Professional Documents
Culture Documents
j=1
u
j
(p
j
)
s.t.
N+1
j=1
p
j
= P
max
=
N
j=1
P
nom
j
(1)
where p
i
and u
i
(p
i
), for i = 1, . . . , N, correspond to the power
supplied by the ith DG and its utility function, respectively;
p
N+1
and u
N+1
are the ctitious power assigned to the
interface agent and its utility function, respectively; u
tot
is the
total utility of the microgrid; and P
max
is the maximum power
that the microgrid can generate. Note that, in each negotiation
period, P
max
is fully allocated among the N + 1 agents, and
that, the power dispatched to the interface agent is used to adjust
the real demanded power P
d
from the microgrid. This desired
power is given by
P
d
=
N
i=1
p
i
where p
i
is the optimal power dispatched to the ith DG in each
negotiation period. Fig. 2 shows the dispatch process and the
interactions between agents.
To solve the optimization problem and according to common
cost functions of generation units [16], quadratic utility func-
tions are used for the DGs. These functions are given by
u
i
(p
i
) =
p
i
c
i
P
nom
i
(p
i
2P
nom
i
) , for i = 1, . . . , N
where the nominal power (P
nom
i
) and the generation cost factor
(c
i
) of the ith DG dene its utility function. The maximum
utility is achieved when the DG generates its nominal power
4562 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 10, OCTOBER 2011
Fig. 3. (Dashed) Utility functions and (solid) marginal utilities for four DGs
with equal cost factors and different nominal powers. In this case, c
i
=
0.7 per unit (p.u.), for i = 1, . . . , 4; P
nom
= [172, 47, 66, 106] kWand P
d
=
225 kW.
(i.e., u
i
(P
nom
i
) = P
nom
i
/c
i
), and it satises the expected
relationship between an economic utility, the nominal power,
and the costs of the generation of the units. Fig. 3 presents a
scheme of the proposed functions. On the other hand, the utility
function of the interface agent is dened as
u
N+1
(p
N+1
) = mp
N+1
where m is a tunable parameter that adjusts the power de-
manded from the microgrid. Given these utilities, the solution
of this constrained nonlinear separable concave problem [18] is
obtained when
du
1
(p
1
)
dp
1
= . . . =
du
N
(p
N
)
dp
N
=
du
N+1
(p
N+1
)
dp
N+1
= m.
In other words, when all rst derivatives of the utility func-
tions of the generators (marginal utilities) are equal to a value
(bidding price) and the constraints are satised, the solution is
optimal. In this case, the optimal resource allocation is reached
when
2
c
i
_
1
p
i
P
nom
i
_
= m, for all i = 1, . . . , N. (2)
If (2) is solved for p
i
and taking the sum over all i =
1, . . . , N, the adjustable parameter m is obtained as a function
of the parameters of the DGs, with a microgrid capacity P
max
and a demanded power P
d
, i.e.,
m =
2
N
j=1
c
j
P
nom
j
(P
max
P
d
). (3)
Now, if (3) is replaced in (2) for a specic value of demanded
power P
d
, the optimal power dispatched for the ith DG is
given by
p
i
= P
nom
i
c
i
P
nom
i
N
j=1
c
j
P
nom
j
(P
max
P
d
). (4)
Notice that, if P
d
= P
max
(i.e., the demanded power is
equal to the microgrid capacity), p
i
= P
nom
i
, which implies
that all DGs must generate their nominal power obtaining the
maximum possible utility. However, if P
d
< P
max
, p
i
depends
on the particular generators parameters, and a merit dispatch
is performed by solving the maximization problem in (1). This
process is shown in Fig. 3 for a specic case.
