Академический Документы
Профессиональный Документы
Культура Документы
, An optimal power ow plus transmission costs solution, Electr. Power Syst. Res.
(2009), doi:10.1016/j.epsr.2009.03.005
ARTICLE IN PRESS
G Model
EPSR-2851; No. of Pages 7
Electric Power Systems Research xxx (2009) xxxxxx
Contents lists available at ScienceDirect
Electric Power Systems Research
j our nal homepage: www. el sevi er . com/ l ocat e/ epsr
An optimal power ow plus transmission costs solution
Y. Pablo O nate, Juan M. Ramirez
pbest
i
x
i
) . . .
+C
2
Rand() (
gbest
i
x
i
) (2)
where iter is the current iteration; C
1
and C
2
are two positive learn-
ing factors; rand() and Rand() are two randomly generated values
within [0,1]; w is known as the inertia weight, and it plays the role
of balancing the global and local searchperformed by the algorithm
[32,33].
3. Objective function
The OPF-SC goal is to optimize an objective function subject to
equality and inequality constraints. The optimization problem is
formulated as follows:
MinF
obj
(u, y) (3)
subject to
g(u, y) = 0 (4)
h
min
h(u, y) h
max
(5)
where g(,) represents the equality constraints set; h(,) is the
inequality constraints set; uare the state variables; y represent both
integer and continuous control variables.
3.1. Objective function
In this paper, the minimization of the total cost includes the pre-
contingency cost (superscript 0) besides each credible contingency
cost (superscript k). Thus, the objective function is constituted by
three terms: (i) the generating costs, (ii) the transmission costs, and
(iii) the consumer benet [34]
F
obj
= min
C
0
+
K
k=1
C
k
(6)
C
0
=
NG
i=1
(U
0
i
f (P
0
Gi
) +SUC
i
) +TC
0
(FLW
0
mn
)
NLoad
j=1
B(P
Loadj
) (7)
C
k
=
NG
i=1
(U
k
i
f (P
k
Gi
) +SUC
i
) +TC
k
(FLW
k
mn
)
NLoad
j=1
(B(P
k
Loadj
)
D(P
k
Loadj
, P
0
Loadj
)) (8)
where Kis the total number of credible contingencies; C
0
represents
the base case operating cost; U
0
i
i-th units pre-contingency state,
1-ON, 0-OFF, NG is the total number of available generators; f (P
0
Gi
)
i-th generators function cost at time t; P
0
Gi
active power supplied by
the i-thgenerator at the pre-contingency state, SUC
i
i-thgenerators
start-up cost; TC
0
() pre-contingency transmission cost, (9) NLoad
total number of load buses, B(P
Loadj
) is the consumer benet curve
for j-thloadat timet; P
Loadj
pre-contingencyactivepower consump-
tion at the j-th load; C
k
credible contingencies cost; D() represents
the load interruption cost; P
k
post-contingency active power con-
sumptionat thej-thload; TC
k
() post-contingencytransmissioncost,
dened as (9)
TC =
M
m=1
c
m
(MW
gi,m
)L
m
MW
gi,m
(9)
where, M lines set; c
m
() cost per MW per unit length of line m;
L
m
length of line m in miles; MW
gi,m
ow in line m, due to the i-th
generator.
3.2. Constraints
In the following, different constraints included in the OPF-SC
formulation are detailed.
3.2.1. Equality constraints at the pre- and post-contingency states
This is the set of nonlinear power balance equations that govern
the steady-state power ow formulation, both at pre- and post-
contingency state, dened as follows:
0 =
NG
i=1
P
0
Gi
NLoad
j=1
(P
0
Loadj
) P
0
Loss
(10)
0 =
NG
i=1
Q
0
Gi
NLoad
j=1
Q
0
Loadj
Q
0
Loss
(11)
0 =
NG
i=1
P
k
Gi
NLoad
j=1
P
k
Loadj
P
0
Loss
, for k = 1, 2, . . . , K (12)
0 =
NG
i=1
Q
k
Gi,t
NLoad
j=1
Q
k
Loadj,t
Q
k
Loss t
, for k = 1, 2, . . . , K (13)
where P
Loss
represents the total active power losses, Q
Loss
are the
total reactive power ones, and K is the total number of credible
contingencies.
