You are on page 1of 8

Electrical Power and Energy Systems 42 (2012) 728735

Contents lists available at SciVerse ScienceDirect

Electrical Power and Energy Systems


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

Multiobjective optimization using Mesh Adaptive Direct Search for power


dispatch problem of microgrid
Faisal A. Mohamed a,, Heikki N. Koivo b
a
b

Department of Electrical Engineering, Omar Al-Mukhtar University, P.O. Box 919, El-Bieda, Libya
Department of Automation and Systems Technology, Aalto University, P.O. Box 15500, 00076 Aalto, Finland

a r t i c l e

i n f o

Article history:
Received 8 August 2010
Received in revised form 4 September 2011
Accepted 16 September 2011
Available online 22 June 2012
Keywords:
Microgrid
Multiobjective optimization
Load management
Mesh Adaptive Direct Search

a b s t r a c t
This paper presents a generalized formulation for determining the optimal operating strategy and cost
optimization scheme and reduction of emissions of a MicroGrid (MG). Multiobjective (MO) optimization
is applied to the environmental economic problem of the MG. The proposed problem is formulated as a
nonlinear constrained MO optimization problem. The proposed problem takes into consideration the
operation and maintenance costs as well as the emission reduction of NOx, SO2, and CO2. The MG considered in this paper consists of a wind turbine, a micro turbine, a diesel generator, a photovoltaic array, a
fuel cell, and a battery storage. The Multiobjective Mesh Adaptive Direct Search (MOMADS) is employed
to minimize the cost function of the system while constraining it to meet the costumer demand and
safety of the system. A comparison is made with Multiobjective Sequential Quadratic Programming
(MOSQP). The results demonstrate the efciency of the proposed approach to satisfy the load and to
reduce the cost and the emissions.
2011 Elsevier Ltd. All rights reserved.

1. Introduction
The need for more exible electric systems, changing regulatory
and economic scenarios, energy savings and environmental impact
are providing impetus to the development of MGs, which are predicted to play an increasing role in future power systems [1]. One
of the important applications of the MG units is the utilization of
small-modular residential or commercial units for onsite service.
The MG units can be chosen so that they satisfy the customer load
demand at compromise cost and emissions all the time.
The management of the MG units requires an accurate environmental economic model to describe the operating problem taking
into account the output power production. Such a model is discrete
and nonlinear in nature, hence optimizations tools are needed to
extract the best compromise solution between the operating costs
and emission.
The environmental/economic problem is conicting in nature.
In reality, the environmental/economic problem is a multiobjective

Abbreviations: MG, MicroGrid; MOMADS, Multiobjective Mesh Adaptive Direct


Search; MOSQP, multiobjective sequential quadratic programming; MO, multiobjective; DG, diesel generator; FC, fuel cell; MT, microturbine; WT, wind turbine; PV,
fotovoltaic; PCC, point common coupling; SD, static switch.
Corresponding author.
E-mail addresses: mohamed.faisal@hut., elabdli@hotmail.com (F.A. Mohamed),
heikki.koivo@hut. (H.N. Koivo).
0142-0615/$ - see front matter 2011 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijepes.2011.09.006

problem with conicting objective, because minimum emissions is


conicting to minimum cost of generation.
Multiobjective optimization (MO) has a very wide range of successful applications in engineering and economics. Such applying
multiobjective optimization for power system problems [2], applications can be found in optimal control systems [3], engineering
design [4], and communication [5], in [6] a new formulation of
multi-objective reactive power and voltage control is presented
for power System basis of the fuzzy sets theory and particle swarm
optimization (PSO). In [7] a multi-objective approach based on the
BellmanZadeh algorithm and fuzzy logic is used to determine
appropriate DG sites. The MO optimization is applied to nd the
optimal solution which is a compromise between multiple and
contradicting objectives. In MO optimization we are mostly interested in the Pareto optimal set which contains all non-inferior
solutions. The decision maker can then select the most preferred
solution out of the Pareto optimal set. The weighted sum method
to handle MO optimization is applied in this paper. Furthermore,
the weighted sum is a simple and straightforward method to handle MO optimization problems.
Solving the environmental/economic problem in the power
generation has received considerable attention. In [8], the problem
has been investigated as a single objective problem by treating the
emissions as constraints in which the total annual emissions were
limited by a certain amount. Moreover, in [1] the optimization is
aimed at reducing the fuel consumption rate of the system while
constraining it to full the local energy demand (both electrical

