Applied Thermal Engineering 107 (2016) 175183

Contents lists available at ScienceDirect

Applied Thermal Engineering

journal homepage: www.elsevier.com/locate/apthermeng

Research Paper

Operational optimization in simple tri-generation systems

_
Ali Nadi nal a,, Ibrahim
Ersz b, Glgn Kayakutlu c
_
The Turkish Air Force Academy, Istanbul,
Turkey
_
_
Energy Institute, Istanbul
Technical University, Istanbul
34357, Turkey
c
_
_
Industrial Engineering Department, Istanbul
Technical University, Istanbul
34357, Turkey
a

h i g h l i g h t s

A linear optimization model for a simple tri-generation system is proposed.

The model includes operational and maintenance costs as well.

Four different PGU alternatives are studied.

Actual consumption and price data are used in the case studies.

Operating a hybrid tri-generation system reduces the annual total operational costs.

a r t i c l e

i n f o

Article history:
Received 7 February 2016
Revised 30 May 2016
Accepted 9 June 2016
Available online 11 June 2016
Keywords:
Distributed generation
Linear programming
Operational optimization
Tri-generation

a b s t r a c t
Combined energy production systems are known to be more profitable than the traditional separate
systems in which electricity and heat are generated or purchased separately. They are evolved to
trigeneration systems to combine electricity, heat and cooling. In order to get the intended advantages of
tri-generation systems, they should be operated in an optimal way. This paper proposes a simple linear
programming model to minimize the total annual variable operation and maintenance costs of a generic
tri-generation system. The optimization procedure is held by comparing four different prime mover
alternatives for three different load conditions of an industrial facility. Results show that, tri-generation
is more cost effective than the separate production for all studied scenarios, and the proposed model helps
to determine the right operational strategy, and the right system design for any given demand profile.
2016 Elsevier Ltd. All rights reserved.

1. Introduction
As the economic wealth and life standards get higher, we are
seeing an increased demand for different forms of energy. It is
reported that the worlds total final consumption reached 8979
MTOE in 2012, while it was 4672 MTOE in 1973, with the increase
in CO2 emissions from 15,633 million tons to 31,734 million tons
[1]. There is a correlation between energy consumption and GHG
emissions, since the world energy supply sticks to fossil fuels. This
rapid change in energy consumption and environmental concerns
lead the researchers and practitioners to develop more efficient
and environmental friendly conversion technologies.
Conventional centralized power generation approaches are
characterized by high rates of energy losses due to waste heat
and distribution inefficiencies [2,3]. In this regard, scientific
literature is recently addressing the local production of energy

Corresponding author.
E-mail address: anunal@hho.edu.tr (A.N. nal).
http://dx.doi.org/10.1016/j.applthermaleng.2016.06.059
1359-4311/ 2016 Elsevier Ltd. All rights reserved.

[4]. Using waste heat from on-site and near-site power

generation systems to fulfil heating and cooling needs has been
discussed for a viable solution to improve overall system efficiency
and reduce the negative environmental impacts [2,5]. Also, multigenerational energy systems that are integrated with renewable
energy sources help increase both energy and exergy efficiencies,
reduce cost and environmental impact, and increase sustainability
[6].
Tri-generation (also known as combined cooling, heating and
power, CCHP) usually refers to simultaneous production of cooling,
heating and power from a single energy source [3,5].
Tri-generation plants are usually built as decentralized systems,
and they are operated close to the end user. It is a common view
that the tri-generation systems are more efficient, profitable,
reliable, and environmentally friendly systems compared with conventional generating plants [3,79]. Nonetheless, tri-generation
systems or any other energy conversion systems should be
designed and operated effectively to gain the expected advantages,
and it is clearly an optimization problem [10].

176

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

Nomenclature
babs
bmc
dabs
dboi
dmc
dpgu

gboi
ghrs
gpgu:el
gpgu:th
rng
rpe
rse
Ed
Eex
Emc
Ep
F boi
F pgu
Q boi
Qd
Q pgu
Q rec:abs
Q rec:reg

COP of ABS
COP of MC
maintenance cost per kW for ABS [USD]
maintenance cost per kW for boiler [USD]
maintenance cost per kW for MC [USD]
maintenance cost per kW for PGU and HRS [USD]
boiler efficiency
HRS efficiency
electrical efficiency of PGU
thermal efficiency of PGU
natural gas price per kW [USD]
grid electricity price (purchasing) per kW [USD]
electricity price (selling) per kW [USD]
electricity demand [kW h]
excess electricity [kW h]
electricity consumed by MC [kW h]
purchased electricity from grid line [kW h]
fuel consumed by boiler
fuel consumed by PGU
generated heat by boiler
heating demand [kW h]
generated heat by PGU
recovered heat for ABS [kW h]
recovered heat for regular heating demand [kW h]

The levels of energy system optimization are threefold [11] (i)

synthesis, (ii) design, and (iii) operation optimization. In the
synthesis level the configuration of the system (i.e., the set of
components included) is defined. In this level several multiple
criteria decision making methods can be used [1216]. Design
optimization implies the technical specifications and the properties of substances at nominal loads, while operational optimization
finds parameters related to desirable operational regimes [11].
Several exact or stochastic optimization methods based on
mathematical modeling can be used for the design and operational
optimization levels [10].
Operational strategy is also an important factor that affects the
economic and environmental performance of a distributed
generation system. The two most distinctive operating strategies
for tri-generation systems are following electrical load, and
following thermal load [2,17,18]. In the following electrical load
operating strategy, depending on the demand profile there may
be some excess heat generated by the PGU (i.e., generating heat
over the thermal demand). The excess heat can be stored for future
use or exhausted. Conversely, in the following thermal load operating strategy, the system may generate excess electricity, depending
on the electricity demand. For grid connected systems it is possible
to exchange electricity with grid line. Choosing either the following
electrical load or the following thermal load operating strategy is a
controversial issue for practitioners.
In this paper, annual (i.e., 8760 h) operational cost of a simple
tri-generation system, which is operated under the following
thermal load strategy, is analyzed by comparing four different
prime mover alternatives for three scenarios namely the worst
case, the most probable case, and the best case. Analysis is held
based on the real operational data of an industrial facility located
_
in Istanbul,
Trkiye. A linear model is used similar to Ref. [19].
The extension to that model is the inclusion of maintenance costs
into the objective function, and the yearly analysis of the system in
comparison to the different alternatives.