An interesting feature of (4) is that for some values of
the generators parameters and demanded power, the optimal
dispatched power could be negative. For instance, if P
d
is low
and the generation cost of the ith generator is high, the second
term of the right-hand side of (4) could be greater than P
nom
i
,
and then, p
i
< 0. This has no sense for generation units, and
as a consequence, that generator must not be dispatched (i.e.,
p
i
= 0). This truncation affects the constraints dened in
(1), i.e.,
N
j=1
p
j
> P
d
. Therefore, the required power from
the microgrid is exceeded, and there will be a period of over-
generation and overcost. Taking this into account, it is neces-
sary to include more constraints in the optimization problem
(e.g., p
i
0, for all i), making the problem difcult to solve
analytically.
For this reason, numerical methods of nonlinear constrained
optimization could be used to solve (1) with the new con-
straints. This could be performed using the proposed MAS
structure, where the MGCC starts the negotiation period cal-
culating m according to the demanded power. After that, each
local agent calculates its utility and the derivative of the utility
with the parameters of each generator, and the optimization
problem is solved iteratively based on some numerical effective
method (e.g., Newton method with the constraint included into
the objective function [18]). These methods could be computa-
tionally expensive, whenever there is a large number of DGs in
a microgrid. In the next section, a simpler model is proposed,
which solves the extended optimization problem using a set of
rst-order differential equations.
B. RD
The RD was introduced in [6], and it is a simple model
of how natural selection, via differential tness, affects the
proportions of individuals that are looking for some nutrients in
different habitats of an environment. A replicator is analogous
to an individual in a population and has associated a tness
function that depends on the parameters in each habitat. The
amount of individuals in each habitat changes as a result of their
mutual interactions and their relative nesses [30]. Originally,
the RD takes two individuals randomly from a homogeneous
population, and then, they play a game, where the one that has
a better payoff will increase its population in a habitat. This can
PANTOJA AND QUIJANO: POPULATION DYNAMICS APPROACH FOR THE DISPATCH OF DGs 4563
be seen as a linear game because the selection of the payoff
matrices is constant [6], [31]. Here, such standard formulation
is extended to represent a nonlinear game dened by logistic
tness functions, assuming that there is a homogeneous popu-
lation as in the original formulation.
In this model, each individual has N pure strategies, which
correspond to choosing which habitat to live in for its entire life,
and the number of individuals in the population is constant, i.e.,
N
i=1
p
i
= P
d
, for some P
d
> 0, and all t 0.
The general model of replicator is given by
p
i
p
i
= [{tness of individuals that play strategy i} . . .
. . . {average tness in population}] .
The left-hand side is the normalized rate of increase in
the population share playing strategy i. The right-hand side
indicates that if i-strategists are more (less) successful than
the average, their population share will increase (decrease).
The tness function f
i
(p
i
) is related to the goodness of the
individuals in the habitat i, and the average tness is given by
f =
1
P
d
N
j=1
p
j
f
j
(p
j
). (5)
Therefore, the RD is
p
i
= p
i
_
f
i
(p
i
)
f
_
. (6)
From [31], the choice of
f = (1/P
d
)
N
j=1
p
j
f
j
guarantees
the invariance of the constraint set
p
dened as
p
=
_
p R
N
+
:
N
i=1
p
i
= P
d
_
.
In other words, if p(0)
p
, then p(t)
p
, for all t 0.
If (6) is taken in steady state (i.e., p
i
= 0), and if p
i
> 0, the
equilibrium point is reached when
f
i
(p
i
) =
f
i
, for the case
of a microgrid with four DGs with different capacities and cost
factors. For a xed demanded power P
d
, the RD model evolves
until it reaches the steady state. At this equilibrium point, the
average tness
f
4
= 0. This
level is given by
f
(P
d
), where the higher the P
d
, the lower the
2
> p
1
). In case of an increment of
the demanded power, the average tness at the equilibrium
is reduced and could be enough to dispatch some power for
the more expensive generator to cover the additional demand.
In general, the resource allocation depends on the generators
characteristics and on the demanded power for the microgrid.
Furthermore, the model has the advantage that, if one or more
DGs are not dispatched, the demanded power is always covered
given the invariance of the constraint set
p
that guarantees that
N
i=1
p
i
= P
d
.
In order to analyze the optimality of the equilibrium point of
the model, the next proposition shows the relationship between
the market-based approach and the RD model.