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

and thermal) and provide a certain minimum reserve power. The


problem has been treated as a single objective problem by neither
considering the emission nor the operation and maintenance costs.
In [9], the problem is also solved as s single objective optimization.
The formulation, however, has a severe difculty in getting the
trade-off relation between the operating costs and emission level.
The problems have been effectively solved by goal programming
[10], classical technique [11], and fuzzy satisfaction-maximizing
approach [12]. The computing speed of these solutions, however,
limits their online applications.
The second objective of this paper deals with solving the optimization problem to explore the benets of having optimal management of the MG. The exploration is based on the
minimization of running costs and reducing the emissions. It is
then extended to cover a load demand scenario in the MG. It will
be shown that by developing a good system model, we can use
an optimization technique to solve the optimization problem accurately and efciently.

inequality constraint vector, G(P) = (g1(P1), g2(P2), . . . , gp(PN))T is less


or equal to zero.
3. Proposed objective function
The major concern in the design of an electrical system that utilizes MG sources is the accurate selection of output power that can
economically satisfy the load demand, while minimizing the emissions. Therefore the system components are required to:
1. Minimize the operation cost ($/h).
2. Minimize the emissions (kg/h).
3. Ensure that the load is served according to the constraints.
3.1. Operating cost
As Shown in Fig. 1, the main utility balances the difference between the load demand and the generated output power from
microsources. Then, the cost function takes the form [14]:

2. Optimization model

CFP
The power optimization model is highlighting the following
points. The output is the optimal conguration of a MG that takes
into account technical performance of supply options, locally available energy resources, demand characteristics, and environmental
levels. Small-scale power generating technologies under consideration include PV, WT, DG, and FC. To run the model, the following
items have to be dened:
 The power demand by the load.
 Locally available energy information: This includes solar irradiation data (W/m2), temperature (C), wind speed (m/s), as well
as cost of fuels ($/l) for the DG and natural gas price for supplying the FC and MT ($/kWh).
 Daily purchased and sold power tariffs in ($/kWh).
 Start up costs in ($/h).
 Technical and economic performance of supply options: These
characteristics include, for example, rated power for PV, power
curve for WT, fuel consumption characteristics DG and FC.
 Operating and maintenance costs and the total emission: Operating and maintenance costs must be given ($/h) for all emissions; the total emission must be given in kg/h for DG, FC, and
MT.
Multiobjective optimization is a method to nd the best solution between different, usually conicting objectives. In the MO
optimization problem we have a vector of objective functions. Each
objective function is a function of the decision (variable) vector
[13]. Mathematically the environmental/economic problem is formulated as follows:
Find the output generator power vector P = [P1, P2, . . . , PN]0 that
minimizes the function:

FP fCFP; EPg

Subject to

729

N
X
C i  F i Pi OMi Pi STC i DCPEi  IPSEi

i1

where
F(Pi)
Ci
Fi
OMi
STCi
Pi
DCPEi
IPSEi

The operating cost of the generating unit i in $/h


Fuel costs of the generating unit i in $/l for the DG, and
in $/kWh for FC and MT
Fuel consumption rate of a generator unit i
Operation and maintenance cost of a generating unit i
in $/h
Start up cost of the generating unit i in $/h
Decision variables, representing the real power output
from generating unit i in kW
Purchased electricity of unit i if the load demand
exceeds the generated power in $/h
Income for sold electricity of unit i if the output
generated power exceeds the load demand in $/h

To model the purchased and sold power, two different conditions are considered. The following equations dene these
conditions:



X
DCPEi C p  max PL 
Pi ; 0
X

IPSEi C s  max
Pi  PL ; 0

where Cp and Cs are the tariffs of the purchased and sold power
respectively in ($/kWh).
3.1.1. Objective constraints
Power balance constraints: To meet the active power balance, an
equality constraint is imposed
N
X
Pi  PL P PV P WT Pbatt 0

i1

hk Pi 0

k 1; . . . ; q

g j P i 6 0

j 1; . . . ; p

6 Pi 6 Pmax
;
Pmin
i
i

8i 1; . . . ; N

where the number of the objective functions P2, and F(P) : Rn ? R.