3.2.2. Inequality constraints at the pre- and post-contingency
states
This is the set of continuous and discrete constraints represent-
ing the systems operational and security limits as bounds.
(i) The active power generated by each unit must satisfy the
maximum and minimum operating limits, both for pre- and
post-contingency states.
P
Gi MIN
P
0
Gi
P
Gi MAX
(14)
P
Gi MIN
P
k
Gi
P
Gi MAX
(15)
Please cite this article in press as: Y. Pablo O nate, et al., An optimal power ow plus transmission costs solution, Electr. Power Syst. Res.
(2009), doi:10.1016/j.epsr.2009.03.005
ARTICLE IN PRESS
G Model
EPSR-2851; No. of Pages 7
Y. Pablo Onate et al. / Electric Power Systems Research xxx (2009) xxxxxx 3
Fig. 1. Proportional sharing concept.
(ii) Voltage magnitudes at each load bus must be close enough to
the reference voltage
min
V
ref
V
j
(16)
where V
ref
is the reference voltage magnitude, V
j
is the voltage
magnitude at the j-th load bus.
(iii) The active power ow through each branch of the network
must satisfy the security limits
|FLOW
0
ij
| FLOW
0
ij MAX
, i, j, i / = j (17)
|FLOW
k
ij
| FLOW
k
ij MAX
, i, j, i / = j (18)
where FLOW
0
ij MAX
, FLOW
k
ij MAX
, represent the maximumactive
power that should ow through the branch connecting
the buses i j, during the pre-contingency and each post-
contingency state, respectively.
4. Transmission cost allocation
To calculate the transmission cost allocation, the average partic-
ipation factor (APF) method [5,8,9] has been used. It is based on a
general transportation problemof howthe ows are distributed on
a meshed network. The only requirement is related to the Kirch-
hoffs rst law fulllment.
Once the power ow scenario is determined, the main idea of
this methodology is to determine the agents (generator and loads)
participationfactor inthe owinlinks. Inthis sense, it will be possi-
ble to trace the owof electricity froma generator to the loadbuses.
The principle adopted to trace these ows is the proportional shar-
ingprinciple, schematizedinFig. 1. Theassumptionmadeis that the
bus is a perfect mixer of all incoming ows, so that it is impossible
to determine which particular inowing electron goes into which
particular outgoing line. It may be assumed that each element ow
leaving the i-th bus can be decomposed into M shares, where M is
the number of incoming ows (generator and elements injecting
power into bus i). The amount of such shares has the same propor-
tion that the incoming ows in the total power injected into bus i
[5,8].
5. Proposed solution
The proposed methodology runs as follows:
Initial population.
In this paper, a particle is composed by continuous and dis-
crete control variables. The continuous ones include the generators
active power output and the loads active power; the discrete
variables are associated with the transformers-tap setting and var-
injection values of the switched shunt capacitors/reactors, Fig. 2.
The population is constituted by K+1 matrices, one for the base
case and one for each of the K credible contingencies (subpopula-
tions) of dimension (NG+ND) NIND; where K represents the total
number of accounted contingencies; NG is the number of continu-
ous variables; ND is the number of discrete variables; NIND is the
number of particles.
For each scenario (pre- and post-contingency states), the initial
active power is a percentile and is randomly allocated among the
NG available thermal units, such that the load is satised (19)
P
n
Gi
=
rand(NG)
rand(NG)
Nload
j=1
P
LOADj
for n = 0, 1, . . . , K (19)
where, rand(NG) represents a vector of size NG randomly generated
within the interval [0,1].