Proposition 3.1: The optimal power dispatched for the ith
DG under the RD, with a tness function such as the one in
(8), is the same optimal power dispatched by the market-based
strategy given by (1).
Proof: In order to prove this proposition, the denition of
the steady state of the system(7) is used to observe the inuence
4564 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 10, OCTOBER 2011
of the tness functions. Replacing the tness (8) and average
tness (5) in (7), it is obtained that
1
c
i
_
1
p
i
P
nom
i
_
=
1
P
d
N
j=1
p
j
c
j
_
1
p
j
P
nom
j
_
. (9)
Solving for p
i
in the left-hand side of (9) and summing over
i = 1, . . . , N
N
i=1
p
i
=
N
i=1
P
nom
i
+
1
P
d
j=1
_
_
p
j
_
2
P
nom
j
p
j
c
j
P
nom
j
_
i=1
c
i
P
nom
i
. (10)
Let
=
N
j=1
_
_
p
j
_
2
P
nom
j
p
j
c
j
P
nom
j
_
(11)
=
N
i=1
c
i
P
nom
i
. (12)
By hypothesis,
N
i=1
p
i
= P
d
and
N
i=1
P
nom
i
= P
max
,
which implies that (10) is given by
P
d
= P
max
+
1
P
d
. (13)
Solving for in (13) and with some additional terms that
hold the equality (i.e., summing P
max
P
max
and multiply-
ing by /), can be expressed as
=
P
d
P
max
2
(P
max
P
max
+ P
d
). (14)
With the hypothesis used to dene P
max
and P
d
, and with
(12), is given by
=
N
j=1
_
P
nom
j
( c
j
P
max
+ c
j
P
d
)(P
d
P
max
)
2
_
. (15)
Using (11) and (15) and solving for the feasible p
j
, it is
obtained that
p
j
= P
nom
j
c
j
P
nom
j
(P
max
P
d
) (16)
which corresponds to the optimal power dispatched with the
market-based strategy given by (4). Then, the equilibrium point
of the proposed RD model is optimal in the maximization of the
total utility of the microgrid.
In a complex system (e.g., large N) with possible uncertain-
ties in the parameters, and using only local information, solving
the rst-order differential (6) can be simplied by a distributed
and iterative process assigning simple tasks to the agents in
the MAS model of a microgrid. In order to approximate the
Fig. 5. Agent scheme for the RD strategy at the kth iteration time. The update
function for the dispatched power (g()) is given by the discrete-time RDmodel
in (17).
continuous-time RD, an option is the discrete-time version of
the model proposed in [33], [34], which is given by
p
i
[k + 1] = p
i
[k]
1/T
s
+ f
i
[k]
1/T
s
+
f[k]
, for k = 0, 1, 2, . . . (17)
where T
s
is a step size related to the sample time in the
discrete dynamical system. Although the equilibrium point of
the discrete RD may differ from the corresponding point in the
continuous one, with an appropriate choice of the step size,
the approximation between continuous and discrete version
is very close (in this case, given the linearity of the tness
functions). Moreover, the invariance at the equilibrium point
N
i=1
p
i
= P
d
is held in the discrete version [33]. With an
enough small sample time, chosen to guarantee the convergence
of the algorithm with a xed absolute error stop criterium,
Fig. 5 shows how the algorithm can be simplied for the agent
information exchange. With the information of the total power
available to dispatch P
d
(obtained previously in a higher level
in the hierarchy of the distribution system), and with random
initial conditions such that p
i
(0)
p
, for all i = 1, . . . , N,
the MGCC coordinates the beginning of the process. At the
kth iteration, the central controller calculates the average tness
f[k] using (5) and sends this information to all the LCs in
the microgrid. Each LC assigned to each DG approximates (6)
and calculates p
i
[k + 1] with (17), which, in practice, is easy
to perform if
f[k] is calculated beforehand. After this, each
LC updates its tness f
i
[k + 1] with the generators parameters
using (8). The information that contains p
i
[k + 1] and f
i
[k + 1]
is sent back to the MGCC, and the process is repeated until
the steady state is reached. The nal dispatched power p
i
will
correspond to the desired power of the ith generator.