The vector of objective functions is denoted by F(P) = (F(P1),
F(P2), . . . , F(Pk))T. The decision variable vector P = (P1, P2, . . . , PN)T
consists of all the design variables and may be bounded. The collection of the equality constraints, H(P) = (h1(P1), h2(P2), . . . , hq(PN))T, is
a q-dimensional equality constraint vector, and similarly the

where
PL
PPV
PWT
Pbatt

The
The
The
The

total power demanded in kW


output power of the photovoltaic cell in kW
output power of the wind turbine in kW
output power of the battery storage kW

Generation capacity constraints: For stable operation, real power


output of each generator is restricted by lower and upper limits as
follows:

730

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

MicroGrid Central
Controller

PCC

SD

Feeder 3

Feeder 2

Feeder 1

Load

LC

LC
LC

Load

Load

Wind
Turbine

Load

Diesel
Engine

Load

LC

LC
Fuel
Cell

Load

Load

Load

PV
array

Heat
load
Charger
Controller

LC

Load

Micro
Turbine

Battery

Fig. 1. MicroGrid architecture.

Pmin
6 Pi 6 Pmax
;
i
i

i 1; . . . ; N

where
Pmin
i
Pmax
i

Minimum operating power of unit i

EP

Each generating unit has a minimum up/down time limits


(MUT/MDT) time limits, once the generating unit is switched on,
it has to operate continuously for a certain minimum time before
switching it off again. On the other hand, a certain stop time has
to be terminated before starting the unit. The violation of such constraints can cause shortness in the life time of the unit. These constraints are formulated as continuous running/stop time constraint
as follows [14]:

on
T off
t1;i =T t1;i is the unit off/on time, while ut1,i denotes the unit off/on
[0, 1] status.
The number of starts and stops (estartstop) should not exceed a
certain number (Nmax).

estartstop 6 Nmax

N
X



102 ai bi Pi ci P2i fi expki P i

11

i1

Maximum operating power of unit i


T on
t1;i  MUT i ut1;i  ut;i P 0


T off
t1;i  MDT i ut1;i  ut;i P 0

mal units can be modeled separately. The total kg/h emission of


these pollutants can be expressed as [15]:

10

3.2. Emission level


The atmospheric pollutants such as sulfur oxides SO2, carbon
oxides CO2, and nitrogen oxides NOx caused by fossil-fueled ther-

where ai, bi,ci, fi, and ki i = 1, . . . , N, are nonnegative coefcients of


the ith generator emission characteristics.
For the emission model introduced in [15,16], we propose to
evaluate the parameters a, b, c, f, and k using the data available
in [17]. Thus, the emissions per day for the DG, FC, and MT is estimated, and the characteristics of each generator will be detached
accordingly.
4. System model
The MG architecture studied is shown in Fig. 1. It consists of a
group of radial feeders, which could be part of a distribution system. There is a single point of connection to the utility called Point
of Common Coupling (PCC). The feeders 1 and 2 have sensitive
loads which should be supplied during the events. The feeders also
have the microsources consisting of a photovoltaic (PV), a wind
turbine (WT), a fuel cell (FC), a microturbine (MT), a diesel generator, and a battery storage. The third feeder has only traditional
loads. The static switch (SD) is used to island the feeders 1 and 2
from the utility when events happened. The fuel input is needed
only for the DG, FC, and MT as the fuel for the WT and PV comes
from nature. To serve the load demand and charge the battery,
electrical power can be produced either directly by PV, WT, DG,
MT, or FC. Each component of the MG system is modeled separately based on its characteristics and constraints. The characteristics of some equipment like wind turbines and diesel generators

731

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

are available from the manufacturer. A charger controller is required to limit the depth of discharge of the battery, to limit the
charging current supplied to the battery, and to prevent overcharging, while making use of the power from the other microsources
when it is available.

