hydraulic balance of the system. Each flow regulator should be adjusted according to heat demand
of that specific customer. Change of setting point in one regulator affects all system, as explained in
section 2.2.2. Therefore, while adjusting new setting point for one building, other buildings at that
branch should be considered.
In the case of region II, this problem becomes obvious. Other buildings which are connected to the
system before region II get hot water more than necessary and the buildings at the end point cannot
get sufficient flow for heating. Also leakage detection efforts should be focused on this region, since
there is 5 l/s average leakage, which has not been positioned yet.
6.1.3.3 Region III
In Figure 13, region III has highpressure loss gradient. Main reason of highpressure gradient is high
speed of flow in this branch. This kind of problems occurs at undersized distribution networks.
An undersized distribution network problem is not easy to identify during averageday conditions. If a
pipe is too small, it may become a problem only during high flow conditions (Walski et al., 2002).
Therefore peak flow simulations are the best way to identify an undersized distribution network
Sizing new piping and rehabilitating existing pipes flattens out the slope of the hydraulic gradient for
a given flow rate. By investigating head loss diagrams of the system, pipes, which need repair or
rehabilitation, can be located.
To see the effect of diameter change on region III, diameters of pipes, which are shown in the figure
22, has been increased 80 mm to 100 mm and 65 mm to 80 mm in the Pipelab simulation programme.
Result is shown in the Figure 23. Head loss gradient for the region III has decreased significantly.
Therefore, it is for sure this part of network contains undersized pipes, which should be changed for
efficient operation of the system.
Şener Report 13 26
FIGURE 22: Resized pipes in region III FIGURE 23: Head loss diagram for resized pipes in region III
6.1.4 Pump control strategy for city circulation loop
Distribution system and characteristics of circulation pumps are explained in sections 2.2.2 and 4.2. In
this section, best pumping strategy for the minimum energy consumption according to the results of
the simulation will be determined.
To provide the circulation of water in the system, pumps must supply enough pressure to water. In
Figure 24, basic pressure loss scheme of the system is show. Required pumping head (H
pump
) is shown
in the figure. Since the pressure loss values of distribution network have already been calculated by
using simulation, the only parameter to add simulation results is main heat exchanger pressure loss in
the pumping station.
FIGURE 24: Pressure loss diagram of distribution system
Report 13 Şener 27
In figure 25, pressure at the outlet of heat exchanger in the pumping station is given as a function of
heat exchanger inlet pressure. Figure 25 has been obtained by hourly measurements of heat exchanger
inlet and outlet pressures. As can be seen from Figure 25, heat exchanger outlet pressure values
fluctuate between 4 bars and 6 bars with average of 4.8 bars. Therefore heat exchanger inlet pressure
(pump head) has no effect on heat exchanger outlet pressure of the system. One of the important
reasons of fluctuations of heat exchanger outlet pressure is changing water temperatures in the heat
exchanger.
0
1
2
3
4
5
6
7
5 6 7 8 9 10
Heat exchanger inlet pressure (bars)
H
e
a
t
e
x
c
h
a
n
g
e
r
o
u
t
l
e
t
p
r
e
s
s
u
r
e
(
b
a
r
s
)
11
FIGURE 25: Pressure difference between heat exchanger inlet and outlet
Since there is no effect of running pumps at pressures higher than 6 bars, pump analysis were done by
keeping pump head of 6 bars constant.
FIGURE 26: Power consumption of circulation pumps for different combinations
Şener Report 13 28
In Figure 26, power consumptions of 3 different combinations were given. As can be seen from figure
until about 100 l/s flow, operation with one pump is the most economic solution, then between 100 l/s
and 150 l/s operation of two pump gives the minimum power consumption. Finally for the flow rates
higher than 150 l/s operation of three pumps is the most economical option. It should be noted once
more that Figure 26 has been obtained by assuming that pump head is kept constant at 6 bars. Since
increase in the pump head is compensated by heat exchanger, it is obvious that there is no need to
operate pumps at higher heads. In Figure 27, heat exchanger pressure drop is given as a function of a
pump head. In Figure 27, which is plotted by using actual data, compensation of excess pump head by
heat exchanger is clearly seen.