Next, some simulation results are presented using RD and
market-based strategies for a typical hourly demand prole in
Colombia.
IV. SIMULATION RESULTS AND DISCUSSION
To avoid an excessive penetration of distributed generation
in a microgrid and to assume that power ow constraints are
respected, the IEEE 13-node test feeder [35] is used as a
microgrid reference model, and some results shown in [29] are
chosen to dene the nominal powers and placement of four
PANTOJA AND QUIJANO: POPULATION DYNAMICS APPROACH FOR THE DISPATCH OF DGs 4565
Fig. 6. Power dispatch for each DG with MBC and RD strategies for the
rst case.
power controllable generators in the system. For a penetration
level around 15% (that can be considered reliable ac-
cording to [36]), the vector P
nom
= [P
nom
1
, . . . , P
nom
4
] =
[172, 47, 66, 106] kW is taken as the xed nominal powers of
the DGs. The generators are located in nodes 671, 611, 652,
and 675, respectively, according to the numbering suggested
in the standard (see [35] for more details). These nodes are
in the end of the radial distribution systems and have bigger
loads associated to them. To simulate the demanded power from
the microgrid, a typical hourly demand prole in Colombia is
dened with the microgrid capacity as the maximum load at
rush hour. To introduce variations on the demand prole, this
curve is scaled by a demand factor d
f
[0, 1], which allows
us to assume adequate values for the demanded power (i.e.,
0 P
d
P
max
). This power is assumed to be constant during
all the negotiation period of 1 h.
To compare the analytical MBC and the RD strategies,
and to observe the inuence of the costs and the demand
factor in the dispatched power from each generator, two
cases are dened: 1) using d
f
= 1 and a cost vector c =
[c
1
, . . . , c
4
] = [1, 0.7, 0.4, 0.8] p.u. and 2) with d
f
= 0.55 and
c = [0.2, 0.1, 0.8, 1] p.u.
Fig. 6 shows the results for the dispatch in the rst case. For
the chosen parameters, both strategies have the same response
given that no constraints are violated, and the dispatch for
all DGs follows the demand prole. At the peak hour, all
DGs generate at their nominal power according to the demand
factor chosen for this case. Moreover, the demand is adequately
covered (i.e.,
4
i=1
p
i
= P
d
, for all t) by the two strategies.
In this case, the dispatched powers p
i
are close to the optimal
ones calculated by (4), showing that the MAS method can be
implemented with appropriate results for both strategies.
The second case exposes some differences between the two
strategies. Since the demand factor is reduced to 55% and the
cost of the fourth generator is signicantly higher than the
others, this DG must be dispatched only in the periods of high
demand. At the same time, the cheaper generator (the second
one) has a resource allocation that is almost constant and close
Fig. 7. Power dispatch for each DG with MBC and RD strategies for the
second case.
to their nominal power. The comparison between the two strate-
gies is shown in Fig. 7, where the greater difference is observed
in the dispatch of the rst DG, which has the greater capacity
and a relative low cost. At the beginning of the day, when
the more expensive generators are not dispatched, the power
allocated to this DG by the MBC strategy is remarkably higher
than the dispatch done by the RD model. In the analytical MBC
case, the optimal power for the expensive DGs in low-demand
periods is negative, and cheaper generators must compensate
this condition to respect the main constraint in the problem
(i.e.,
N+1
j=1
p
j
= P
max
, including the interface agents power).
Given the inclusion of the additional constraints in the numeri-
cal MAS optimization problem (i.e., p
i
0, for all i), some of
these conditions are activated, and the assigned power for these
generators is truncated to zero without optimal reallocation of
the negative power dispatched originally. Because of this, the
rst DG is highly dispatched, and the solution is not optimal.
This situation is more visible in the left graph in Fig. 8, which
shows the demand covered by all generators in the microgrid.
