Evaluation of Stability Index by Swarm

Optimization Techniques
1
Harish Kiran S, 1Subhransu Sekhar Dash, 1Subramani C,
1
Department of EEE, SRM University, Chennai, India.
[email protected]
Abstract: This paper highlights the use of optimization techniques to predict the voltage stability of the system using the
different voltage stability indices. The context of this paper is to develop a coding for the techniques used and to relate the
effectiveness of the index based on the optimization techniques. The reactive power of the line plays a vital role in identifying
to what extend the system is stable from collapse and also helps to find the weakest busses in the power system network. Based
on the results one FACTS controller will be connected to enhance the stability of the stability of the power system network.
The optimization techniques are coded with the newton raphson load flow analysis. The system considered for this study is 62
buses Indian utility system.

Keywords: Fast Voltage Stability Index (FVSI), Global Voltage Stability Index (GVSI), Ant Colony Optimization (ACO),
Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC) Algorithm, Newton Rapshon method (NR).

1. Introduction: techniques are explained and the way it is coded in

Power system voltage stability [1] is a big the matlab editor environment are deals in section 3.
problem with the resent development. For his The section 4 deals with the possible FACTS
stability problem there are many solution and controller that could be connected in the network.
research works are still in progress to find a better The coding result for the 62 bus Indian utility system
solution. The stability of the system can be evaluated is tabulated in section 5 with the conclusion and the
by the voltage profile of the line commonly known as further scope of analysis.
the Voltage Stability Analysis (VSA). The best way
to find the stability of the system is to find the index 2. Formulation of Stability Indices
of the system and its corresponding voltage at the Stability Index of the system gives the distance
bus. to the collapse point of the power system or the
Initially the stability of the system was found by maximum operating range of the power system. As
the use of PV curve and QV curve [2] by P. Kunder stated in the section 1 there are many ways to identify
et al. Then slowly many other method where used to the stability index of the system of which the Fast
find such as L-Index [3], Modal analysis[4], Line Voltage Stability Index (FVSI) and the Global
Stability Index (Lmn) [5], Line Stability Index (LQP) Voltage Stability Index (GVSI).
[6], Bus Power Index [7], Power Transfer Stability The formation of the indices are based on a
Index (PTSI) [8], New Voltage Stability Index simple two bus system and then developed to be used
(NVSI) [9], Fast Voltage Stability Index (FVSI) [10], in the actual system and not any other bus in the
Global Voltage Stability Index (GVSI) [11] and system. The two bus line diagram is shown in fig: 1.
more. Each index has its own merits and demerits.
In this paper we will use the Fast Voltage
Stability Index (FVSI) and the Global Voltage
Stability Index (GVSI). The effectiveness of the
indices is analyzed with the optimization techniques
such as Enhanced Genetic Algorithm (EGA) [12],
Big Bang-Big Crush Algorithm (BB-BC) [13] and
Artificial Bee Colony (ABC) algorithm [14].
The conventional load flow method Newton
Rapshon method is used to get the necessary data’s
for the indices which are given as the input data to
the optimization techniques. The optimization
techniques used will give the index of the system for
the corresponding loading condition. fig 1: Two bus modal
In the following section 2, the indices used are Where,
elaborated and derived. Different optimization
Vi, Vj: voltage at the sending and receiving end 4 Z 2Q j
buses. 1 (11)
Pi, Qi – active and reactive power at the sending end (Vi ) 2 ( R sin   X ij cos  )
bus. Since δ is normally very small we can assume δ≈0,
Pj, Qj – active and reactive power at the receiving end Rsin δ≈0, Xcos δ≈X, then the equation (11) can be
bus. rewritten as
δ is angular difference between the sending and the 4 Z ij2Q2
receiving end. FVSI ij  (12)
Req or R is the line resistance and Vi 2 X ij
Xeq or Xij is the line reactance. Where Z – line impedance
The system is said to be in stable operating Xij – line reactance
conduction when the indices obtained is between the Qj – reactive power at the receiving end
ranges 0.0000 to 0.9999, above which the system Vi – sending end voltage.
stability will collapses. The value indicated by the FVSI that is close to
1 indicates that the particular line is close to its
2.1 Fast Voltage Stability Index (FVSI): instability point. Hence the FVSI should be
The FVSI index is based on the line voltage and maintained less than or equal to 0.9999 in order to
the reactive power of the receiving end bus of the maintain the system stability.
system. The general current equation of the two bus
system shown in fig: 1 is 2.2 Global Voltage Stability Index (GVSI):
Vi 0  V j  For the formation of the GVSI is also similar to
I (1) the formation of the FVSI index. The same two bus
R  jX ij system is considered shown in fig: 1.
The apparent power at the receiving end bus 2 The apparent powers at the sending and
can be written as receiving end of the buses are
S j  Vj I * (2) Si  Pi  jQi (13)
Rearranging equation (2) we get; S j  Pj  jQ j (14)
*
S  The line current in the line from sending end to
I  j  (3) receiving end is as given below
 Vj 
 
