PMU Based Centralized Adaptive Load Shedding

Scheme in Power System

H. BENTARZI
Signals and Systems Laboratory, IGEE, UMBB University, Boumerdes, Algeria
[email protected]

Abstract- One of major contingencies of an electric power load to be shed and its appropriate location [4-5].
system is due to imbalance between loads and generated electric The most important feature and advantage of this scheme
power. This leads to instability in the power system and hence to compared to conventional schemes is that bus voltages have
a blackout. Conventional Under frequency Load Shedding
an important role in the load shedding scheme. Also the
Schemes have been widely used, to restore power system
stability. However, the analysis of recent blackouts suggests that intelligent selection of loads to be shed is another important
voltage collapse and voltage related problems are also important feature of the proposed scheme. Additionally, the amount of
concerns in maintaining system stability The aim of this work is load to be shed is determined adaptively and it is proportional
to develop a new approach applied to a load shedding scheme to the magnitude of disturbance. By this way, the problems of
based on both parameters frequency and voltage that can be under-shedding and over-shedding, existing in the
provided by smart sensor named Phasor Measurement Unit. All
conventional schemes can be avoided.
generator buses frequencies and voltages are sent to center
where a magnitude of disturbance will be calculated and hence
the amount of load to be shed as well as its location will be II. LOAD SHEDDING SCHEME
determined. Under frequency load shedding scheme is the most
Keywords- Smart Grid, Centralized load shedding, Adaptive load commonly used control system to balance the generation and
shedding scheme, critical frequency, disturbance, Phasor load (power demand) and it is the last control step for
Measurement Unit and voltage sensitivity. preventing electric power system from blackouts. It deals
with shedding the appropriate amount of load for removing
I. INTRODUCTION
the overload situation. This may be performed in many steps
An electric power system is a large interconnected system
with each step having its own setting frequency and percent
that produces, transmits and distributes an electric energy to
of load to be shed. Three main types of load shedding
different consumers. Stability of the power system is of a
schemes may be distinguished: traditional (conventional),
great concern, since it is subjected to different disturbances
semi-adaptive and adaptive load shedding scheme [3]. In the
that may cause a local or complete system collapse if no
latter, the amount of load as well as the percent of load to be
adequate action is taken to prevent it. Therefore, many
shed in each step are selected adaptively according to the
techniques have been developed to make the power system
magnitude of the disturbance. It is determined using the initial
survives during disturbances and continue to operate. One
rate of frequency decline and is based on the System
common disturbance is the imbalance between generation and
Frequency Model (SFR) [1, 4]. From the reduced SFR model,
load due to an overload situation caused by generator outage
the relation between the frequency decline and the size of the
or loss of transmission lines. Generally, this situation has an
disturbance Pd is obtained as follows:
effect on frequency behavior of the system and hence the
df P
frequency decreases below the rated operating value. The d
, (Hz/s) (1)
dt 2 H
system spinning reserve can compensate small overload, sys

whereas large one requires rapid emergency control actions to

be taken by under frequency load shedding schemes that trip Where Pd is the disturbance magnitude in per unit,
temporary certain loads in order to balance the system and and Hsys is the inertia constant of the system, in seconds.
consequently recover the nominal operating frequency [1-3]. Load is typically shed by opening the circuit breakers that
Under frequency Load Shedding (UFLS) schemes have installed at the terminals of the feeders in the distribution
been widely used, to restore power system stability. However, substations. The architecture of load shedding schemes can
the analysis of recent blackouts suggests that voltage collapse be: local, distributed or centralized [5]. In this work,
and voltage related problems are also important concerns in centralized load shedding is used where each bus frequency
maintaining system stability [2]. For this reason, both information is sent to a central location where processed by a
frequency and voltage need to be taken into account in load computer program, which is used to perform the overall load
shedding schemes. In the proposed Centralized Adaptive shedding. Decision is then sent from the central station to the
Load Shedding Scheme algorithm, we consider both distribution substation to trip breakers as selected by the
parameters; voltage and frequency to determine the amount of computer control program.