5. The Mesh Adaptive Direct Search (MADS) algorithm


The MADS class of algorithms, introduced in [18], is designed
for nonsmooth optimization problems. The convergence analysis
of MADS ensured necessary optimality conditions of the rst [18]
and second [19] orders under certain assumptions.
A general optimization problem may be stated as,

minP2X CFP

12

where CF : X ! R [ f1g; X fP 2 Rn : l 6 AP 6 ug; A 2 Rmn is a


real matrix, l; u 2 R [ f1gn , and l 6 u. Each iteration k of the
MADS algorithm is characterized by two steps. First, an optional
search step over the space of variables is carried out, as long as it
is a nite process and all trial points lie on a mesh. If no better point
is found or no global search is used, the algorithm goes to a mandatory local exploration step (mandatory because it ensures convergence). Second is the poll step, at most 2n trial mesh points near
the incumbent solution are selected (the poll set) and evaluated.
If no better neighbor is found, the mesh is rened. If an improved
mesh point Pk+1 2 X is found, the mesh is kept the same or coarsened, and then Pk+1 is the next incumbent. The exploration directions vary at each iteration, and become dense with probability 1.
This is the main difference between the pattern search and MADS
algorithms. General constraints are handled with a barrier approach, which redenes the objective as follows:

CF X

CFP if P 2 X
1

otherwise

13

Then, MADS is applied to the unconstrained barrier problem

minP CF X P

14

The feasible region X can be nonlinear, non-convex, non-differentiable, or disjoint. There are no hypotheses made on the domain, except that the initial point must be feasible. The convergence
results depend on the local smoothness of CF (and not CFX, which
is obviously discontinuous on the boundary of X). They also depend
on the tangent cone at the limit point produced by the algorithm.

The proposed optimization method is compared with the results obtained in [2022]. It also incorporates an explicit cost minimization criterion applied to the MG architecture as well as
minimizing the emission. The formulation in this work seeks the
most environmental/economical generation to satisfy the load demand and the constraints. The problem is decomposed into two
stages, starting with building the system model, which is an important stage to understand the problem. The next stage is the application of the algorithm developed. The algorithm consists of
determining at each iteration the optimal use of the resources
available, such as wind speed, temperature, and irradiation as they
are the inputs to the model. If the produced power from the wind
turbine and the photovoltaic cell is less than the load demand, then
the algorithm goes to the next stage which is the use of the other
alternative sources according to the load and the objective function
of each one.

6. Implementation of the algorithm


When designing MGs, several goals could be set, including
reduction in emissions and generation cost. To achieve this, it is
important to highlight all factors inuencing the main goal. The
following items summarize the key characteristics of the implemented strategy:
 Power output of WT is calculated according to the relation
between the wind speed and the output power.
 Power output of PV is calculated according to the effect of the
temperature and the solar radiation that are different from
the standard test condition.
 We assume that the WT and PV deliver free cost power in terms
of running as well being emission free. Furthermore, their output power is treated as a negative load, determine the different
between the actual load and WT and PV output power. If the
output from PV and WT is greater than the load, the excess
power is directed to charge the battery.
 The power from the battery is needed whenever the PV and WT
are insufcient to serve the load. Meanwhile the charge and discharge of the battery is monitored.
 The net load is calculated if the output from PV and WT is smaller than the total load demand and the battery is discharged.
 Choose serving the load by other sources (FC or MT or DG)
according to the objective functions.