0
1
2
3
4
5
6
3 4 5 6 7 8 9 10 1
Pump head (bars)
H
e
a
t
e
x
c
h
a
n
g
e
r
p
r
e
s
s
u
r
e
d
r
o
p
(
b
a
r
s
)
1
FIGURE 27: Heat exchanger pressure drop as a function of pump head
6.2 Geothermal pipeline system
Structure of geothermal pipeline system and problem of finding the minimum well pump operation
policy has been explained in section 2.2.1 and 4.1 respectively. The results of programme are given in
Table 4. According to requirements, programme input parameters can be changed so that, best
combination for other heat loads can be found. While choosing the best option among other
alternatives, continuation of operation for same pump has been considered.
TABLE 4: Well operation policy for different heat loads
Flow (kg/s)
Power cons. (kWe)
Heat production
(kWth)
Total flow (kg/s)
BD2 BD4 BD6 BD7 B4 B5 B10 B11
371 49043 210 22 39 33 17 14 42 23 20
309 44750 181 22 39 33 12 5 42 16 12
258 40239 168 17 28 33 12 8 42 16 12
213 35806 151 17 28 24 7 5 42 16 12
172 31371 122 17 28 24 7 0 30 16 0
138 26796 106 17 28 15 0 0 30 16 0
108 22350 88 17 18 15 0 0 30 9 0
82 18461 70 8 18 15 0 0 30 0 0
59 13701 55 8 18 0 0 0 30 0 0
40 9299 43 13 0 0 0 0 30 0 0
26 4637 18 13 0 0 0 0 5 0 0
Report 13 Şener 29
In Figure 28, presentation of geothermal pipeline system in Pipelab is shown.
FIGURE 28: Geothermal pipeline system all dimension are in m
In Figure 29, results of the simulation for 49043 kW
th
heat production (Table 4) is given. Thick lines
in the Figure 29 are main pipeline and other lines are connections to production and consumption
points. As can be seen from figure, distribution of geothermal water temperature is sufficient for the
operation of heat exchanger stations in the system.
FIGURE 29: Temperature distribution along geothermal pipeline
Şener Report 13 30
FIGURE 30: Pressure distribution along geothermal pipeline
In Figure 30, pressure distribution of the pipeline system from well heads to heat exchangers is shown
for maximum heat production (Table 4). The largest pressure need to pump the water into the system
is for BD4. Operation of well head pressures should be done according to head loss diagram of the
pipeline system.
6.3 Estimation of heat demand
As stated in section 5, time series analysis of hourly measurements taken from pumping station has
been done, to fit estimation model for a heat load of the system. Data used in this analysis are:
• Daily average of flow of circulation water.
• Daily average of outside temperature.
• Daily average of water temperature at the pumping station inlet.
• Daily average of water temperature at the pumping station outlet.
Data belongs to 188 days of 20012002 heating session. Original outputs of the programme, which
uses the ARX function from Matlab’s System Identification Toolbox to create a prediction model, are
given in Figure 31.
Figure 31 can be rewritten as:
E T D T T B q A q
outside outside outside today tomorrow
+ ⋅ + ⋅ + ⋅ + ⋅ =
1 2 3
where:
A= ) 03696 . 0 8125 . 0 ( ±
B= ( ) 34 . 46 3 . 197 ± −
C= ) 85 . 59 9 . 148 ( ± −
D=( ) 32 . 52 84 . 71 ±
E= ) 1274 6714 ( ±
Report 13 Şener 31
Discretetime IDPOLY model: A(q) y(t) = B(q)u(t) + e(t)
A(q) = 1  0.8125 (+0.03696) q^1
B1(q) = 197.3 (+46.34)  148.9 (+59.85) q^1 + 71.84 (+52.32) q^2
B2(q) = 6714 (+1274)
Estimated using ARX
Loss function 1.64798e+006 and FPE 1.73804e+006
Sampling interval: 1
Created: 07Oct2002 23:57:05
Last modified: 07Oct2002 23:57:05
FIGURE 31: Original programme output for time series analysis
Success of model has been has been tested on two separate data sets. The first data set belongs to
heating season of 20012002, from which model has been derived. In Figure 32, estimated heat loads
are shown by dots while variation of real load is given by continuous curve. It should be noted that,
while deriving model only the data of 20012002 heating session is used. Therefore, success of model
can be checked by using another data set and comparing the actual values with the model. In Figure
33, measurements of 20002001 heating season were used. Estimated heat load is shown by dots and
real heat load is shown by continuous curve in Figure 33. As can bee seen from figure, results of the
prediction model and actual heat load variation are fitting each other. Comparison of results shows
that using a prediction model to predict the head load demand of next day can give rather close
estimation of the actual value.