Q rec
Rabs
Rd
Rmc
Sabs
Sboi
Shrs
Smc
Spgu
W pgu
ABS
ATOC
CCHP
CHP
COP
GHG
GT
HRS
ICE
MC
MTOE
PGU
TOC

recovered heat [kW h]

cooling from ABS [kW h]
cooling demand [kW h]
cooling from MC [kW h]
rated capacity of ABS
rated capacity of boiler
rated capacity of HRS
rated capacity of MC
rated capacity of PGU
generated electricity by PGU [kW h]
absorption chiller
annual total operational cost
combined cooling, heating, and power
combined heat and power
coefficient of performance
green house gases
gas turbine
heat recovery system
internal combustion engine
mechanical chiller
million tons of oil equivalent
power generation unit
total operational cost

This paper contributes to the field of energy optimization not

only by the extension of a simple model but also constructing a
realistic operational model including the maintenance activities.
The decision making process for the operating strategies broaden
the view of energy operation management. This paper first reviews
the literature on operational optimization, then constructs the
model and gives solutions. Lastly the conclusion is the section
where achievements are discussed. The proposed methodology
helps decision makers choose the right system architecture, and
helps practitioners operate the chosen system in an optimal way
with the given demand profile.
2. Literature review
Several studies use linear programming to optimize
tri-generation systems [27,2026,19,2831]. Arosio et al. [20]
developed and implemented a model for automatic optimization
of the operating policy of trigenerative plants. The constitutive
equations which formalize the relationships between the plant
components and energetic and economic target functions are
expressed using linear terms only. The implemented optimization
study can be a useful instrument for designers and stakeholders of
trigenerative plants.
In a two part study, Cardona et al. [21,22] investigated the
operation and long term planning of the CHCP plant supplying the
Malpensa 2000 international airport, by means of profit-oriented
linear optimization. They stated that purely profit-oriented management could significantly reduce the annual energy saving; however
slight changes in the operational mode allowed to achieve near optimal economic results in respect of the objective for a reduction of
energy consumption and pollutant emissions.
Cho et al. [23] presented an optimization of the operation of
CCHP systems for different climate conditions based on operational
cost, primary energy consumption, and CO2 emissions using an
optimal energy dispatch algorithm. They developed a network flow

177

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

model which facilitates setting up the objective function and

constraints for the linear programming problem. They concluded
that if CCHP systems increase the cost of operation, as long as
energy savings and reduction of emissions are guaranteed, the
implementation of these systems should be considered. Hu and
Cho [24] considered the uncertainties in the energy demand and
extended the deterministic linear programming model proposed
in [23] as a stochastic model with some probability constraints.
Ilic et al. [25] evaluated the potential of integrating a
lignocellulosic ethanol plant into a district heating system. The
district heating system has been studied using a model framework
based on linear programming, which has been developed for
optimization of dynamic energy systems with time dependent
components and boundary conditions. They concluded that introducing the plant would lead to a significant reduction in the cost
of heat production. Ilic et al. [26] used the same optimization
framework for a different study, and reported an opportunity to
reduce fossil fuel consumption by between 20% and 65%.
Kong et al. [27] presented a simple linear programming model to
determine the optimal strategies that minimize the overall cost of
energy for the CCHP system. They showed that the optimal
operation of the system is dependent upon load conditions. They
also concluded that to operate the CCHP system may not be
optimal, especially when the electricity-to-gas cost ratio is very low.
Rong and Lahdelma [28] developed an algorithm
(tri-commodity simplex) that exploits the special structure of
tri-generation models efficiently. They minimized simultaneously
the production and purchase costs of three energy components,
as well as CO2 emission costs. In another study, Rong et al. [29]
modeled the tri-generation planning problem as a linear programming problem. They developed an effective method for long term
planning. Qin et al. [30] also presented a piecewise elimination
method for hourly LP models, and a graphical method to improve
the computational efficiency. Ren et al. [31] also developed a linear
programming model for the design and evaluation of a biomass
energy system. They elaborated sensitivity analyses to show how
the optimal solutions would vary due to changes of some key
parameters such as electricity and city gas tariffs, biogas price
and etc.
Lozano et al. [19] analyzed the operation of a simple
tri-generation system. A linear programming model for a single
hour provides the operational mode with the lowest variable cost,
but the maintenance costs are not included in the model. They also
used a thermoeconomic analysis. As a contribution, we extended
the model described in [19] by adding the maintenance costs for
all components into the objective function, and also we solved
the problem using a practical and widespread solver for a year
period instead of a single hour.

Electricity grid
Ep
Fpgu

E ex

W pgu

PGU

Ed

Qpgu
HRS

Qrec

Qrec.reg

Qd

Qrec.abs

ABSORPTION
CHILLER

Fboi

BOILER

Rabs

Qboi

E mc
Rd

Rmc

MECHANICAL
CHILLER
Fig. 1. Tri-generation system.