5.1. Description of MADS algorithm

7. Results and discussion

A general and exible algorithmic framework for MADS was


proposed in [18]. This is then specialized to a specic algorithmic
implementation.The main steps of the algorithm are summarized
as follows:

At rst, the optimization model is applied to the load shown in


Fig. 2 which is rescaled from [23,24]. The load demand varies from

14

full load
load remined from the PV,WT

12

Load (KW)

 Initialization
The user denes the starting point and the initial mesh size.
The algorithm initializes other parameters for subsequent steps.
 Quest for an improved mesh point
Global search (optional): evaluation of CF over a nite subset of
points dened by the mesh.
Local poll (mandatory): denition of a poll set and evaluation of
CF over points in that set.
 Parameter update
Parameters are updated.
 Termination
If some stopping criterion is reached, stop; if not, go back to step
2.

10
8
6
4
2

10

12

14

Time (hour)
Fig. 2. Hourly load.

16

18

20

22

24

732

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

2
Emission

1.2

1.5

Cost

0.8

0.5

0.6

Cost [$/h]

1.2
Emission

1.15

Cost

1.1

1.2
1

10

15

1.05
25

20

10

15

20

4
3.5

Fig. 3. Convergence of cost and emission objective functions using MOSQP and
MOMADS.

Emission [kg/h]

MOSQP
MOMADS

2.5
2
1.5

Cost [$]

Fig. 5. Hourly emission using the MOSQP and MOMADS.

Iteration

Time [hour]

Emission [kg/h]

1.25

1.4

Convergence of cost and emission objective functions MOMADS


1.6

Iteration

1.8

MOSQP
MOMADS

Emission [kg]

Cost [$/h]

1.4

Emission [kg/h]

Convergence of cost and emission objective functions MOSQP

1
0.5

2
3
Operating Cost [$/h]

Fig. 6. Trade-off in cost and emission using MOSQP and for P = 0.12 and S = 0.

10

15

20

Time [hour]
Fig. 4. Hourly operating cost using the MOSQP and MOMADS.

4 kW to 14 kW. The available power from the PV and the WT are


used rst.
The best results of the cost and emission functions, when optimized individually, are given in Table 1. Convergence of the operation cost and emission objectives for both approaches, when the
purchased tariff is 0.12 $/kWh and the sold tariff 0.07 $/kWh, is
shown in Fig. 3.
Figs. 4 and 5 illustrate the hourly operating costs and emissions
for the two proposed techniques. However, the costs and emissions
are high when the generators are on and the load is high.
Then the problem is solved as a MO optimization problem
where both operation cost and emissions are optimized simultaneously. Considering the diversity of the Pareto optimal set over
the trade-off surface, the trade-off relation can be obtained by minimizing the function [16]:

CP hCFP 1  hEP

15

subject to power balance, and upper and lower limits on the generation. Here h 2 [0, 1]. Value h = 1.0 implies minimum operating cost
and full emissions. We obtain the optimum solution of the operating cost objective. The importance of the emission objective increases when h decreases. Then the optimum solution will move
toward the optimal emission objective value h = 0.0, which implies
minimum emission with no attention being paid to operating costs.
The optimum solution of the emission objective is now obtained.
Function C is minimized for successive values of h to cover the entire range from 0 to 1, the two objectives are given the same
weights. For non-dominated solution points, an improvement in
one objective requires degradation of the other objective. The proposed model is highly nonlinear. Since each generator has different
behavior that inuences the operating cost, the solutions are diverse and acceptably distributed over the trade-off curve.
Figs. 69 show the relationship (trade-off curves) of the operating cost and emission objectives of the non-dominated solutions
obtained by MOSQP and MOMADS approaches for different purchased and sold tariffs. Considering the denition of the multiobjective problems, a non-dominated solution becomes a feasible
solution. Then at least one of the objective values is better than

Table 1
The objective functions when optimized individually.
PL (kW/day)

Case 1
Case 2

171.4009
171.4009

TE (kg/day)

TC ($/day)

OG (kW/day)

SP (kW/day)

PP (kW/day)

MOSQP

MOMADS

MOSQP

MOMADS

MOSQP

MOMADS

MOSQP

MOMADS

MOSQP

MOMADS

149.9198
12.3885

68.6616
13.0826

76.1973
124.6914

61.3615
113.7293

241.9197
36.0860

144.3805
42.5693

70.5188
00.0000

17.8455
00.0000

14.8635
135.3149

27.0204
128.8316

Where TE, total emissions; TC, total operating cost; OG, total optimal generated power; SP, total sold power; PP, total purchased power.