150
 II. A PROPOSED ALGORITHM analysis is carried out in the following manner. The equation
In this work, a new approach which is applied to the for reactive power is:
centralized adaptive load shedding scheme is described. The Q = V V Y sin (5)
frequencies measured by Phasor Measurement Unit (PMU)
[6] are used for calculating the rate of change of frequency as Where: Qi: is the reactive power of bus (i),
well as the magnitude of the disturbance in the power system. Vi: is the voltage at bus (i),
The advantage of this approach, as compared to the Vj: is the voltage at bus (j),
conventional under-frequency load shedding scheme, is to Yij: is line admittance,
estimate the magnitude of overload occurring from different : is the voltage angle difference,
disturbances and accordingly to determine the necessary is The load angle at bus (i).
amount of load to be shed as well as its location. Therefore, it Differentiation of Eq.(5) with respect to Vi, gives,
avoids unnecessary shedding actions. dQ
= V Y sin (6)
Once the magnitude of the disturbance is determined, the dV
location and the amount of load to be shed at each bus have to Thus individually, for each bus the relation dQ/dV can be
be determined. This determination is based on the voltage written as:
sensitivity at each bus measured also by PMU. dQ1
= V Y sin (7)
2.1 Determination of the disturbance magnitude dV1
After measuring the frequency by PMU at each generator in
the power system, the rate of frequency decline of each dQ2
= V Y sin (8)
generator is determined in the center, and then the system dV2
mean frequency decline of the system is calculated according And so on till the nth equation:
to the following formula [7]: dQn
= V Y sin (9)
df dVn
n i
i 1 H i dt
df The Q-V curve shown in Fig.1 is obtained after plotting the
c , (Hz/s ) (2) dQ/dV for one such sample bus. The system is said to be
dt n unstable beyond the knee point which represents the critical
i 1 Hi situation. As the knee point is approached, the dQ/dV values
Where: dfc/dt is the rate of mean frequency decline, become smaller. Thus a system bordering on instability will
And dfi/dt is the rate of generator (i) frequency decline, have a small value of the slope at the knee point.
df and Hi is the inertia constant of generator (i).
c In Figure (1), the Y axis is the Mvar values of Q and the X
dt axis is the p.u. values of the bus voltage. The plot considers
Then, once the mean rate of frequency decline is known, two states of voltage stability represented by two points “A”
the size of the disturbance in the system may be determined and “B”. It can be noticed that, the point “B” is closer to the
using the following formula [8]: knee point as compared to point “A”. Thus, the dQ/dV of
H sys df c point “A” will be higher than the dQ/dV of point “B". Thus
Pd 2 , (pu) (3) more load needs to be shed from a bus with the dQ/dV value
f n dt of point “B”. Therefore:
Where fn is the nominal frequency of the system in hertz dQ 1
(50Hz), and Hsys is the equivalent inertia constant (in dV the load amount be shed from each bus
second) of the system given by the following formula [4]:
n
Or, the amount of load to be shed from each bus.
S Hi
i 1 i To estimate this load quantity, we consider the reciprocal of
H sys , (MVA) (4)
n
Si the voltage sensitivity as a fraction of the sum of all the
i 1
reciprocals of voltage sensitivities. The reciprocal is
Where: Si is the rated apparent power of generator (i), and considered because for a higher slope (i.e. a more stable
n is the total number of generators. case), the reciprocal will be smaller, hence a less amount of
2.2 Determination of the amount and location of the load to load will be shed from it. Now, the summation of the dV/dQ
be shed values of all the load buses is given by:
Once the magnitude of the disturbance is determined using
dVi dV1 dV2 dVn
the above equivalent swing equation (2). The location and the = + + (10)
amount of load to be shed at each bus have to be determined. dQi dQ1 dQ2 dQn
This determination is based on the voltage sensitivity at each The load that must be shed from each bus is a small
bus. Thus the bus with voltage sensitivity very close to the portion of the total load required to be shed in order to
instability limit will have a maximum load shed based on the recover the power balance. This portion of load at each bus is
reciprocal of its sensitivity as a fraction of the sum of the proportional to the fraction of dV/dQ value at each bus with
reciprocals of all the load bus sensitivities. Now the QV respect to the sum calculated above. This is represented as:

151
 Table I The generation unit parameters of the system
Generator
Gen. 1 Gen.2 Gen.3
Parameters
Volt Ampere (MVA) 247.5 192.0 128.0

Power factor (PF) 0.9 0.85 0.85

Inertia constant ( H) 9.55 3.33 2.35

Droop factor ( R) 0.06 0.05 0.08

High Pressure Power Fraction ( FH) 0.2 0.3 0.4

Reheat Time Constant (TR) 8 7 5

Damping Factor (D) 1 1 1

Mechanical Power Gain factor (Km) 0.95 0.9 0.85

dVi
dQi Figure (3): IEEE 9 Bus Power Network Model
for each bus (i) (11)
dV1 dV2 dVn
(dQ1 + dQ2 + dQn)
This is the fraction of the total voltage sensitivities. Thus,
the closer bus ‘i’ is to the knee point of the Q-V curve the
higher value of the above fraction will be. Hence, when this
fraction is multiplied by the total amount of load to be shed,
the load to be shed from each bus is obtained.