P  jQ j I 2

P i
2
 jQi2 
line (15)
I j (4) Vi
2

V j  
The total real and reactive powers injected in the
Equating equation (1) and (4) sending end bus are
i j     V j 0   R  jX ij   Pj  jQ j  (5)
2
VV Pi  Req I line
2
 Pj (16)
The real part of the equation (5) is
Qi  X eq I line
2
 Qj (17)
i j cos   V  RPj  X ij Q j
2
VV j (6)
Substituting the current equation (15) in
And the imaginary of the equation (5) is equation (16) and (17) the following is obtained,
i j sin   X ij Pj  RQ j
VV (7)
Pi  Req
P i
2
 jQi2 
 Pj
Rearrange equation (7) with respect to Pj (18)
Vi 2
i j sin 
RQ j  VV
Pj 
X ij
(8)
Qi  X eq
P i
2
 jQi2 
 Qj (19)
Substituting equation (8) in (6) and rearrange it Vi 2
with respect to Vj;
Rearrange equation (18)
 R   R2 
V 
j
2
 X ij
sin   cos   VV   X  Q 0
 i j  ij X ij  j Qi  X eq
 P  P  Q
i j
(20)
    Req
j

(9)
And substitute in equation (20) in (19)
To get the real roots for V2 the discriminant
must be greater than or equal to “0”; B2 – 2AC ≥ 0 Req  2  X eq  Pi  Q j   Req Q j  
Pi   Pi      Pj
 R    R2  Vi 2   Req  
 sin   cos   i 
V  4  ij
X   Q j  0 (10)  
X
 ij    X ij  (21)
 Assume Vi as 1 in equation (21) and frame the The peso code for the PSO algorithm: Initially
equation with respect to Pi in quadratic form, run the load flow analysis for the test system and
Pi 2  Req2  X eq2   Pi  2 X eq2 Pj  2 X eq Req Q j  Req   form a matrix format with the necessary parameters
that required for the PSO algorithm.
X 2
eq Pj2  Req2 Q 2j  Req Pj  2 X eq Req Pj Q j   0 (22)
Step 1: Assignee the necessary parameters for the
The value of Pi must be real and positive. Hence
process such as no of lines, weight factor,
the coordinates B2-4AC must be greater than zero.
pBest, gBest, etc,.
 2X P  2X R Q  R   4 R  X 
2 2 2 2
eq j eq eq j eq eq eq Step 2: based on number of index N number of
random population are generated.
 X P  R Q  R P  2 X R P Q   0 (23)
2
eq j
2 2
eq
2
j eq j eq eq j j Step 3: For each individual line calculate the fitness.
On solving the above equation (23) we obtain Step 4: Based on the pBest compare it with the
the desired index as fitness for better solution if better swap the
value.
4  X eq Pj  Req Q j   X eq Q j  Req Pj   1 (24)
2

  Step 5: Set gBest as the best solution for that line.