Mechanical chillers, auxiliary boilers and the connection between

the CCHP system and the electricity grid also decrease the risk of
shortage and enhance the reliability of the system. In some cases,
depending on the legal regulations excess heat or electricity, if
any, can be sold to the grid, or both the excess heat and electricity
are allowed to be discharged easily with no cost [19,34].
The system analyzed in this paper is depicted in Fig. 1, and
consists of the following units: the PGU, a heat recovery system,
a single effect absorption chiller, an auxiliary boiler fed by natural
gas, and a mechanical chiller. Four different PGU types are studied.
One is a gas turbine and other three are internal combustion
engines fed by natural gas. The PGU generates heat and electricity.
The generated electricity can be used to feed mechanical chillers
and meet regular electricity demand. The heat generated by the
PGU can be used for regular heating demand and absorption
chillers. The auxiliary boiler generates heat for regular heating
demand only. Cooling demand can be fulfilled by absorption
chillers and mechanical chillers. All system components can be
operated in full or partial load.
3.2. General mathematical model of the system

3. Materials & methods

3.1. Simple tri-generation system
A tri-generation system is actually an extension of a CHP
system, that is, the production of a threefold energy vector
requested by the user from a unique source of fuel [4,32]. In other
words,the CCHP system is the form of the CHP system coupled
with a heat driven refrigeration system (e.g., absorption chiller)
that produces cooling when needed. COP of the absorption chillers
is generally lower than that of the electric chillers. For this reason,
the hybrid cooling system with electric chillers and absorption
chillers is recommended not only to use the recovered waste heat,
but also to improve the cooling efficiency [33].
The cogenerated heat is also used for regular heating demand.
When the recovered heat is less than the heat requirement, the
remainder of the heat requirement is met by auxiliary boilers.

A linear optimization model is used for hourly operational cost

optimization of the proposed system mentioned in Section 3.1.
Decision variables, objective function, and constraints are given
in the following subsections.
3.2.1. Decision variables
Eleven type of decision variables, for operational optimization of
the ith hour, are defined as follows: W ipgu ; Eip ; Eimc ; Eiex ; Q ipgu ; Q irec ;
Q irec:abs ; Q irec:reg ; Q iboi ; Riabs ; Rimc . These decision variables are observed
in Fig. 1. The model consists of 96,360 (i.e., 8760 11) decision
variables.
3.2.2. Objective function
The objective function to be minimized (i.e., annual total operational cost) in the linear mathematical model mainly consists of fuel

178

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

costs, maintenance costs, purchased electricity cost, and sold electricity income as given in Eq. (1). The first and the second terms
of the objective function represent the cost of fuel that is used in
the PGU and the boiler, respectively. The third through the sixth
terms represent the maintenance costs for the PGU and HRS, boiler,
absorption chiller, and mechanical chiller, respectively. The seventh
term gives the cost of purchased electricity from the grid. The last
term represents the income from the sold electricity. Nevertheless,
in the case study, selling electricity price is assumed zero, since the
sale of electricity is not permitted. The excess electricity, if any, is
transferred to the grid with no cost and no gain.

ATOC

8760
X

F ipgu rng F iboi rng W ipgu dpgu

i1

Q iboi dboi Riabs dabs Rimc dmc Eip ripe  Eiex rse

8760
X

W ipgu

i1

gpgu:el

rng

Q iboi

gboi

rng W ipgu dpgu Q iboi dboi

Riabs dabs Rimc dmc Eip ripe  Eiex rse

The second equality in Eq. (1) follows, since the fuel consumed
in the PGU and the boiler is dependent on the efficiencies of the
PGU, and the boiler, respectively. Mathematically speaking:

F pgu
F boi

W pgu

gpgu:el
Q boi

gboi

Demand constraints, capacity constraints, and balance equations are also taken into account in the mathematical model, and
given in the following sub sections.

3.2.5. Balance equations

As depicted in Fig. 1, the system has several control volumes
that the energy balances should be taken into the consideration.
The first one is the PGU which produces the electricity and heat.
The second one is the HRS which recovers the waste heat from
the PGU. The third one is the separation of the recovered heat.
The fourth one is the absorption chiller, and the last one is the
mechanical chiller. Eqs. (10)(14) are energy balance equations.

W ipgu

gpgu:el

gpgu:th  Q ipgu 0

10

Q ipgu ghrs  Q irec 0

11

Q irec:abs Q irec:reg  Q irec 0

12

Q irec:abs babs  Riabs 0

13

Eimc bmc  Rimc 0

14

where i 1; 2; . . . ; 8760. Eq. (10) follows since the PGU generates

electricity and heat simultaneously. Eq. (11) states that the recovered heat depends on the efficiency of the HRS. The recovered heat
is used for both to feed absorption chiller and to fulfil the regular
heating demand, as shown in Eq. (12). Cooling capacity of the chillers depends on the COPs as shown in Eqs. (13) and (14).
In addition to the above mentioned constraints, all decision
variables should be non-negative. In summary, the linear programming model of the system consists of 96,360 decision variables,
70,080 equality constraints, and 43,800 inequality constraints
(non-negativity constraints not included).
3.3. Solution method

3.2.3. Demand constraints

Demand constraints are defined as follows:

W ipgu

Eip

Eid

Eimc

Eiex

Rimc Riabs  Rid 0

Q iboi Q irec:reg  Q id 0

where i 1; 2; . . . ; 8760. Eq. (2) represents the electricity demand

constraint. Excess electricity may be generated depending on the
electricity demand, heating demand, and legal regulations. Eqs.
(3) and (4) represent the cooling, and heating demand constraints,
respectively. Since the system is operated in the following thermal
load strategy, the generated heat should totally fulfil the heating
demand.
3.2.4. Capacity constraints
Capacity constraints are defined as follows:

W ipgu  Spgu 6 0

Q irec  Shrs 6 0

Riabs  Sabs 6 0

Rimc  Smc 6 0

Q iboi  Sboi 6 0

where i 1; 2; . . . ; 8760. As shown in Eqs. (5)(9), capacity constraints consist of the capacities of the PGU, HRS, absorption chiller,
mechanical chiller, and boiler, respectively.