Here, it is notable that there are periods of overgeneration in
the MBC dispatch, which implies an increment of total costs in
the microgrid (as it can be seen in the right graph in Fig. 8).
However, the RD strategy always covers the demanded power
from the microgrid in 100%, optimizing the costs of generation.
In other words, in periods of low demand for the RD case, the
more expensive generators are not dispatched until the tradeoff
cost capacity is reached in order to cover the desired power.
This is due to the invariance of
4
i=1
p
i
= P
d
.
Based on these simulation results, the efciency of the RD
algorithm when there are active constraints is shown. Other
than the optimality of the solution, which maximizes the total
utility of the DGs in the microgrid, the computational cost of
the algorithm is reduced, compared with the MBC strategy. For
4566 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 58, NO. 10, OCTOBER 2011
Fig. 8. Demand coverage and total costs for MBC and RD strategies for the
second case.
instance, the execution time for the test programs in the rst
study case is 7.63 s for MBC and 2.18 s for RD. Although
this difference is remarkable, the more important aspect of the
proposed strategy is the facility of implementation in an agent
scheme with a high number of generators in each microgrid.
According to the simple study cases presented, if signicant
changes in the DGs parameters, demand factors, and/or utility
functions occur during the negotiation period, a new approxi-
mation must be done in order to adjust the information and pro-
cedures of the MGCC in the MBC strategy. Therefore, another
advantage of the RD strategy is the needless reconguration
for the changes in the system. Moreover, if the parameters of a
generator change between periods of negotiation, the correction
in the strategy is only the change in the information for the local
agent in that DG. The only thing that is necessary to guarantee
is that, at the beginning of each negotiation period, the initial
conditions satisfy p
i
(0)
p
, for all i = 1, . . . , N.
V. CONCLUSION
An approach based on evolutionary game theory applied to
the dispatch of DGs in a microgrid has been proposed. The main
advantages of the RD strategy are the simplicity of the model,
the adaptability of the tness functions, and the optimality of
the solution in a dynamic resource allocation problem. Taking
this into account, this strategy can be implemented in a MAS
scheme for large-scale power distribution systems, and it is
shown that the RD dispatch maximizes the total utility of the
microgrid. Moreover, a comparison between the performance of
the RD model and a market-based strategy has been presented
for simple simulation cases. These simulation results show
interesting possible applications in real distribution networks.
Some of the future directions include studies for changes in the
tness functions including other parameters of the system (e.g.,
to provide ancillary services in microgrids) and the use of the
RD strategy in larger test systems.
ACKNOWLEDGMENT
The authors would like to thank the SILICE group, Uni-
versidad de los Andes, Bogot, Colombia, for their input. The
authors would also like to thank the anonymous reviewers for
their suggestions.
REFERENCES
[1] T. Ackermann, G. Andersson, and L. Sder, Distributed generation: A
denition, Elect. Power Syst. Res., vol. 57, no. 3, pp. 195204, Apr. 2001.
[2] D. iljak and A. Ze cevi c, Control of large-scale systems: Beyond decen-
tralized feedback, Annu. Rev. Control, vol. 29, no. 2, pp. 169179, 2005.
[3] E. Bueno, . Hernndez, F. Rodrguez, C. Girn, R. Mateos, and
S. Cobreces, A DSP- and FPGA-based industrial control with high-speed
communication interfaces for grid converters applied to distributed power
generation systems, IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 654
669, Mar. 2009.
[4] J. Lopes, N. Hatziargyriou, J. Mutale, P. Djapic, and N. Jenkins, Inte-
grating distributed generation into electric power systems: A review of
drivers, challenges and opportunities, Elect. Power Syst. Res., vol. 77,
no. 9, pp. 11891203, Jul. 2007.
[5] N. Hatziargyriou, Microgrids, the key to unlock distributed energy
resources?, IEEE Power Energy Mag., vol. 6, no. 3, pp. 2629,
May/Jun. 2008.