Even from the above equation it is the same that Step 6: Repeat step 4 & step 5 for all the line.
is the index has to be less than ‘1’ in order to have Step 7: Return all the gBest as the best solution for
the system to be is more stable. The line with index each line.
more close to ‘1’ that line has more tendencies to be Step 8: End.
the weakest line of the system.
3.2 Artificial Bee Colony (ABC) Algorithm:
3. Optimization Techniques for VSA Karaboga[19] in 2005 proposed the Artificial
Optimization is a process of finding the maxima Bee Colony Algorithm[20 – 22] based on the honey
and the minimum of the function concerned. It was bee swarm behavior. In a normal bee colony the bees
first used by George Dantzig is 1940, initially for can be categorized into three such as the onlooker
military application and later it is has found its way bees, scout bees and the employer bees. Based on the
in many area. food sources the numbers of employer bee are
There are many category in optimization selected, that are the based on the number of lines the
techniques of which this paper uses 3 different index to be found. The onlooker bees look out for the
algorithms from a particular category known as the patron of the movement of the employee bee to find
“Swarm Intelligence” category and the algorithms are the best food source, i.e., the best index value of each
“Particle swarm Optimization Algorithm”, “Ant line is found by the employee bee and identified by
Colony Optimization Algorithm” and “Artificial Bee the onlooker bee based on the movement patron.
Colony Algorithm”. When the employee bee completes its tasks the
entire employee bee in that food sources will become
a scout bee and moves in again in that direction to
3.1 Particle Swarm Optimization Algorithm:
find a new food source, otherwise when the best
PSO algorithm was attributed by Dr. Kennedy,
index solution for the line is obtained than the
Dr. Eberhart and Dr. Shi [15] intended based on the
employee bee converts it-self into a scout bee and
social behavior simulation. It has a simplified and
moves into the next line to find the possible best
optimization performing, in the book written by
index in it. The fitness function for the ABC
Kennedy and Eberhart [16 - 18] the algorithm is
algorithm is the same as the PSO algorithm given in
groped in the swarm intelligence. This algorithm
equation (25) & (26).
finds its application in various places where the local
The peso code for the ABC algorithm is given in
methods fails to optimize such as in places where it is
the following steps.
not incessant and inflexible to derive.
Step 1: Run the load flow analysis and form matrix
But the PSO requires only the evaluated solution
format tabulation for the necessary
by the fitness function based on the parameters to be
parameters required.
optimized. The values are restricted to its maximum
Step 2: The best solution to be identified in each line
and minimum values, of its current positions. Each
is defined as 0.9999.
particles move randomly to the possible best position
Step 3: The number of food sources is initialized as
of all or to its own best position based on the fitness
the number of line.
functions. The fitness functions for this problem are
Step 4: The employee bees sent to find the best
identified as
possible solution in the first line or the first
Fij=max (FVSIij) (25)
food sources.
Fij=max (GVSIij) (26)
Step 5: Calculate the fitness function for the
corresponding value.
Step 6: Check for the best solution is either fitness In this paper we will be using two controllers
function value or the initialized value. one in series is Thyristor Controlled Series
Step 7: Repeat step 4 to step 6 until the Compensator (TCSC) and the other controller is
fitness function value becomes the best combination of both shunt & series is Unified Power
solution. Flow Controller (UPFC).
Step 8: Check for the maximum number of line The selection of these controllers is based on the
reached else increase line count by 1 and go steady state and the dynamic state stability of the
to step 4. controllers. For the possible problem such as voltage
Step 9: Return all the line fitness function value. limits, loop flow thermal limits, short circuit level,
Step 10: End. transient stability, damping, post contingency and
voltage stability. By the use these controllers a better
3.3 Ant Colony Optimization (ACO): solution can be obtained. The other advantage of
Marco Dorrigo in 1992 [23] proposed the Ant these controllers is that they have the ability to
Colony Optimization (ACO) algorithm [24-25] to improve the power factor Extra High Voltage (EHV)
find the shortest path by graphically, on the bases of Transmission line.
an ant finding its path between the food sources and The description of the controller is available in
its colony. varies sources, so only the rating of the controller is
The algorithm is initialized as 1 for the trail described. As per the Siemens AG database [26] the
intensities. Then the calculation of probability for the rating of the controller are found to be,
first ant to find a short path is performed. Based on RTCSC  rf *0.45  0.25 (27)
the disturbance a value is calculated as the first ant RUPFC  rf *180 (28)
variable. When the first ants of all the variables are
initialized, then the counter is increased by 1. Where
The assignment continues with the second ant rf- is the rating factor of the controller.
and the process is repeated until the whole colony is
created. After the first round the colony will have The rating of the controller connected in the
sufficient data to form for the round and update all system is taken as 1p.u, in 1 st case and -1p.u, in the
the values of the ant periodically based on the path 2nd case the results are tabulated based on the index.
modified. The algorithm check in each ant for the
goal reached at the end of the completion of the first 5. Test results and Discussions
cycle. This process is repeated until the best solution 5.1 Result with optimization techniques with
is obtained by all the ants in all the branches or food FVSI index:
sources. All the load busses in the taken system are
loaded uniformly until the critical index of 0.9999 is
4. FACTS Controller: obtained. The loading value for FVSI index is found
There are many types of Flexible AC to be 867.16% of the nominal load value in the NR
Transmission System (FACTS) controllers classified method. For the corresponding loading the index for
based on who it is connected to the transmission line the optimization techniques used are tabulated.
such as shunt, series or both combined.