F
i

V bus
.Figure (1): Q-V analysis.
Therefore, the formula to be tested is:
dVi
dQi
Si = P (12)
dVi
dQi Figure (3): The flowchart of the proposed Algorith.
The table I gives the different parameters of the three
IV. SIMULATION RESULTS AND DISCUSSION generators.
The PSSE software has been used in the simulation for two
cases causing disturbances which are: Loss of transmission
The model of IEEE-9 bus power system that is shown in
line (3) and Generator G3 Outage with Increase of loads A B
Figure (2) is used to test and evaluate the proposed algorithm
and C by 50% (Over load). Then its results have been used
whose flow chart shown in Figure (3). The power flow from
for testing our algorithm programmed by Matlab.
one bus to another is indicated as well as the size of each load It can be noted that before the use of the load shedding
and the generating capacity. scheme, the load bus voltages decrease below their
predetermined standards and become stable at the lower
values as given in table II. The voltage sensitivities; dV/dQ
values are calculated individually for each load bus and the
results are also listed in the table II.

152
 The summation of all of the dV/dQ values is V. CONCLUSION
0.001880236.This value is used in the voltage sensitivity We have considered several cases here. In the first case
formula to calculate the load to be shed at each bus. The study, the disturbance was very small. The values of
obtained results are also given in table II. The total load to be frequency and voltage were in the safe margin and the
shed is 196.696 MW. After applying the load shedding disturbance can be removed by the spinning reserve with no
algorithm, the frequency generators and the load bus voltages need to perform any further load shedding. However, in the
plots are shown in Figure (4) and Figure (5) respectively. second case, the disturbance was very large; as a result both
The generator bus frequencies have been improved above frequency and voltage settle at low unacceptable values.
the specified lower limit. They settle at a value very close to The obtained simulation results for any disturbance sizes
50 HZ. The load bus voltages also show an improvement due are really satisfactory. It can be said that the load shedding
to the appropriate amount of shedding. based on the voltage sensitivities definitely improves the
frequency and voltage of the power system. The load
Table II: Voltage Sensitivities at each Load Bus (Case 2) shedding algorithm is activated at the threshold values;
Load Voltage dQ/dV dV/dQ Load to frequency of 49.7 Hz and voltage of 0.97 p.u for a 50 Hz
Bus (pu) be shed power system.
number (MW)
The average time required to start shedding load is 0.3
5 0.95477 1080 0.000926756 96.946
seconds. After load shedding has been applied an
6 0.97328 9691 0.000103185 10.8 improvement in the frequency and voltage profiles is
8 0.98175 1176 0.000850300 88.95 observed. In totality, it takes around 20 seconds for the
system to reach an acceptable value.
The advantage of this method is that the amount of load to
be shed is not large for all the disturbances unlike the
conventional one. Therefore, unnecessary shedding is avoided
which allows both a better service to different consumers and
the system collapse prevention.

REFERENCES
[1] H. Bentarzi, A. Ouadi, N. Ghout, F. Maamri and N. E.
Mastorakis, “A New Approach Applied to Adaptive
Centralized Load Shedding Scheme controllers”, in Proc.
8th WSEAS Int. Conf. on (CSECS ), Tenerife, Spain,
December 14-16, 2009.
[2] H.Seyedi and Sanaye-Pasand, “New Centralized Adaptive
Load-Shedding Algorithms to Mitigate Power System
Blackouts”, IEEE Paper, Published in IET Generation,
Transmission & Distribution, 2008.
[3] B. Delfino, S. Massucco, A. Morini, P. Scalera, and F.
Figure (4): Generator bus Frequency (After applying Load Silvesto, “Implementation and comparison of different
shedding). under frequency load shedding schemes”, Proc. IEEE
Power Engineering Society Summer meeting, Vol.1, 2001.
[4] P.M. Anderson, Power System Protection, IEEE press, New
York, 1990.
[5] P.M. Anderson and A.A. Fouad, Power System Control and
Stability, IEEE Press, Sec. Ed., New York, 2002.
[6] A. Ouadi, H. Bentarzi and J. C. Maun, “A new computer
based phasor measurement unit framework”, (IEEExplore),
SSD’09, Pp.1-6, Djerba, Tunisia, 2009, DOI:
10.1109/SSD.2009.4956773
[7] H. Seyedi, M. Sanaye-Pasand, M. R. Dadashzadeh,
“Design and simulation of an adaptive load shedding
algorithm using a real network”, IEEE Power India
Conference, April 10-12, 2006.
[8] V. Terzija, “Adaptive underfrequency load shedding based
on the magnitude of the disturbance estimation”, IEEE
Transactions on Power Systems, Vol.21, No. 3, Aug. 2006.

Figure (5): Load Bus Voltages (After Load Shedding).

153