0 20 40 60 80 100 120 140 160 180 200
0
0.5
1
1.5
2
2.5
3
3.5
x 10
4
Days
H
e
a
t
d
e
m
a
n
d
(
k
W
)
FIGURE 32: Comparison of real heat load with estimations (20012002 session)
Şener Report 13 32
0
5000
10000
15000
20000
25000
30000
0 5 10 15 20 25 30 35
Days
H
e
a
t
l
o
a
d
(
k
W
)
FIGURE 33: Comparison of real heat load with estimations (20002001 session)
7. CONCLUSION
The main goal of this study was to simulate the Balcova geothermal distribution system, and by
considering results of simulation determine the operation strategy for the system. Simulation of city
circulation and geothermal distribution networks were achieved by using Pipelab district heating
simulation programme. During modelling phase, system data taken from the Balcova geothermal
company was used. However roughness values and heat loss coefficients of pipes were estimated by
considering the catalogues of district heating pipes in the market, since there is no existing actual
data.
The primary objective of a simulation is to reproduce the behaviour of a real system in a useful way.
To achieve this aim, data are supplied that depict the physical characteristics of the system, the loads
placed on the system, and the boundary conditions in effect. Even if all the data gathered describing
model match the real system exactly, it is unlikely that the pressures and flows computed by the
simulation model will absolutely agree with observed pressures and flows. Modelling is essentially a
balance between reality, a simulated reality, and the effort necessary to make the two agree (Walski et
al., 2002).
By looking at Figure 15 where the actual system pressure loss is compared with simulation result, it
can be said that results of simulation follows the trend of actual system behaviour. It should also be
noted the most problematical buildings according to company record experience high pressure or
temperature drops in the simulation. This is also one of the indications of success of the simulation.
Report 13 Şener 33
Roughness values used in the
simulation are for new pipes, however
the system is 5 years old. Therefore it
is for sure that actual roughness values
are bigger than the one used in
simulation. In Figure 33 variation of
maximum head loss of supply network
with roughness value is given.
According to Walski (2002) two pipes
of the same size, material, and age can
have different effective diameters and
roughness based on the quality of the
water historically flowing through pipe.
Although results of simulation give
quite reasonable results, further study
about calibration of roughness values in the model, will improve the accuracy of results obtained from
system. In addition to changing roughness values by time, formation of scale in the pipes is not a low
possibility since water in the circulation loop contains high amounts of calcite. Scaling which occur
over time after the pipe has been installed, creates the difference between actual and nominal
diameters. Reduction of diameters in the system creates higher pressure losses in the system. Since
allowable pump head to provide circulation in the system is at very critical level, effect of these two
factors (roughness and scaling) on the system should be investigated to provide insufficient pumping
operations.
15
17
19
21
23
25
27
29
0 0.00005 0.0001 0.00015 0.0002 0.00025
Roughness (m)
S
u
p
p
l
y
n
e
t
w
o
r
k
h
e
a
d
l
o
s
s
(
m
)
FIGURE 34: Variation of maximum head loss by roughness
Leakage is also one of the problems in Balcova district heating system. Amount of leakage is
changing according to system operation conditions. During peak demand times leakage reaches up to
5 l/s. Although this amount is small compared to total flow of the system, effect of the leakage can be
much bigger than its value depending on where it is placed. It is general approach to distribute
leakage homogenously in the system during simulation, when the location of leakage is unknown.