FVSI
Without FACTS With UPFC With TCSC
INDEX 
Line No 88 89 30 88 89 30 88 89 30
Bus From 61 62 20 61 62 20 61 62 20
Bus To 62 25 23 62 25 23 62 25 23
NR 0.9999 0.9015 0.5742 0.4572 0.2000 0.3285 0.6726 0.4653 0.4317
PSO 0.9999 0.9103 0.5612 0.4492 0.1859 0.3162 0.6685 0.4684 0.4402
ABC 0.9999 0.8935 0.5698 0.4629 0.2102 0.3302 0.6702 0.4607 0.4385
ACO 0.9999 0.9068 0.5712 0.4562 0.2035 0.3254 0.6795 0.4702 0.4286
Table 1: FVSI index for with and without FACTS controller

The table 1 shows the FVSI indices obtained for bus 20 to 23. Based on the analysis the week bus is
with and without FACTS controller. The first most found to be as bus no 62. Hence, the FACTS
critical line is the line no 88 connecting bus 61 to controller is connected in bus no 62 with line no 89.
62.The second critical line is line no 89 connecting The indices when the FACTS controller such as
bus 62 to 25 and line no 30 is the third connecting
UPFC and TCSC are connected to the bus no 62 are improved as well as the system loadability of the
tabulated in table 1. system is improved.

5.2 Result with optimization techniques with

GVSI index:

The same process as of with FVSI index is done with

the GVSI index and its maximum loading for NR
method is found to be 848.5% the nominal load
value. The indices obtained for with and without
FACTS controller are tabulated in table 2.

Fig 2: Voltage Profile of weak bus by FVSI index

Based on the voltage profile the busses with a

voltage margin less than 0.9500p.u, are bus no 62,
20, 11, 12 and 12. The Figure 2 shows the voltage
margin obtained for NR method with UPFC and
TCSC as controller.
The bus no 62 is ranked first with voltage of
0.7506p.u, hence the bus no 62 is the best place for
the FACTS controller to be connected. The voltage
profile of the weak bus after connecting the FACTS
controller is also shown in Figure 2. When UPFC is
connected the voltage at bus no 62 is found to be
0.9367p.u, and for TCSC as 0.9783p.u, this proves
that the voltage stability of the system has been

GVSI
Without FACTS With UPFC With TCSC
INDEX
Line No 29 88 89 29 88 89 29 88 89
Bus From 17 61 62 17 61 62 17 61 62
Bus To 21 62 25 21 62 25 21 62 25
NR 0.9999 0.8379 0.4795 0.8975 0.8283 0.4775 0.4039 0.8379 0.4795
PSO 0.9999 0.8275 0.4775 0.8805 0.8135 0.4575 0.4105 0.8248 0.4682
ABC 0.9999 0.8325 0.4732 0.9051 0.8227 0.4701 0.3951 0.8329 0.4724
ACO 0.9999 0.8297 0.4805 0.8924 0.8560 0.4757 0.4106 0.8395 0.4802
Table 2: GVSI index for with and without FACTS controller

From table 3 the top most critical line is found to

be as line no 29 bus connected from bus 17 to 21.
The second critical line is line no 88 connected from
bus 61 to 62 and the third critical line is line no
number 89 connected from bus 62 to 25. But for the
placement of the FACTS controller the voltage
profile has to be considered as the critical bus no 62
comes in the second and third line.

Figure 3: Voltage Profile of weak busby GVSI index

Hence by the voltage profile value for the

corresponding load value are tabulated for the busses
with a voltage margin less than 0.9500p.u, in Figure
3. From the table the weakest bus is found to be bus
no 62 by NR method and also the FACTS controller
is connected at bus no 62 through line no 89.
The indices obtained after connection of the
FACTS controller are tabulated in table 2. The
indices value after connecting the controller has
reduced to some extent which shows the system is in
stable state with a wider loading rang.

6. Conclusion:
By both the indices the weak bus identified are
the same and also a similar results were obtained
when optimization techniques are used, this show the
effectiveness of the index to identify the weak bus.
Even the voltage profiles of the proposed system
shows a marginal increase when the FACTS
controller is connected.
To extent the research work with the real time
practical system and data to identify the stability state
of the system using artificial intelligence techniques
by using the indices.

References:
[1]