3.3.1. Linear programming and simplex method

Linear programming (also known as linear optimization) is the
process of maximizing or minimizing a linear objective function,
P
z ni1 ci xi , under linear equality and inequality constraints
[35,36]. These constraints form a convex polyhedron. As a wellestablished optimization model, linear programming has a wide
variety of real world applications from manufacturing to diet planning. A linear program (LP) can be formulated in canonical form:

Maximize z cx

15

Subject to Ax 6 b

16

xP0

17

where x is a column vector of n decision variables, c is a row vector of

n coefficients, b is a column vector with m elements (m represents
the number of constraints), and A is a m  n matrix.
The Simplex Algorithm is the most widely used, remarkably
simple, and an efficient method for solving LPs. Thousands of LPs
have been solved by practitioners or researchers using the Simplex
Algorithm since it was first introduced in 1947 by Dantzig [37].
Although its worst case complexity is exponential in the number
of variables [36,38,39], it converges in polynomial time for many
problem instances [36]. At each iteration, the method tries to
optimize the objective function finding feasible solutions.
Geometrically speaking, constraint functions defines the borders
of a convex polytope (simplex), which is actually an intersection
of half-spaces in n-dimensional Euclidean space. The sides of an
hyperplane is formed by half-spaces [36]. Starting from a vertex
of the simplex, at each iteration the algorithm moves along the
other vertices adjacently. Convexity ensures the optimal solution
to lay on the edges or borders of the feasible region. If the objective
function is not improved for the next iteration, it means that an

179

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

optimal solution is achieved. For the theoretical details of the

Simplex algorithm, we refer the interested reader to [40].

1,000

heating, and cooling especially for production process. In other

words, energy demand is highly dependent on the production
quantity and schedule. Thermal energy is mainly used for drying
process, while cooling is used for cold storage. Refrigeration is a
must for almost all plants in the food sector.
We have studied three different scenarios, namely the best, the
worst, and the most probable case scenarios, with respect to the
production plan of the plant. In the worst case scenario, the plant
works 5 days a week with 60% production capacity. In the best case
scenario, it works 7 days a week with full production capacity.
Similarly, in the most probable case scenario, the plant works
6 days a week with full production capacity. Also, the plant is off
for two weeks in a year for maintenance. Monthly average demand
in three energy vectors for all the three scenarios (included
building/space cooling and heating) is shown in Fig. 2.
As shown in Fig. 2, the plant does not have severe fluctuations
in monthly average demand, except in August. There is a crackdown in all demand vectors in August, since the average demand
decreases. This is due to the two week-off period that takes place
in this month.
Currently, the plants energy need is fulfilled via conventional
centralized manner. In other words, electricity is purchased from
the grid for regular demand, and mechanical chillers and boilers
are used for cooling, and heating demand, respectively
(i.e., separate production). Annual total operational costs for each
scenario (i.e., the worst, the best, and the most probable cases)
are given in Table 1. For all cases in separate production, a
1200 kW boiler, and a 620 kW mechanical chiller is used.
The firms top management consider investing on a
tri-generation plant. To this end, there are several alternatives that
serve as PGUs. These alternatives consist of three internal
combustion engines, and a gas turbine. The nominal capacities,
and efficiencies of these alternatives are given in Table 2. The
efficiency and COP values given in this table may change with
the outer conditions such as ambient temperature. However, for
the sake of simplicity, these values are taken to be constant.
Tri-generation is planned as an additional system into the
current conventional one. This means that the facility can cover
all energy need without the tri-generation system. A 620 kW

400

0
1

Electricity

6
7
Months

Heating

Cooling

10

11

12

10

11

12

10

11

12

600

Average demand [kW]

The firms in the food industry can be considered as good field

for tri-generation implementations, since their production and
post-production processes require electricity, heating, and cooling
simultaneously. We tested the proposed linear programming
model using actual demand data of a production plant in food
_
industry, located in Istanbul,
Trkiye. The plant uses electricity,

600

200

400

200

0
1

Electricity

6
7
Months
Heating

Cooling

1,200
1,000

Average demand [kW]

3.4. Case study

Average demand [kW]

800

3.3.2. MS excel solver and visual basic for applications

The Simplex algorithm is used by many commercial solver
platforms. One of these platforms is Microsoft Excel Solver, and it
is not only used effectively to solve linear models but also
nonlinear ones. In the linear model proposed in this paper, Excel
Solver gives the hourly optimal operational loads of the components. Since, each hours demand profile and production strategy
is independent of each other (because no storage is considered),
running the solver in an 8760 h loop is applicable for the optimization of annual total operational cost. To this end, Excel Solver is
embedded in a loop using Visual Basic for Applications. VBA is a
powerful programming and coding platform bundled in MS Office
applications. Excel Solver is used, because it is easy to implement
the model using spreadsheets, and easy to reach within almost
every personal computer.

800

600
400
200
0
1

Electricity

6
7
Months

Heating

Cooling

Fig. 2. Demand profiles for three scenarios.

Table 1
Annual total operational cost of separate production for three cases.
Case

Conditions

ATOC ($)

413680.89
Around 6000 h active production with 60%
capacity, works 5 days a week, 2 weeks-off for
maintenance
938830.94
The best
Around 8400 h active production with full
capacity, works 7 days a week, 2 weeks-off for
maintenance
The most probable Around 7200 h active production, works
810079.56
(i.e., normal)
6 days a week, 2 weeks-off for maintenance
The worst

180

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

single-effect absorption chiller is chosen for the planned

tri-generation system, since the needed cooling temperature is
greater than zero. Mechanical and absorption chillers are planned
to operate simultaneously if needed. All the system components
can be operated in partial or full load.
The unit natural gas price is assumed to be unchanged throughout a year. On the other hand, there are three different tariff
through a day for electricity, as shown in Table 3. The facility can
not sell the excess electricity, if any, but feed the grid for free, so
we assume that the selling electricity price is zero. Operational
and maintenance costs are also important inputs for the model,
and given in Table 3. As shown, unit operational and maintenance
cost for the absorption chiller is greater than that of the mechanical
chiller. This follows since absorption chillers use extra
cooling water for evaporative condensers. Additionally, parasitic
electricity, which is necessary for equipments functionality, is
negligible.
Using the data on hand, the total annual optimal operational
cost of possible systems is calculated with respect to the three
scenarios. Also, the values of decision variables at optimal
solutions are recorded for each hour of the year.