Results of the simulation should be investigated and compared with actual measurements taken from
system, so that simulation can be used as a very efficient tool to locate leakages.
By investigating system head loss diagrams and available heads supplied from pumps to circulate in
the city, it can be seen that available head is at very critical level. Since the region having highest
head loss in the system has continuous problems about insufficient heating, it is for sure that there is
insufficient problem in the system. Since the available head supplied from pump station cannot be
increased other solution techniques 6.1.2 should be considered. However it should also be taken into
account that there is a leakage in the system, and leakage detection efforts should be focused on this
region.
Results of programme finding the optimum well pump operation strategy were given in section 6.2.
Programme determines the right combination of wells that should be operated to get minimum power
consumption while meeting heat demand of the system. The results then used as an input parameter
for the pipelab simulation programme and the interaction of wells to each other checked. From the
results of the geothermal pipeline simulation, it can be said that the most important parameter to
control system is well head pressure valves. Since the requirement of head to pump geothermal fluid
into the system is different for each well. Therefore, pressure between pipeline and well should be
under control to provide sufficient head to the system.
The ARX prediction model provides the heat load estimation of one day later. As can seen from
comparison of the model results with actual measurements, model gives the very accurate estimation
of heat load of one day later. Therefore, the results of the model should be considered while
determining the pumping strategy of next day. In Figure 34, suggested control diagram of system is
Şener Report 13 34
shown. By using estimated heat load of tomorrow, system operation strategy (Heat exchanger
temperatures, pump settings, valve arrangements) can be determined.
FIGURE 35: Suggested operation strategy for the system
It should always be remembered simulation programs cannot reproduce the actual operation
conditions exactly. Using simulation programme for this kind of system is a continuous process of
calibration. Accuracy of simulation can only be increased with more data. Since it is impossible to
place measurement devices at every point of the system, to provide the necessary data to the
programme, mobile measurement devices can be good solution to know more about the actual system.
ACKNOWLEDGEMENTS
I would like to express my gratitude to Dr. Ingvar Fridleifsson for giving me the opportunity to
participate in the UNU Geothermal Training Programme in Iceland. I am very grateful to Mr. Lúdvik
S. Georgsson and Gudrún Bjarnadóttir for kind and patient help during last six months. Their great
organization skills made life easier to me.
Thanks to my professors in İzmir Institute of Technology Geothermal Energy Centre. Dr. Macit
Toksoy he introduced the area of geothermal energy and offered this project to me. Dr. Zafer İlken as
a Director of Geothermal Energy Centre he provided me necessary permissions to come Iceland. Dr.
Gülden Gökçen, as a supervisor of my M.Sc. project she held very valuable lectures for me during last
two years.
Great appreciation should be expressed to Balçova Geothermal Company, and chief engineer Cihan
Çanakçı for his tireless efforts to create the system database and permission to use unpublished
system data.
Special and final thanks go to great teacher and advisor, Páll Valdimarsson, for his guidance and
patience during last six months. His enormous enthusiasm to teach motivated me every new morning.
