Professional Documents
Culture Documents
I. INTRODUCTION
when i ≠ j: where:
subscripts G, L : generator and load parameters
𝜕𝑃𝑖,𝑢𝑝𝑓𝑐 respectively.
𝐻𝑖𝑗 𝑢𝑝𝑓𝑐 = (11) subscripts Cal : calculated parameters by load flow
𝜕𝜃𝑗 program .
𝑢𝑝𝑓𝑐 𝜕𝑃𝑖,𝑢𝑝𝑓𝑐
𝑁𝑖𝑗 = 𝑉𝑗 (12)
𝜕𝑉𝑗
𝜕𝑄 𝑖,𝑢𝑝𝑓𝑐 III. RANDOM SEARCHING APPROACH (RSA)
𝐽𝑖𝑗 𝑢𝑝𝑓𝑐 = (13)
𝜕𝜃𝑗
𝜕𝑄𝑖,𝑢𝑝𝑓𝑐 RSA is a contribution method to ACO algorithm, to
𝐿𝑖𝑗 𝑢𝑝𝑓𝑐 = 𝑉𝑗 (14) ensure that ACO program will converge to optimal
𝜕𝑉𝑗
solution at very short time. Without RSA, ACO may
requires more than 30 seconds computation time to find
the most optimal UPFC parameters. This method is
For bus j, when j = i: performed by running load flow program consisting of
UPFC repeatedly, i.e. the UPFC voltage constant and
𝜕𝑃𝑗 ,𝑢𝑝𝑓𝑐
𝐻𝑗𝑗 𝑢𝑝𝑓𝑐 = (15) angle are selected step-by-step within the range of (0.01<
𝜕𝜃𝑗 ε < 0.1) and (0° < γ < 360°), and then send to power
𝑢𝑝𝑓𝑐 𝜕𝑃𝑗 ,𝑢𝑝𝑓𝑐 system model . The UPFC parameters within the ranges
𝑁𝑗𝑗 = 𝑉𝑗 (16)
𝜕𝑉𝑗 that result to the lowest losses and FVSI will be converted
𝜕𝑄 𝑗 ,𝑢𝑝𝑓𝑐
to narrow ranges, which then will be used in ACO engine.
𝐽𝑗𝑗 𝑢𝑝𝑓𝑐 = (17)
𝜕𝜃𝑗
𝜕𝑄𝑗 ,𝑢𝑝𝑓𝑐 IV. ANT COLONY OPTIMIZATION (ACO)
𝐿𝑗𝑗 𝑢𝑝𝑓𝑐 = 𝑉𝑗 (18)
𝜕𝑉𝑗 Ant Colony Optimization (ACO) was introduced by
Marco Dorigo in his Ph. D thesis in 1992 [10,11], and
when j ≠ i: was developed it in his further work with his colleagues,
as summarized in [12]. ACO algorithm is inspired by the
𝜕𝑃𝑗 ,𝑢𝑝𝑓𝑐 behavior of real ant colonies used to solve combinatorial
𝐻𝑗𝑖 𝑢𝑝𝑓𝑐 = (19) optimization problem. In this study, there are some
𝜕𝜃𝑖
modifications performed on the ACO algorithm in order
to make it suitable for the application in UPFC. The β : parameter, which determines the relative
algorithm was modified to solve the continuous importance of pheromone versus distance
optimization problems instead of graphical optimization (β > 0)
problems in its original philosophy [13]. The algorithm ρ : heuristically defined coefficient (0 < ρ < 1)
for UPFC optimization using ACO has been described in α : pheromone decay parameter (0 < α < 1)
Fig. 2. The process involves initialization, state transition q0 : parameter of the algorithm (0 < q0 < 1)
rule, local updating rule, fitness evaluation and global τo : initial pheromone level
updating rule [11,14].