[6] P. Taylor and L. Jonker, Evolutionary stable strategies and game
dynamics, Math. Biosci., vol. 40, no. 1/2, pp. 145156, Jul. 1978.
[7] C. Caizares and S. Kodsi, Power system security in market clearing
and dispatch mechanisms, in Proc. IEEE Power Eng. Soc. Gen. Meeting,
2006, pp. 16.
[8] J. Guerrero, J. Vasquez, J. Matas, M. Castilla, and L. Garcia de Vicuna,
Control strategy for exible microgrid based on parallel line-interactive
UPS systems, IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 726736,
Mar. 2009.
[9] A. Pantoja and N. Quijano, Modeling and analysis for a temperature
system based on resource dynamics and the ideal free distribution, in
Proc. Amer. Control Conf., 2008, pp. 33903395.
[10] J. Vidal, Learning in multiagent systems: An introduction from a game-
theoretic perspective, in Adaptive Agents, vol. 2636, Lecture Notes in
Computer Science. New York: Springer-Verlag, 2003, pp. 202215.
[11] N. Quijano and K. Passino, The ideal free distribution: Theory and
engineering application, IEEE Trans. Syst., Man, Cybern. B, Cybern.,
vol. 37, no. 1, pp. 154165, Feb. 2007.
[12] M. Anastasopoulos, D. Petraki, R. Kannan, and A. Vasilakos, TCP
throughput adaptation in WiMax networks using replicator dynamics,
IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 40, no. 3, pp. 647655,
Jun. 2010.
[13] A. Dimeas and N. Hatziargyriou, Operation of a multiagent system for
microgrid control, IEEE Trans. Power Syst., vol. 20, no. 3, pp. 1447
1455, Aug. 2005.
[14] Z. Qiu, G. Deconinck, N. Gui, R. Duan, and R. Belmans, A market-
based MAS framework for microgrids, in Proc. 17th IFAC World Congr.,
Jul. 2008, pp. 11 05311 058.
[15] M. Baran and I. El-Markabi, A multiagent-based dispatching scheme for
distributed generators for voltage support on distribution feeders, IEEE
Trans. Power Syst., vol. 22, no. 1, pp. 5259, Feb. 2007.
[16] M. Aunedi, D. Skrlec, and G. Strbac, Optimizing the operation of dis-
tributed generation in market environment using genetic algorithms, in
Proc. 14th IEEE MELECON, 2008, pp. 780785.
[17] D. Srinivasan and J. Chazelas, Heuristics-based evolutionary algorithm
for solving unit commitment and dispatch, in Proc. IEEE Congr. Evol.
Comput., 2005, vol. 2, pp. 15471554.
[18] F. Ygge and H. Akkermans, Power load management as a computational
market, in Proc. 2nd ICMAS, 1996, pp. 393400.
[19] R. Cressman and V. Krivan, Migration dynamics for the ideal free distri-
bution, Amer. Naturalist, vol. 168, no. 3, pp. 8497, 2006.
[20] S. McArthur, M. Davidson, M. Catterson, L. Dimeas, D. Hatziargyriou,
F. Ponci, and T. Funabashi, Multiagent systems for power engineering
applicationsPart I: Concepts, approaches, and technical challenges,
IEEE Trans. Power Syst., vol. 22, no. 4, pp. 17431752, Nov. 2007.
[21] F. Wu, K. Moslehi, and A. Bose, Power system control centers: Past,
present, and future, Proc. IEEE, vol. 93, no. 11, pp. 18901908,
Nov. 2005.
[22] G. Pappas, G. Lafferriere, and S. Sastry, Hierarchically consistent control
systems, IEEE Trans. Autom. Control, vol. 45, no. 6, pp. 11441160,
Jun. 2000.
[23] G. Pepermans, J. Driesen, D. Haeseldonckx, R. Belmans, and
W. Dhaeseleer, Distributed generation: Denition, benets and issues,
Energy Policy, vol. 33, no. 6, pp. 787798, Apr. 2005.