Table 2
Capacities, efficiencies, and coefficient of performance values of equipments.
Equipment

Capacity (kW)

gel

gth

COP

1st PGU alternative

2nd PGU alternative
3rd PGU alternative
4th PGU alternative
HRS 1
HRS 2
HRS 3
HRS 4
Boiler
Absorption chiller
Mechanical chiller

509.69
885.55
1,012.82
1162.00
800.00
1200.00
1200.00
2400.00
1200.00
620
620

40.52
39.86
40.32
29.00

53.85
50.69
49.18
58.00
0.90
0.90
0.90
0.90
0.95

0.67
3.2

Table 3
Price list.
Items

Price ($/kW h)

rpe 22:0005:00
rpe 06:0016:00
rpe 17:0021:00
rse
rng

0.0455
0.0808
0.1296
0.0000
0.0400
0.0055
0.0076
0.0027
0.0060
0.0030

dpgu for ICE

dpgu for GT
dboi [41]
dabs
dmc [41]

4. Results and discussion

Summary of the values of decision variables, and annual total
operational cost values for all cases with all the PGU alternatives
are given in Table 4. In the first column of this table N stands

Table 4
Summary of optimization results.
Case

Decision variables in (kW)

W pgu

N509

N885

N1012

N GT

B509

B885

B1012

B GT

W509

W885

W1012

W GT

Q pgu

ATOC ($)
Q rec

Q rec:reg

Q rec:abs

Ep

Emc

Rmc

Rabs

Q boi

Min
Max
Avg

0.00
509.69
284.39

0.00
677.36
377.95

0.00
609.63
340.16

0.00
607.70
286.86

0.00
177.55
53.29

0.00
1244.63
624.30

10.98
190.63
105.21

35.12
610.02
336.68

0.00
118.25
35.49

0.00
619.50
147.64

697183.14

Min
Max
Avg

0.00
885.55
490.37

0.00
1126.15
623.61

0.00
1013.54
561.25

0.00
607.70
286.86

0.00
581.46
274.39

0.00
1244.63
372.30

0.00
190.63
59.20

0.00
610.02
189.43

0.00
387.25
182.74

0.00
619.50
147.64

667790.36

Min
Max
Avg

0.00
1012.82
540.69

0.00
1235.38
659.50

0.00
1111.84
593.55

0.00
607.70
286.86

0.00
679.77
306.69

0.00
1244.63
315.27

0.00
190.63
52.47

0.00
610.02
167.92

0.00
452.72
204.26

0.00
619.50
147.64

656798.94

Min
Max
Avg

0.00
766.67
222.63

0.00
1533.35
445.25

0.00
1380.01
400.73

0.00
607.70
286.86

0.00
930.93
113.86

0.00
1244.63
673.46

0.00
193.75
92.61

0.00
620.00
296.34

0.00
620.00
75.83

0.00
619.50
147.64

735883.74

Min
Max
Avg

0.00
509.69
327.31

0.00
677.36
434.98

0.00
609.63
391.49

0.00
607.70
334.67

0.00
177.55
56.81

0.00
1244.63
725.55

12.57
190.63
118.73

40.22
610.02
379.93

0.00
118.25
37.84

0.00
619.50
172.25

807984.64

Min
Max
Avg

0.00
885.55
567.58

0.00
1126.15
721.79

0.00
1013.54
649.61

0.00
607.70
334.67

0.00
581.46
314.94

0.00
1244.63
431.55

0.00
190.63
65.00

0.00
610.02
208.01

0.00
387.25
209.75

0.00
619.50
172.25

773604.73

Min
Max
Avg

0.00
1012.82
626.25

0.00
1235.38
763.86

0.00
1111.84
687.47

0.00
607.70
334.67

0.00
679.77
352.80

0.00
1244.63
365.01

0.00
190.63
57.12

0.00
610.02
182.80

0.00
452.72
234.96

0.00
619.50
172.25

760795.66

Min
Max
Avg

0.00
766.67
258.46

0.00
1533.35
516.91

0.00
1380.01
465.22

0.00
607.70
334.67

0.00
930.93
130.55

0.00
1244.63
779.05

0.00
193.75
103.38

0.00
620.00
330.82

0.00
620.00
86.94

0.00
619.50
172.25

852728.41

Min
Max
Avg

0.00
509.69
241.46

0.00
677.36
320.89

0.00
609.63
288.80

0.00
364.62
143.43

0.00
350.38
145.37

0.00
746.78
199.78

0.00
114.38
34.99

0.00
366.01
111.97

0.00
233.35
96.81

0.00
371.70
73.82

338060.39

Min
Max
Avg

0.00
663.53
276.20

0.00
843.81
351.24

0.00
759.43
316.12

0.00
364.62
143.43

0.00
492.60
172.69

0.00
746.78
159.35

0.00
114.38
29.30

0.00
366.01
93.77

0.00
328.07
115.01

0.00
371.70
73.82

336089.01

Min
Max
Avg

0.00
668.79
280.61

0.00
815.75
342.27

0.00
734.17
308.04

0.00
364.62
143.43

0.00
467.34
164.61

0.00
746.78
156.62

0.00
114.38
30.98

0.00
366.01
99.15

0.00
311.25
109.63

0.00
371.70
73.82

334641.58

Min
Max
Avg

0.00
460.00
113.07

0.00
920.01
226.14

0.00
828.01
203.53

0.00
364.62
143.43

0.00
558.56
60.09

0.00
746.78
345.91

0.00
116.25
52.74

0.00
372.00
168.76

0.00
372.00
40.02

0.00
371.70
73.82

375949.23

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

for normal case (i.e., the most probable), B stands for the best
case, and W stands for the worst case. The minimum, maximum,
and the average values of decision variables through 8760 h are
tabulated for a general view. Annual total operational costs are
given in the last column.
Confirming previous studies [19,21,22,25,26], optimization
results show that to operate a tri-generation system reduces ATOC
for all cases and alternatives with given conditions, compared with
the separate production (check Table 1). Besides, the 1012 kW gas
engine gives the minimum ATOC value for all scenarios, while the
gas turbine gives the maximum ATOC. This follows since the power
to heat ratio value of gas turbines is considerably lower than that
of internal combustion engines. Although gas turbines can gener-