Report 13 Şener 35
NOMENCLATURE
Scalars
A = Heat exchanger area [m
2
]
a
ij
= Connectivity matrix entry []
a, b, c = Pump efficiency curve coefficients []
c
p
= Water heat capacity [J/(kg °C)]
G
s
= Subgraph
G
n
= Graph
g = Acceleration due to gravity [m/s
2
]
h
pump
= Pump head [m]
T = Tree
L = Cotree (the set of links)
m = Water flow [kg/s]
m
h
= Flow of hot fluid [kg/s]
m
well
= Well flow [kg/s]
N = Pump rotational speed [rpm]
n
L
= Number of links []
n
n
= Number of nodes []
n
q
= Number of constant heat flow elements []
n
T
= Number of tree branches []
n
t
= Number of constant temperature elements []
P
pump
= Pump power consumption [W]
P
total
= Total pump power consumption [W]
Q
well
= Heat energy produced from well [W]
q = Heat flow [W]
q
q
= Constant heat flow [W]
q
t
= Heat flow in constant temperature element [W]
q
x
= Heat exchanger duty [W]
T = Temperature [°C]
T
c,in
= Cold fluid inlet temperature [°C]
T
c,out
= Cold fluid outlet temperature [°C]
T
h,in
= Hot fluid inlet temperature [°C]
T
h,out
= Hot fluid outlet temperature [°C]
T
well
= Geothermal fluid temperature at well head [°C]
T
return
= Geothermal fluid temperature at heat exchanger outlet [°C]
U = Heat transfer coefficient [W/(m
2
°C)]
U
eq
= Equivalent heat transfer coefficient [W/°C]
V = Volumetric flow rate [l/s]
Greek Symbols
∆T
m
= Logarithmic mean temperature difference [°C]
η
pump
= Pump efficiency
η
motor
= Motor efficiency
Şener Report 13 36
Vectors and Matrices
A = Flow elements connectivity matrix []
A
f
= Flow connectivity matrix []
A
L
= Cotree connectivity matrix []
A
q
= Constant heat flow connectivity matrix []
A
T
= Tree connectivity matrix []
A
t
= Constant temperature connectivity matrix []
A
x
= Heat exchanger connectivity matrix []
D = Cutset matrix []
E = Element flow origin matrix []
I
hT
= Tree head source identity matrix
I
sT
= Tree storage tank identity matrix
I
rT
= Tree resistor identity matrix
I
pT
= Tree pipe identity matrix
i = Index vector
m = Flow vector [kg/s]
m
hT
= Tree head source flow vector [kg/s]
m
mL
= Link flow source flow vector [kg/s]
m
pL
= Link pipe flow vector [kg/s]
m
pT
= Tree pipe flow vector [kg/s]
m
rT
= Tree resistor flow vector [kg/s]
m
sT
= Tree storage tank flow vector [kg/s]
F
ij
= Submatrix of the cutset matrix
q
f
= Vector of heat flow in flow elements [W]
q
q
= Constant heat flow vector [W]
q
t
= Vector of heat flow in constant temperature elements [W]
q
x
= Heat exchanger duty vector [W]
T
n
= Node temperature vector [°C]
U
eq
= Heat exchanger transfer matrix [W/°C]
Report 13 Şener 37
REFERENCES
Battocletti, L., 1999: Geothermal resources in Turkey. Bob Lawrence & Associates, Inc., USA,72 pp.
Helmke, U., Moore, J.B., 1994: Optimization and dynamical systems. SpringerVerlag, London, 389
pp.
Gökçen, G., 1999: History of Balçova geothermal field. Unpublished study. IZTECH, Turkey, 10 pp.
Manoutchehr, H., Chow, V.T., and Dale, D.M., 1971: Water resources systems analysis by discrete
differential dynamic programming. Civil Engineering Studies, Hydraulic Engineering Series No:24.
University of Illinois, Urbana, 118 pp.
Valdimarsson, P., 1995: Graphtheoretical calculation model for simulation of water and energy flow
in district heating systems. 5
th
International symposium on automation of district heating systems.
Nordic Energy Research Program,39.
Valdimarsson, P., 1993: Modelling of district heating systems. Ph.D. thesis. University of Iceland,
Reykjavík 144.
Walski, M.T., Chase, V.D., Savic, A.D., 2002: Water distribution modelling. Haestad Press,
Waterbury, CT, U.S.A, 441 pp.