[24] W. Li, G. Joos, and J. Belanger, Real-time simulation of a wind turbine
generator coupled with a battery supercapacitor energy storage system,
IEEE Trans. Ind. Electron., vol. 57, no. 4, pp. 11371145, Apr. 2010.
[25] S. Nishikata and F. Tatsuta, A new interconnecting method for wind
turbine/generators in a wind farm and basic performances of the
PANTOJA AND QUIJANO: POPULATION DYNAMICS APPROACH FOR THE DISPATCH OF DGs 4567
integrated system, IEEE Trans. Ind. Electron., vol. 57, no. 2, pp. 468
475, Feb. 2010.
[26] C. Liu, K. Chau, and X. Zhang, An efcient wind-photovoltaic hybrid
generation system using doubly-excited permanent-magnet brushless ma-
chine, IEEE Trans. Ind. Electron., vol. 57, no. 3, pp. 831839, Mar. 2010.
[27] M. Cacciato, A. Consoli, R. Attanasio, and F. Gennaro, Soft-switching
converter with HF transformer for grid-connected photovoltaic systems,
IEEE Trans. Ind. Electron., vol. 57, no. 5, pp. 16781686, May 2010.
[28] N. Rahim and J. Selvaraj, Multi-string ve-level inverter with novel
PWM control scheme for PV application, IEEE Trans. Ind. Electron.,
vol. 57, no. 6, pp. 21112123, Jun. 2010.
[29] C. Tautiva, A. Cadena, and F. Rodriguez, Optimal placement of distrib-
uted generation on distribution networks, in Proc. 44th Int. UPEC, 2009,
pp. 15.
[30] A. Menon, K. Mehrotra, C. Mohan, and S. Ranka, Optimization
using replicators, in Proc. 6th Int. Conf. Genetic Algorithms, 1995,
pp. 209216.
[31] J. Hofbauer and K. Sigmund, Evolutionary Games and Population
Dynamics. Cambridge, U.K.: Cambridge Univ. Press, 1998.
[32] N. Britton, Essential Mathematical Biology. New York: Springer-
Verlag, 2003.
[33] J. W. Weibull, Evolutionary Game Theory. London, U.K.: MIT Press,
1997.
[34] D. Fudenberg and D. Levine, The Theory of Learning in Games.
London, U.K.: MIT Press, 1998.
[35] W. Kersting, Radial distribution test feeders, in Proc. IEEE Power Eng.
Soc. Winter Meeting, 2001, vol. 2, pp. 908912.
[36] A. Azmy and I. Erlich, Impact of distributed generation on the stability
of electrical power system, in Proc. IEEE Power Eng. Soc. Gen. Meeting,
2005, pp. 10561063.
Andrs Pantoja received the B.S. degree in elec-
tronics engineering from Universidad Nacional,
Manizales, Colombia, in 1999 and the M.S. degree
in electronics engineering from Universidad de los
Andes, Bogot, Colombia, in 2008, where he is
currently working toward the Ph.D. degree in the
School of Engineering.
Since 2003, he has been an Assistant Professor
with the Departamento de Ingeniera Electrnica,
Universidad de Nario, Pasto, Colombia. His re-
search interests are dynamic resource allocation,
distributed generation, distributed control in smart grids and buildings, and
coordination in large-scale systems.
Nicanor Quijano (S02M07) received the B.S.
degree in electronics engineering from Ponticia
Universidad Javeriana, Bogot, Colombia, in 1999
and the M.S. and Ph.D. degrees in electrical and
computer engineering from The Ohio State Univer-
sity, Columbus, in 2002 and 2006, respectively.
In 2007, he was with the Departamento de In-
geniera Elctrica y Electrnica, Universidad de los
Andes (UNIANDES), Bogot, as an Assistant Pro-
fessor. He is currently an Associate Professor and the
Head of the research group in control and automation
systems (GIAP, UNIANDES). His research interests include smart grids and
distributed generation, hierarchical and distributed methods using bioinspired
techniques (e.g., foraging theory and evolutionary game theory), dynamic
resource allocation, and switched and hybrid systems.