Usage [kW]

1,000

800

600

400

200

0
1

25

49

73

Wpgu

Hours
Qrec.reg

97

121

145

Qrec.abs

1,200

181

ate thermal energy in higher values, the electrical output is not

much. Consequently, since the electrical output from the PGU does
not meet the electricity demand totally, it is unavoidable to
purchase electricity from the grid line. As a result, operational cost
increases. Additionally, maintenance costs for gas turbines are
greater than that of internal combustion engines.
To see how the optimally operated system works within the
given conditions, we choose four representative weeks (i.e., winter,
spring, summer, and autumn) from the most probable scenario,
and give the graphical representation of the weekly operation of
the system. Results are depicted for 1012 kW PGU in hourly
manner (i.e, 168 h) in Fig. 3.
As mentioned earlier in Table 3, the facility purchases
electricity with three different price tariffs throughout a day.
As shown in Fig. 3, the PGU doesnt operate every hour. The
key factor is the relationship between the grid electricity price
and the natural gas price while deciding on whether to operate
the PGU or not. When the unit electricity price is less than or
equal to the unit fuel price, the PGU does not work. This is an
expected behavior for cost minimization, and confirms a
previous study [27].
When the PGU operates, either in partial or full load, it generates heat and work simultaneously. Since the facility needs heating
and cooling simultaneously throughout the year, recovered heat is
used almost levelled by cooling and heating as shown in Fig. 3.
Anyway, the quantity of the recovered heat for heating is slightly
more than cooling for winter, and vice versa for summer. In
transition seasons, the quantity of the recovered heat for heating
and cooling is almost equal.

Usage [kW]

1,000
800

5. Conclusion

600
400
200
0
1

25

49

73

Wpgu

Hours
Qrec.reg

97

121

145

Qrec.abs

1,200

Usage [kW]

1,000
800

600
400
200
0
1

25

49

73

Wpgu

Hours
Qrec.reg

97

121

145

121

145

Qrec.abs

1,200

Usage [kW]

1,000
800

600
400
200
0
1

25

49

73

Wpgu

Hours
Qrec.reg

97

Qrec.abs

Fig. 3. Hourly analysis for representative weeks (1012 kW ICE, the most probable
case scenario).

In this study, we proposed a linear optimization model for the

operational cost minimization of a simple tri-generation system.
We tested the proposed model for a manufacturing plant which
_
operates in the food sector, and is located in Istanbul,
Trkiye.
We used actual demand data, current prices, and current tariff
regulations. Using the data at hand, we studied four different
PGU alternatives within three demand scenarios.
Results show that operating a hybrid tri-generation system
reduces the annual total operational costs for all cases and alternatives, compared with the separate production in given conditions.
Among the alternatives, 1012 kW gas engine yields the minimum
annual total operational cost. Although 1012 kW capacity system
gives the minimum operational cost, this information is not
enough for decision making. Some other factors, such as initial
investment cost, should also be taken into consideration to choose
the proper system size. Results of a life cycle cost analysis would be
helpful for such a decision making process.
For an investment, especially for energy conversion systems,
the capacity usage ratio is an important factor to increase the
return on investment. If an installed power plant is not operated
with acceptable capacity use throughout a year, expected benefits
may not be gained. In this regard, the relationship between the
natural gas price and unit electricity price comes into prominence
when to decide whether to operate the system. In our case, for the
periods when the unit electricity price is less than or equal to the
unit natural gas price, the tri-generation system is not used.
Another important problem is allocating the recovered heat.
When the PGU operates, the recovered heat must be allocated
effectively, between the absorption chiller and regular heat
demand. The proposed model can be considered as a helpful
instrument for such allocation problems.
In summary, the proposed model and the solution method can
be used to operate an installed power plant optimally, and also it

182

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

gives the initial support for the investment decision. It is observed

that the decision achieved by operating the model can be enriched
by coupling with the life cycle costing. This study will extend the
trigeneration studies focused on the annual total operational cost
optimization by the execution of the simplest model with the real
life data and different scenarios showing the annual total
operational cost minimization by the choice of technology.