733

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

14

Load
Optimal Generation Case 1

12

Optimal Generation Case 2


10

Power [kW]

Emission [kg/h]

8
6
4

2
0

4
Operating Cost [$/h]

10

15

20

Time [hour]

Fig. 7. Trade-off in cost and emission using MOMADS and for P = 0.12 and S = 0.

Fig. 10. Effect of purchased power tariffs on the MG optimal operation using
MOMADS.

14
Load

Optimal Generation Case 1

12

Power [kW]

Emission [kg/h]

Optimal Generation Case 2


10

3
2

8
6
4

2
0

10

Operting Cost [$/h]


Fig. 8. Trade-off in cost and emission using MOSQP and for P = 0.16 and S = 0.1.

15

20

Fig. 11. Effect of purchased power tariffs on the MG optimal operation using
MOSQP.

3.5

MT
FC

DG
2.5

Power [kW]

Emission [kg/h]

10

Time [hour]

2
1.5
1

0.5
0

4
6
Operating Cost [$/h]

10

Fig. 9. Trade-off in cost and emission using MOMADS at scenario 3 and for P = 0.16
and S = 0.1

the corresponding objective of all the other feasible solutions. The


non-dominated solutions are those from which the multi-objective
decision algorithm attempts to select the best compromise solution according to the preferences of the decision makers. Consequently, the two objectives of all the non-dominated solutions

10

15

20

Time [hour]
Fig. 12. Power generation distribution using MOSQP.

are located along the left and lower boundaries of the feasible domain as minimization is desired. The operating costs of the nondominated solutions thus appear to be inversely proportional to
their emissions, as illustrated in Figs. 69.
Table 2 and Figs. 10 and 11 show the effect of changing the purchased and sold tariffs on the optimal setting of the MG. There are

734

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

Table 2
The effect of the purchased and sold tariffs on the optimal generation.
Load (kW/day)

Case
Case
Case
Case

1
2
3
4

171.4009
171.4009
171.4009
171.4009

Optimal generation (kW/day)

Total cost ($/day)

Total emission (kg/day)

MOSQP

MOMADS

MOSQP

MOMADS

MOSQP

MOMADS

112.5081
127.1114
117.7365
117.7365

110.7196
112.4749
111.2769
111.2769

58.5862
61.3490
60.4168
60.4168

51.0058
64.5349
56.3351
56.3351

40.3235
49.4893
44.6083
44.6083

47.3697
47.1226
47.5260
47.5260

Table 3
Cost savings and emissions reductions of the MG using multiobjective optimization Scenario 3.
Average cost and emissions

Case
Case
Case
Case

A
B
C
D

MOSQP
MOMADS

Average difference with respect to the optimal case

Cost ($/day)

Emissions (kg/day)

Cost ($/day)

Emissions (kg/day)

Cost (%)

Emissions (%)