Every parameter requires to be set for limiting the search
range in order to avoid large computation time.
n 1
d max max di (27)
i 1
di r max u (28)
where:
r : current node
u : unvisited node
di : distance between two nodes
arg max u J k ( r ) (r , u ). (r , u ) }, if q q0 (exp loitation )
s
S , otherwise (biased exp loration )
(29)
where:
q : random number uniformly distributed in
[0…1]
S : random variable selected according to the
probability distribution given in eq. (29)
[ (r , s)].[ (r , s) ]
, if s J k ( r )
Pk (r , s) u J k ( r ) [ (r , u )].[ (r , u ) ]
0, otherwise
(30)
where:
Fig. 2 ACO flowchart algorithm τ : pheromone
Jk(r) : set of nodes that remain to be visited by ant
Step 1: Initialization; during the initialization process n, k positioned on node (to make the solution
m, tmax, dmax, β, ρ, α and q0 are specified. feasible)
η : 1/δ, is the inverse of the distance δ(r,s).
where:
n : no. of nodes
m : no. of ants
tmax : maximum iteration
dmax : maximum distance for every ants tour
Step 4: Apply local updating rule; while constructing a The pheromone level is updated by applying the global
solution of UPFC optimization, ants visit edges and updating rule in eq. (33).
change their pheromone level by applying the local
updating rule of eq. (31). τ(r,s) ‹— (1– α) τ(r,s)+ α.Δ τ(r,s) (33)
Δ τ(r,s) = τo (r , s) gb
0, otherwise
Step 5: Determine tuned parameters; two variables (x1, x2)
required to represent the UPFC parameters (i.e. UPFC Lgb : the length of the globally best tour from
voltage constant, 𝜀, and angle, 𝛾) and are selected within the beginning of the trial
the specified ranges from RSA method.
Step 7: End condition; the algorithms stop the iteration
Step 6: Fitness evaluation; it is performed after all ants when a maximum number of iterations have been
have completed their tours. In this step, the control performed otherwise, repeat step 3. Every tour that was
variable is computed using the following equation:- visited by ants should be evaluated. If a better path is
d discovered in the process, it will be kept for next
x x max (32)
reference. The best path selected between all iterations
d max
where: engages the optimal scheduling solution to UPFC tuning
d : distance for every ants tour problem. The overall steps of the ACO algorithm can be
xmax : maximum x represented in the flow chart of Fig. 2.
2 14
13
4 C
15
30
12 16
18 23
20
6
10 17 19 24
5 7
C
21
29
22
27
25
11
9 28
C
26
and 27 to 30 are compulsory. Fast Voltage Stability Index for Bus 26 and on location 27 to 30 for Bus 30, and with
(FVSI) which was developed by Associate Professor Dr. combination of three UPFCs. It is shown that effects of
Ismail Musirin [15] was used as the indicator for power UPFC combination in power system gives further
system stability. To avoid system collapse, FVSI must be optimization that cannot be achieved by individual UPFC
lower than unity, or should be minimized as low as on location 25 to 26 and 27 to 30. For e.g. looking
possible [15]. A method called as Maximum Loadability towards load of 30.3 MVAR from Table 1, in spite of
Identification (MLI) [13] was first conducted to find the individual UPFC on line 25 to 26 resulted to reduced
most sensitive bus (the test busses for ACO program), i.e. FVSI value, the UPFC combination gives better than that,
the bus that experienced convergence failure of load flow i.e. 0.6947 is further reduced to 0.6349, which is about 9
program at the lowest load MVAR value. From the test % further reduction. Thus, system stability that was
results, the most sensitive busses were 26 and 30, which improved before is further improved by combinational
diverge at load MVAR of 30.3 and 33.3 respectively. effect. Furthermore, both of the system’s Total Losses
After performing MLI, RSA was performed to find the and Voltage Profile are well improved as well as FVSI
most narrow ranges of UPFC parameters (𝜀, 𝛾), which value, for e.g. at 30.3 MVAR, reduced losses of 23.7523
then will be used by ACO engine to perform optimization MW and improved voltage of 0.7319 p.u. by individual
process. For e.g. if ε = 0.07 and γ =150° result to the UPFC on location 25 to 26 are further improved to
lowest losses and FVSI at MVAR of 15, the best narrow 21.8423 MW and 0.7721 p.u. respectively by UPFC
ranges for the usage of ACO are (0.065< ε < 0.075) and combination.