References
[1] International Energy Agency, Key World Energy Statistics 2014, 2015. <http://
www.iea.org/publications/freepublications/publication/key-world-energystatistics-2014.html> (accessed online 21.05.2015).
[2] M. Jradi, S. Riffat, Tri-generation systems: energy policies, prime movers,
cooling technologies, configurations and operation strategies, Renew. Sustain.
Energy Rev. 32 (2014) 396415, http://dx.doi.org/10.1016/j.rser.2014.01.039.
<http://www.sciencedirect.com/science/article/pii/S1364032114000501>.
[3] F.A. Al-Sulaiman, F. Hamdullahpur, I. Dincer, Trigeneration: a comprehensive
review based on prime movers, Int. J. Energy Res. 35 (3) (2011) 233258,
http://dx.doi.org/10.1002/er.1687. <http://search.ebscohost.com/login.aspx?
direct=true&db=a9h&AN=58094407&site=ehost-live>.
[4] G. Chicco, P. Mancarella, Distributed multi-generation: a comprehensive view,
Renew. Sustain. Energy Rev. 13 (3) (2009) 535551, http://dx.doi.org/10.1016/
j.rser.2007.11.014.
<http://www.sciencedirect.com/science/article/pii/
S1364032107001578>.
[5] H. Cho, A.D. Smith, P. Mago, Combined cooling, heating and power: a review of
performance improvement and optimization, Appl. Energy 136 (2014) 168
185,
http://dx.doi.org/10.1016/j.apenergy.2014.08.107.
<http://
www.sciencedirect.com/science/article/pii/S0306261914009301>.
[6] I. Dincer, C. Zamfirescu, Renewable-energy-based multigeneration systems,
Int. J. Energy Res. 36 (15) (2012) 14031415, http://dx.doi.org/10.1002/
er.2882. <http://search.ebscohost.com/login.aspx?direct=true&db=enr&AN=
83584192&site=ehost-live>.
[7] M. Liu, Y. Shi, F. Fang, Combined cooling, heating and power systems: a survey,
Renew. Sustain. Energy Rev. 35 (2014) 122, http://dx.doi.org/10.1016/j.
rser.2014.03.054.
<http://www.sciencedirect.com/science/article/pii/
S1364032114002263>.
[8] A.A. Knizley, P.J. Mago, A.D. Smith, Evaluation of the performance of combined
cooling, heating, and power systems with dual power generation units, Energy
Policy 66 (2014) 654665, http://dx.doi.org/10.1016/j.enpol.2013.11.017.
<http://www.sciencedirect.com/science/article/pii/S0301421513011245>.
[9] H.A. Moussawi, F. Fardoun, H. Louahlia-Gualous, Review of tri-generation
technologies: design evaluation, optimization, decision-making, and selection
approach, Energy Convers. Manage. 120 (2016) 157196, http://dx.doi.org/
10.1016/j.enconman.2016.04.085.
<http://www.sciencedirect.com/science/
article/pii/S0196890416303375>.
[10] A.N. nal, S. Ercan, G. Kayakutlu, Optimisation studies on tri-generation: a
review, Int. J. Energy Res. 39 (10) (2015) 13111334, http://dx.doi.org/
10.1002/er.3342. <http://dx.doi.org/10.1002/er.3342>.
[11] M.M. Stojiljkovic, M.M. Stojiljkovic, B.D. Blagojevic, Multi-objective
combinatorial optimization of trigeneration plants based on metaheuristics,
Energies 7 (12) (2014) 85548581.
[12] W. Jiang-Jiang, Z. Chun-Fa, J. You-Yin, Multi-criteria analysis of combined
cooling, heating and power systems in different climate zones in China, Appl.
Energy
87
(4)
(2010)
12471259,
http://dx.doi.org/10.1016/j.
apenergy.2009.06.027.
<http://www.sciencedirect.com/science/article/pii/
S0306261909002578>.
[13] P.-L. Chang, C.-W. Hsu, C.-Y. Lin, Assessment of hydrogen fuel cell applications
using fuzzy multiple-criteria decision making method, Appl. Energy 100
(2012) 9399, http://dx.doi.org/10.1016/j.apenergy.2012.03.051.
<http://
www.sciencedirect.com/science/article/pii/S0306261912002711>.
[14] Y.-Y. Jing, J.-H. Zhao, J.-J. Wang, Using the multi criteria analysis method
PROMENTHEE II to select the optimal CCHP system: a case study, ASME 2009
3rd International Conference on Energy Sustainability, San Francisco,
California, USA, vol. 2, 2009, pp. 3541.
[15] M. Ebrahimi, A. Keshavarz, Prime mover selection for a residential micro-CCHP
by using two multi-criteria decision-making methods, Energy Build. 55 (2012)
322331,
http://dx.doi.org/10.1016/j.enbuild.2012.09.001.
<http://
www.sciencedirect.com/science/article/pii/S0378778812004495>.
[16] A. Nieto-Morote, F. Ruz-Vlila, F. Canovas-Rodrigez, Selection of a trigeneration
system using a fuzzy AHP multi-criteria decision-making approach, Int. J.
Energy Res. 35 (9) (2011) 781794, http://dx.doi.org/10.1002/er.1739.
<http://dx.doi.org/10.1002/er.1739>.
[17] J.-J. Wang, H. Bai, Y.-Y. Jing, J.-L. Zhang, Economic analysis and optimization
design of a solar combined cooling heating and power system in different
operation strategies, in: 7th IEEE Conference on Industrial Electronics and
Applications (ICIEA), 2012, 2012, pp. 108112, http://dx.doi.org/10.1109/
ICIEA.2012.6360706.
[18] S. Jayasekara, S. Halgamuge, A review on optimization strategies of combined
cooling heating and power generation, in: IEEE 6th International Conference
on Information and Automation for Sustainability (ICIAfS), 2012, 2012, pp.
302307, http://dx.doi.org/10.1109/ICIAFS.2012.6419921.