95.3091
68.6616
113.7293
53.5643

229.4895
61.3615
13.0826
58.9397

47.0079
20.3604
65.4281
53.5643

187.6880
19.5600
28.7189
11.7628

97.32
42.15
135.46
10.90

449.00
46.79
68.70
28.14

49.3980
48.3012

44.0851
41.8015

1.0968
00.0000

2.2836
00.0000

2.27
0

5.46
0

all together four cases. In case 1, the effect of the changing the purchased tariffs is studied. The sold power is rst 0.04 $/kWh and the
purchased tariff is 0.1 $/kWh. In Case 2 the value of the purchased
tariff is increased to 0.16 $/kWh and the sold is the same as in Case
1. Considering the purchased tariff values, it was noticed that when
the tariff is low, it was preferable to buy as much power from the
main grid as possible. However, when the tariffs were higher, it
was more economic to generate the required power from the MG
according to MOSQP and MOMADS approaches.
In Cases 3 and 4, the purchase power tariff is kept constant at
0.12 $/kWh, while the sold tariff was 0.0 $/kWh in Case 3 and
0.04 $/kWh in Case 4. It is noticeable that the MG generates more
power when the sold tariffs are increased and MOMADS optimization algorithm is applied. In the MOSQP, the small variation of the
sold power tariff has no signicant effect This leads to conclude
that the MOMADS is more effective when the variation of the sold
tariffs is small. The higher values of the sold power tariff make it
possible to produce more power to meet the load demand in
MOMADS.
Table 3 illustrates the cost savings and emission reductions of
the MG using different cases and compares them with the proposed MOMADS technique. The results obtained using the proposed technique to minimize the total cost and total emissions is

5
MT
FC

Power [kW]

DG

Minimize
P

xCFP 1  xkEP

16

where k is the scaling factor and x is the weighting factor. The scaling factor k is used to balance the two objectives. The increase of the
scaling factor favors the predominance of the total emission objective function over the total operating cost objective function. The
value of k = 3000 was found to be the best compromise between
the two objectives. In this study the weighting factor x is selected
to vary randomly x = rand[0, 1] and k is chosen to be 3000. With the
proposed MOMADS method, the total operating cost and emissions
are reduced to 48.3012 $/day and 41.8015 kg/day respectively compared to MOSQP. MOMADS is more capable of handling the multiobjective optimization problem of the MG when the problem
becomes more complex, e.g, when more constraints and purchased
and sold power are considered.
It is also noticeable from Figs. 12 and 13 that MOMADS has a
good distribution of the powers given to the micro-sources.
8. Conclusion

compared with some conventional strategies. In the rst case the


DG, FC, and MT operate at their rated power for the whole day
(Case A). In the second case costs are optimized individually (Case
B). In the third scenario emissions objective function is optimized
individually (Case C). Case A gives a higher operating cost and
higher emissions which indicates that it is not acceptable. Larger
generating power, larger costs and emissions are attained. In Case
B, the cost is somewhat reduced, while the emissions increase. In
the third case, the cost increased while the emissions decreased
and the optimal choice was to purchase more power from the main
grid. For achieving the completeness and checking the effectiveness of the proposed cost function and proposed solution, the
problem is also treated as a single objective optimization problem:
[25] (case D):

10

15

20

Time [hour]
Fig. 13. Power generation distribution using MOMADS.

Objective functions to determine the optimum operation of a


MG with respect to the load demand and environmental requirements are constructed and presented in a new form. The optimization problem includes a variety of energy sources that are likely to
be found in a microgrid: fuel cells, diesel engines, microturbines,
PV arrays, wind generators, and battery storages. Constraint functions are added to the optimization problem to reect some of the

F.A. Mohamed, H.N. Koivo / Electrical Power and Energy Systems 42 (2012) 728735

additional considerations often found in a small-scale generation


system. The results of the optimal power operating costs and emission curves for the MG ensure that the optimization works very
well and can give the optimal power to the generators after taking
the operating costs and emissions into account. The effectiveness
of the suggested approach is conrmed through the agreement between the optimized settings and the output from the algorithm.
The responses are effected by several variables such as weather
conditions, sold and purchased tariffs, and the actual power demand. The results show the capability of the proposed system
model and the proposed algorithm to achieve both reduction in
the operating costs and meeting the load demand. The proposed
procedure can be implemented with different loads and for periods
more than one day.