The percentages of further improvement for losses and same time, system’s Total Losses and Voltage Profile
voltage magnitude are approximately 8 % and 6 % have been optimized from individually optimized values
respectively. (values optimized by individual UPFC on location 27 to
Fig. 4 illustrates the graphical representation of FVSI 30), i.e. from 26.3244 MW and 0.629 p.u. to 24.074 MW
values in accordance with variation of load MVARs. By (9 % further reduction) and 0.6739 p.u. (7 % further
inspection, the UPFC combination results to the most improvement) respectively.
reduction of FVSI values, which means that system
stability has been optimally improved beyond the Fig. 6 shows graphical view of FVSI variation in
capability of individual UPFC on location 25 to 26. accordance with load MVAR changes. From that
Moreover, UPFC combination increases the gap margin diagram, it is observed that for all values of load MVAR
(losses margin) between without and with UPFC system, levels, there exists further improvement in terms of power
i.e. the combinational involvement of UPFC provides system stability, i.e. the FVSI values has been extensively
further reduction in terms of Total Losses, and hence reduced from its individually optimized values.
simultaneously further increase the quality of power Moreover, the Total Losses also being improved beyond
transmission, as depicted in Fig. 5. the capability of individual UPFC on location 27 to 30 at
Similar observation can be deducted from Table 2, all levels of load MVARs, as shown in Fig. 7.
which represents the outputs for Bus 30. UPFC
combination still capable of producing further improved
results in terms of FVSI, Total Losses, and Voltage
Profile. Considering load of 33.3 MVAR, the FVSI that
had been reduced by individual UPFC on location 27 to
30 before is now further reduced by UPFC combination
from 0.8128 to 0.7542 (further reduction of 7 %). At the
Fig. 4 : FVSI comparison with bus 26 loaded Fig. 7 : Total Losses comparison with bus 30 loaded
VII. CONCLUSION
VIII. REFERENCES [10] Dorigo, M. (1992). " “Optimization, Learning and Natural
Algorithms”." Ph.D Thesis, Dipartimento, Politecnico di
[1] N.Tambey.Prof M.L.Kothari, “Unified Power Flow Milano, Italy.
Controller (UPFC) Based Damping Controllers for Damping [11] M.R.Kalil, I. M. M., M.M Othman, T.K.A.Rahman (2009).
Low Frequency Oscillations in a Power System” "Loss Minimization by Optimal Reactive Power Dispatch
[2] Y.Ye,M. Kazerani, Power Flow Control Schemes for Series- Using Ant Colony Optimization (ACO) Technique."
Connected FACTS Controllers, Electric Power System International Journal of Power, Energy and Artificial
Research 76 (2006) 824-831 Intelligence.
[3] Nagaraju, S. V. R. K. a. S. S. (2007). ""Loss Minimization [12] Caro, M. D. a. G. D. (1997). "“The Ant Colony Optimization
by Incorporation of UPFC in Load Flow Studies"." Meta-Heuristic: in D. Corne, M. Dorigo, F. Glover, editors,
International Journal of Electrical and Power Engineering 1 New Ideas in Optimization"." McGraw-Hill: pp. 11 - 32.
(3): pp. 321 - 327. [13] Khalil, M. R. (2008). "Ant Colony Optimization (ACO)
[4] S.E.Mubeen, R.K.Nema,G.Agnihotri, "Power FLow Control Technique for Reactive Power Planning in Power System
with UPFC in Power Transmission System", World Stability Assessment" MSc. Thesis, Mara University of
Academy of Science, Engineering & Technology 47 2008 Technology.
[5] G.Radman, R.S.Raje, "Power Flow Model/Calculation for [14] ALLAOUA, B. G. a. B. (2009). "Ant Colony Optimization
Power System with Multiple FACTS Controllers", Electric Applied on Combinatorial Problem for Optimal Power Flow
Power System Research 77 (2007) 1521 - 1531 Solution." Leonardo Journal of Sciences: pp. 1 - 17.
[6] Vural, M. T. a. A. M. (2004). ""Analysis and Modeling of [15] M.R.Kalil, I. M. M., M.M Othman (2009). "Optimal
Unified Power Flow Controller : Modification of Newton- Transformer Tap Changer Setting for Voltage Stability
Raphson Algorithm and User-Defined Modeling Approach Improvement " International Journal of Power, Energy and
for Power Flow Studies"." The Arabian Journal for Science Artificial Intelligence, No. 1, Vol. 2: pp. 89 - 95.
and Engineering Volume 29(Number 2B): pp. 136 - 14