[19] M.A. Lozano, M. Carvalho, L.M. Serra, Operational strategy and marginal costs
in simple trigeneration systems, Energy 34 (11) (2009) 20012008, http://dx.
doi.org/10.1016/j.energy.2009.08.015. <http://www.sciencedirect.com/science/
article/pii/S0360544209003545>.
[20] S. Arosio, M. Guilizzoni, F. Pravettoni, A model for micro-trigeneration systems
based on linear optimization and the Italian tariff policy, Appl. Therm. Eng. 31
(1415) (2011) 22922300, http://dx.doi.org/10.1016/j.applthermaleng.
2011.03.027. <http://www.sciencedirect.com/science/article/pii/S135943111
1001591>.
[21] E. Cardona, P. Sannino, A. Piacentino, F. Cardona, Energy saving in airports by
trigeneration. Part II: Short and long term planning for the Malpensa 2000
CHCP plant, Appl. Therm. Eng. 26 (1415) (2006) 14371447, http://dx.doi.
org/10.1016/j.applthermaleng.2006.01.020. <http://www.sciencedirect.com/
science/article/pii/S1359431106000305>.
[22] E. Cardona, A. Piacentino, F. Cardona, Energy saving in airports by
trigeneration. Part I: Assessing economic and technical potential, Appl.
Therm. Eng. 26 (1415) (2006) 14271436, http://dx.doi.org/10.1016/j.
applthermaleng.2006.01.019.
<http://www.sciencedirect.com/science/article/
pii/S1359431106000299>.
[23] H. Cho, P.J. Mago, R. Luck, L.M. Chamra, Evaluation of CCHP systems
performance based on operational cost, primary energy consumption, and
carbon dioxide emission by utilizing an optimal operation scheme, Appl.
Energy 86 (12) (2009) 25402549, http://dx.doi.org/10.1016/j.apenergy.
2009.04.012. <http://www.sciencedirect.com/science/article/pii/S030626190
9001421>.
[24] M. Hu, H. Cho, A probability constrained multi-objective optimization model
for CCHP system operation decision support, Appl. Energy 116 (0) (2014) 230242,
http://dx.doi.org/10.1016/j.apenergy.2013.11.065. <http://www.sciencedirect.
com/science/article/pii/S0306261913009732>.
[25] D. Djuric Ilic, E. Dotzauer, L. Trygg, District heating and ethanol production
through polygeneration in Stockholm, Appl. Energy 91 (1) (2012) 214221,
http://dx.doi.org/10.1016/j.apenergy.2011.09.030. <http://www.sciencedirect.
com/science/article/pii/S0306261911006337>.
[26] D. Djuric Ilic, E. Dotzauer, L. Trygg, G. Broman, Introduction of large-scale
biofuel production in a district heating system an opportunity for reduction
of global greenhouse gas emissions, J. Clean. Prod. 64 (2014) 552561, http://
dx.doi.org/10.1016/j.jclepro.2013.08.029.
<http://www.sciencedirect.com/
science/article/pii/S0959652613005696>.
[27] X. Kong, R. Wang, X. Huang, Energy optimization model for a CCHP system
with available gas turbines, Appl. Therm. Eng. 25 (23) (2005) 377391, http://
dx.doi.org/10.1016/j.applthermaleng.2004.06.014. <http://www.sciencedirect.
com/science/article/pii/S1359431104001784>.
[28] A. Rong, R. Lahdelma, An efficient linear programming model and optimization
algorithm for trigeneration, Appl. Energy 82 (1) (2005) 4063, http://dx.doi.
org/10.1016/j.apenergy.2004.07.013. <http://www.sciencedirect.com/science/
article/pii/S030626190400162X>.
[29] A. Rong, R. Lahdelma, P.B. Luh, Lagrangian relaxation based algorithm for
trigeneration planning with storages, Eur. J. Oper. Res. 188 (1) (2008) 240257,
http://dx.doi.org/10.1016/j.ejor.2007.04.008. <http://www.sciencedirect.com/
science/article/pii/S0377221707003840>.
[30] C. Qin, J. Tang, Y. Zhang, An efficient algorithm for {CCHP} system sizing and an
operational optimization model based on {LP}, J. Nat. Gas Sci. Eng. 25 (2015)
189196, http://dx.doi.org/10.1016/j.jngse.2015.05.001.
<http://www.
sciencedirect.com/science/article/pii/S1875510015001857>.
[31] H. Ren, W. Zhou, K. Nakagami, W. Gao, Integrated design and evaluation of
biomass energy system taking into consideration demand side characteristics,
Energy 35 (5) (2010) 22102222, http://dx.doi.org/10.1016/j.energy.
2010.02.007. <http://www.sciencedirect.com/science/article/pii/S036054421
0000599>.
[32] G. Chicco, P. Mancarella, From cogeneration to trigeneration: profitable
alternatives in a competitive market, IEEE Trans. Energy Convers. 21 (1)
(2006) 265272, http://dx.doi.org/10.1109/TEC.2005.858089. <http://ieeexplore.
ieee.org/lpdocs/epic03/wrapper.htm?arnumber=1597346>.
[33] J. Wang, Z.J. Zhai, Y. Jing, C. Zhang, Particle swarm optimization for redundant
building cooling heating and power system, Appl. Energy 87 (12) (2010)
36683679, http://dx.doi.org/10.1016/j.apenergy.2010.06.021. <http://www.
sciencedirect.com/science/article/pii/S0306261910002485>.
[34] M. Li, H. Mu, H. Li, Analysis and assessments of combined cooling, heating and
power systems in various operation modes for a building in China, Dalian,
Energies (19961073) 6 (5) (2013) 24462467, http://dx.doi.org/10.3390/
en6052446.
<http://search.ebscohost.com/login.aspx?direct=true&db=enr&
AN=87773688&site=ehost-live>.
[35] N. Ploskas, N. Samaras, Efficient GPU-based implementations of simplex type
algorithms, Appl. Math. Comput. 250 (2015) 552570, http://dx.doi.org/
10.1016/j.amc.2014.10.096. <http://www.sciencedirect.com/science/article/
pii/S0096300314014714>.
[36] M. Vujosevic-Janicic, F. Maric, D. Tosic, Using simplex method in verifying
software safety, Yugoslav J. Oper. Res. 19 (1) (2009) 133148, http://dx.doi.
org/10.2298/YUJOR0901133V.
<http://yujor.fon.bg.ac.rs/index.php/journal/
article/view/649>.
[37] G.B. Dantzig, Origins of the Simplex Method Tech. Rep., Department of
Operations Research, Stanford University, CA, 1987.
[38] T. Kitahara, S. Mizuno, On the number of solutions generated by the dual
simplex method, Oper. Res. Lett. 40 (3) (2012) 172174, http://dx.doi.org/
10.1016/j.orl.2012.01.004.
<http://www.sciencedirect.com/science/article/
pii/S0167637712000053>.

A.N. nal et al. / Applied Thermal Engineering 107 (2016) 175183

[39] V. Klee, G.J. Minty, How good is the simplex method, in: O. Shisha (Ed.),
Inequalities-III, Academic Press, New York, 1972, pp. 159175.
[40] D.G. Luenberger, Y. Ye, Linear and Nonlinear Programming, International
Series in Operations Research & Management Science, Springer, US, 2008.
<http://www.springer.com/us/book/9781441945044>.

183

[41] S. Sanaye, H. Hajabdollahi, 4E analysis and multi-objective optimization of

CCHP using MOPSOA, Proceedings of the Institution of Mechanical Engineers,
Part E: Journal of Process Mechanical Engineering 228 (1) (2014) 4360,
http://dx.doi.org/10.1177/0954408912471001. <http://pie.sagepub.com/content/228/1/43.abstract>.