Şener Report 13 38
TABLE OF CONTENTS
ABSTRACT.................................................................................................................................. 1
1.INTRODUCTION...................................................................................................................... 1
2.BALCOVA GEOTHERMAL DISTRICT HEATING SYSTEM.............................................. 2
2.1 Brief description and history of Balcova geothermal field................................................. 2
2.2 Presentation of existing geothermal utilization in Balcova................................................ 3
2.2.1 Geothermal pipeline system....................................................................................... 3
2.2.2 City distribution system.............................................................................................. 5
3. MODELLING OF DISTRICT HEATING SYSTEMS ............................................................ 6
3.1 Theory behind the model.................................................................................................... 6
3.1.1 Graph theory............................................................................................................... 6
3.1.2 Flow solution.............................................................................................................. 7
3.1.3 Heat solution............................................................................................................... 8
3.2 Pipelab district heating simulation program..................................................................... 12
4. OPTIMISATION OF PUMP OPERATIONS......................................................................... 12
4.1 Well pumps ...................................................................................................................... 12
4.1.1 Statement of the problem ......................................................................................... 12
4.1.2 Dynamic programming............................................................................................. 13
4.1.3 Formulation of the problem...................................................................................... 14
4.2 Circulation pumps ............................................................................................................ 15
5. MODELLING OF HEAT DEMAND BY USING TIME SERIES ANALYSIS.................... 16
6. RESULTS................................................................................................................................ 17
6.1 Distribution system........................................................................................................... 17
6.1.1 Results of the simulation .......................................................................................... 17
6.1.2 Problems of distribution system............................................................................... 18
6.1.3 Possible sources of the problems.............................................................................. 20
6.1.3.1 Region I ............................................................................................................ 20
6.1.3.2 Region II........................................................................................................... 25
6.1.3.3 Region III ......................................................................................................... 25
6.1.4 Pump control strategy for city circulation loop........................................................ 26
6.2 Geothermal pipeline system............................................................................................. 28
6.3 Estimation of heat demand............................................................................................... 30
7. CONCLUSION....................................................................................................................... 32
ACKNOWLEDGEMENTS ........................................................................................................ 34
NOMENCLATURE.................................................................................................................... 35
Report 13 Şener 39
LIST OF FIGURES Page
1. Map of Turkey................................................................................................................................. 3
2. Basic scheme of geothermal pipeline system in Balçova................................................................ 4
3. Basic control scheme of city distribution network .......................................................................... 5
4. Heat exchanger schematic (Valdimarsson, 1995............................................................................. 9
5. Graph representation of a heat exchanger (Valdimarsson, 1995)................................................. 10
6. Presentation of Balcova city distribution system in the Pipelab screen ........................................ 12
7. Performance curve of circulation pump ........................................................................................ 15
8. Outside temperature vs. system heat load ..................................................................................... 16
9. Length from source vs. head loss diagram for the Balcova distribution system........................... 17
10. Pressure loss curve for the critical path of city distribution loop ................................................ 18
11. Length from source vs. node temperature diagram for the Balcova distribution system............ 18
12. Buildings with heating problem.................................................................................................. 19
13. Presentation of regions with heating problem in hl diagram of supply network........................ 19
14. Presentation of regions with heating problem in Tl diagram of supply network ....................... 20
15. Comparison of actual head difference and required head difference .......................................... 21
16. Enlarged view of region I ........................................................................................................... 23
17. Temperature drop of indicated branch......................................................................................... 23
18. Temperature drop after decreasing the piping size...................................................................... 23
19. Temperature drop after increasing flow ...................................................................................... 23
20. Pressure drop curve after increasing the flow capacity of the branch ......................................... 24
21. Maximum temperature drop in the system according to insulation quality................................. 24
22. Resized pipes in region III........................................................................................................... 26
23. Head loss diagram for resized pipes in region III ...................................................................... 26
24. Pressure loss diagram of distribution system............................................................................... 26
25. Pressure difference between heat exchanger inlet and outlet ...................................................... 27
26. Power consumption of circulation pumps for different combinations ........................................ 27
27. Heat exchanger pressure drop as a function of pump head ......................................................... 28
28. Geothermal pipeline system....................................................................................................... 29
29. Temperature distribution along geothermal pipeline................................................................... 29
30. Pressure distribution along geothermal pipeline.......................................................................... 30
31. Original programme output for time series analysis.................................................................... 31
32. Comparison of real heat load with estimations (20012002 session) ......................................... 31
33. Comparison of real heat load with estimations (20002001 session) .......................................... 32
34. Variation of maximum head loss by roughness........................................................................... 33
35. Suggested operation strategy for the system............................................................................... 34
LIST OF TABLES
1. Production wells in Balcova (2001 November)............................................................................ 3
2. Reinjection wells in Balcova (2001 November) .......................................................................... 4
3. Heat exchanger stations directly connected to geothermal pipeline system (2001 November) .... 4
4. Well operation policy for different heat loads............................................................................. 29