References
[1] Hernandez-Aramburo CA, Green TC, Mugniot N. Fuel consumption
minimization of a microgrid. IEEE Trans Ind Appl 2005;41(3):67381.
[2] Kothari DP, Dhillon JS. Power system optimization. second ed. Prentice Hall of
India Private Limited; 2011.
[3] Liu G, Yang J, Whidborne J. Multiobjective optimisation and
control. Baldock: Research Studies Press Ltd.; 2003.
[4] Andersson J. A survey of multiobjective optimization in engineering design.
Technical report, Department of Mechanical Engineering, Linkoping
University, LiTH-IKP-R-1097, 2000.
[5] Elmusrati M, Riku J, Koivo H. Multiobjective distributed power control
algorithm for CDMA wireless communication systems. IEEE Trans Vehicular
Technol 2007;56(2):77988.
[6] Zhang W, Liu Y. Multi-objective reactive power and voltage control based on
fuzzy optimization strategy and fuzzy adaptive particle swarm. Int J Electr
Power Energy Syst 2008;30(9):52532.
[7] Barin A, Pozzatti LF, Canha LN, Machado RQ, Abaide AR, Arend G. Multiobjective analysis of impacts of distributed generation placement on the
operational characteristics of networks for distribution system planning. Int J
Electr Power Energy Syst 2010;32(10):115764.
[8] Manisa Pipattanasomporn. A study of remote area internet access with
embedded power generation. Faculty of the Virginia Polytechnic Institute
and State University, PhD-thesis, December 2004.

735

[9] Mohamed Faisal A, Koivo Heikki. System modelling and online optimal
management of microgrid using mesh adaptive direct search. Int J Electr
Power Energy Syst 2010;32(5):398407.
[10] Nanda J, Kothari DP, Lingamurthy KS. Economic emission load dispatch
through goal programming technique. IEEE Trans Energy Conv 1988;3:2632.
[11] Nanda J, Lakshman H, Kothari ML. Economic emission load dispatch with line
ow constraints using a classical technique. IEE Proc Gen Inst Transm Distrib
1994;141(1):110.
[12] Huang CM, Yang HT, Huang CL. Bi-objective power dispatch using fuzzy
satisfaction-maximizing decision approach. IEEE Trans Power Syst
1997;12:171521.
[13] Miettinen K. Nonlinear multiobjective optimization. Boston: Kluwer Academic
Publishers; 1998.
[14] Azmy AM, Erlich I. Online optimal management of PEM fuel cells using neural
networks. IEEE Trans Power Deliv 2005;29(2):10518.
[15] Abido MA. Multiobjective evolutionary algorithms for power dispatch
problem. IEEE Trans Evolut Comput 2006;10(3):31529.
[16] Talaq JH, El-Hawary F, El-Hawary ME. A summary of environmental/economic
dispatch algorithms. IEEE Trans Power Syst 1994;9:150816.
[17] Morgantown W. Emission rates for new DG technologies. The regulatory
assistance project. <http://www.raponline.org/ProjDocs/DREmsRul/Collle/
DGEmissionsMay2001.pdf>.
[18] Audet C, Dennis Jr JE. Mesh adaptive direct search algorithms for constrained
optimization. SIAM J Optim 2006;17(1):188217.
[19] Abramson MA, Audet C. Convergence of mesh adaptive direct search to second
order stationary points. SIAM J Optimi 2006;17(2):60619.
[20] Mohamed F, Koivo H. Online management of microgrid with battery storage
using multiobjective optimization. In: The rst international conference on
power engineering, energy and electrical drives (POWERENG07), 1214 April
2007, Setubal, Portugal.
[21] Mohamed F, H Koivo. System modelling and online optimal management of
microgrid with battery storage. In: 6th international conference on renewable
energies and power quality (ICREPQ07), 2628 March 2007, Sevilla, Spain.
[22] Mohamed F, Koivo H. Microgrid online management and balancing using
multiobjective optimization. In: Proceedings of IEEE power tech, 2007, 15
July 2007, Lausanne, Switzerland.
[23] Michel V, Sascha B, Britta B, Herve C, Nipon K, Franz K, et al. Expandable hybrid
systems for multi-user mini-grids. In: Proceedings of 17th European
photovoltaic solar energy conference, October 2001, Munich, Germany. p.
23305
[24] Zahedi A, Kalam A. Balancing cost and performance in a PV /wind/battery
hybrid power system. In: Proceedings of Australasian universities power
engineering conference (AUPEC00), Brisbane, Australia, 2000. p. 2803
[25] Abido MA. Environmental/economic power dispatch using multiobjective
evolutionary algorithms. IEEE Trans Power Syst 2003;18(4):152937.