The transformer-tap setting and var-injection values of the
switched shunt capacitor/reactor (discrete variable) are randomly
generated between upper and lower limits. An operator is included
to ensure that each discrete variable is rounded to its nearest deci-
mal integer value that represents the physical operating constraint
of a given variable.
round(random[ST
min
i
, ST
max
i
], ) (20)
where represents the most signicant decimal.
5.1. Reconstruction operators
In this paper, the reconstruction operators have been used to
satisfy the units constraints. Taking into account the inequalities
within a conventional formulation, by using penalty functions, for
instance, results in an execution where it is very likely to make
wrong decisions, due to the use of excessively high penalty factors.
Altogether, the PSO and the operators, accomplish such handling
in a more efcient way, limiting the use of penalization. That is,
such mechanisms control that each continuous variable fullls the
generation limits, load and discrete variable limits for the pre- and
post-contingencies states. For each state the following reconstruc-
tion operators have been applied.
Please cite this article in press as: Y. Pablo O nate, et al., An optimal power ow plus transmission costs solution, Electr. Power Syst. Res.
(2009), doi:10.1016/j.epsr.2009.03.005
ARTICLE IN PRESS
G Model
EPSR-2851; No. of Pages 7
4 Y. Pablo Onate et al. / Electric Power Systems Research xxx (2009) xxxxxx
5.2. Control variables operating limits
This operator is dened as
P
Gi
=
P
Gi MAX
if P
G,i
> P
Gi MAX
P
Gi
if P
G,i MIN
< P
Gi
< P
G,i MAX
P
Gi MIN
if P
Gi MIN
< P
Gi
P
Gi MIN
0 otherwise
(21)
Theoperator, besides satisfyingtheactivepower generationlim-
its, is useful for avoiding the use of additional variables in order to
determine the units state; thus, for those units in which the active
power is below a predened percentage, the off state is chosen
(P
Gi
=0). Likewise,
P
k
Load,j
=
P
0
Load,j
if P
k
Load,j
> P
0
Load,j
P
0
Load,j
Load,j
if P
k
Load,j
< P
0
Load,j
Load,j
P
k
Load,j
otherwise
, k = 0, 1, . . . , K and j = 1, 2, . . . , ND (22)
ST
k
j
=
u
max
j
if ST
j
> u
max
j
u
min
j
if ST
j
< u
min
j
ST
j
otherwise
, k = 0, 1, . . . , K and j = 1, 2, . . . , ND (23)
where
Load,j
is the maximum allowed load interruption at bus-j.
5.3. Handling constraints for load balance
After the execution of the precedent operation, the total active
power assigned to the thermal units is not necessarily equal to
the demanded load. Thus, a re-dispatch is required for taking into
account the available units. Such re-dispatch must deal with con-
straints [34].
5.4. NewtonRaphson load ow
Once the reconstruction operators have been applied and
the control variables values are determined, for each particle a
load ow run is performed. Such ow run allows evaluating the
branches active power owandthe total losses, whichare assigned
to the slack bus.
5.5. Fitness function
To handle the active power ow limits and voltage constraints,
the objective function penalization is used. By this technique, the
tness function is composed by the objective function (6) plus
penalty terms for particles violating some power ow and/or volt-
age constraint. Such tness function can be expressed as follows:
F
fit
i
= F
obj
i
+Cte(iter)
N OFLW
j=1
OverFlow
j
+F
obj
i
NLoad
j=1
|V
ref
V|,
i = 1, . . . , NIND (24)
where, N OFLWrepresents the total number of lines with overow;
NIND is the number of individuals; OverFlow
j,t
is dened as in (25).
abs(FLOW
i,j MAX
FLOW
i,j
) (25)
FLOW
i,j MAX
, is the maximumallowable owacross the line con-
nectingbuses i j; Cte(iter) is adynamicallymodiedpenaltyvalue:
Cte(iter) = 300