Congestion Management in a Deregulated Environment

Using Corrective Measures

M. Doll, Member, IEEE J. F. Verstege, Senior Member, IEEE

University of Wuppertal University of Wuppertal
42o97Wuppertal,GerrnmY 42097Wuppertal,Germany

Abstract: This paper presents a Congestion Management System The functions of the CMS are the detection of multiple con-
(CMS) based on corrective measures. In an unbundled power system gested conditions and the suggestion of corrective measures.
the schedules of the generation units and of the demands are given to Multiple congestion is some of the followings:
thetransmission system operator whohasto realize them. This may
cause multiple congested conditions, which can be eliminated by the a) The maximum flow limit on all lines.
following corrective measures: Adjustment of the network elements b) The upper and lower voltage limit at every node.
like transformer taps or shunt elements, corrective switching, correc- c) a) and b) for the results of the (n-1) contingency analysis.
tive redispatch of generation. Aim of the CMS is to speci~ a state of d) The breaking power at the circuit breakers as a result of the
the transmission system, which represents an undisturbed state and
short-circuit calculation.
enables all schedules. The search for a state of the transmission sys-
tem, which fulfills these requirements, represents a mixed-integer- To observe these requirements of the security analysis the
nonlinear-problem, which cannot be described in an analytical way. transmission system operator search for a system state, which
Particularly the topology changes generate a problem of large
fulfills these requirements. The following measures may avoid
dimension. The paper presents the mathematical model representing a
congestion:
CMS to formulate an optimization problem and gives exemplary
study results. The described problem is solved by an Evolution Corrective variation of the transmission system control
Strategy. options (e.g.: voltage level, complex transformation ratios
Keywords: Congestion management, power system security, power or settings of the shunt elements).
system operation, corrective switching, Evolution Strategy, Corrective switching [4,7,9] of network elements.
liberalized energy markets. To provide ancillary services the transmission system
operator has concluded contracts with generation
companies. In the margin of these contracts congestion can
I. INTRODUCTION
be avoided.
If there is no possibility to avoid the congestion the redis-
The deregulation of the electric utility industry causes a fun-
patch of generation maybe permitted.
damental reorganization. Many different structures are arisen
in different regions. Typical is the splitting in the competitive The detection of multiple congested conditions and the
generation market and the transmission system as a regulated search for the considering corrective measure are tasks for an
monopoly. This unbundling of the vertical integrated CMS, which gives proposals to the transmission system
companies generates a new role for the transmission system operator. Because of the discrete topology changes the CMS is
operator. The role depends on the market structure and the a mixed-integer-nonlinear-problem which cannot be described
chosen market rules and gives new tasks to the operator. One in an analytical way.
of these functions is the Transmission Dispatch and Con- This paper presents a new approach based on an Evolution
gestion Management (TDCM) [8]. Strategy (ES) that solves the described problem.
One method to prevent congestion are price signals like
marginal costs [1,2,3]. This method can only be used if the
II. TERMINOLOGY
congestion is permanent and already known. Then the conges-
tion management is a task of power system planning.
To formulate the problem and its solution mathematically,
But how to handle with temporary and unexpected conges-
the following notation is introduced:
tion which arises during operation? In a meshed transmission
system there is no possibility to specify just one originator, a, A index, quantity of contingency cases
who has caused the congestion. So the transmission system b, B index, quantity of circuit breaker
operator has to use all options of the power grid to avoid the
e index of shunt element
congestion. This is a difficulty work, which can be done by a
g, G index, quantity of generation units
Congestion Management System (CMS).
The presented approach of the CMS is independent of the 1,L index, quantity of lines
market structure and the market rules. The assumption is, that nom index of nominal magnitade
all schedules of generations and demands are known by the n, N index, quantity of nodes
transmission system operator. The required information are
r index of transformer
provided by forecasting andlor announcement of the
x index of decision vector elements
schedules.

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 393
 r contingency analysis requirement the current requirement Ii’ at line 1,
Ii’ current requirement
Y short-circuit calculation requirement I + sgn(Il – Il,ma ) ~ (11 – Ii,max )
17[ = (2)
(9 voltage requirement 2 11.max
I current
AS the short-circuit calculation requirement Y at circuit breaker b
P contracted active power for ancillary services
AS
Q contracted reactive power for ancillary service I + Sgn(S~C,b – s~c,b,mu )
Yb =
PSD 2
scheduled active power of the generation (3)
GN * (S;c,b – ‘;c,b,max )
P active power of generation
~GN ‘;c,b,max
redispatch of active power
APLS changing of the active power losses and the contingency analysis requirement r for contingency
Z@ assessed redispatch of active power case a
c coefficient
ra . Zf=l ‘a,. + Zf=l ‘aJl .
(4)
v voltage
v REF
voltage at reference node Equation (4) is the sum of all voltage and current
S’rse initial symmetrical short-circuit power requirements in the contingency case a. It is not necessary to
~ IR
study all contingency cases A to specifi the security of the
position of the in-phase regulator taps power system [6]. It is sufficient to check the essential
~QR position of the quadrature regulator taps contingency cases.
~ST
position of the shunt element taps
B. Decision variables and constraints
SLN
position of the switch at lines
SBC The task of the CMS is to find a network configuration that
position of the switch at bus couplers
~ GN fulfills the requirements (l), (2), (3) and (4). The state space,
position of the switch at generation units
where a solution can be found is defined by the continuous
SLD
position of the switch at loads vector i
SSE
position of the switch at shunt elements
objective function “AS
o P
2 vector of discrete decision variables “AS
Q

;
i vector of continuous decision variables
; IR
i discrete step size vector
; QR (5)
6 continuous step size vector

‘
ndf normal distribution function ST
rand random function
REF
sgn sign function v
- GN
AP

III. PROBLEM FORMULATION

and the discrete vector d
A. Requirements

The main task of the CMS is to determine a secure power

system state without congestion. This means that all operation (6)
limits are strictly adhered. These requirements can be written
as the follows:

The voltage requirement @at node n,

The vectors ~As -‘s
and Q describe the active and reactive

1+ sgn(Vn – Vn,~OI) * (v. – ‘n,mox )

power, which is contracted between the transmission system
@. = operator and the generation companies to provide the ancillary
z YB,~*M 14Y
(1) services. These contracts include restrictions like the follows:
I + ~gn(v’a,mjn – V’n) * (V’n,mjn– Vn )
+
2 v n,nom ~AS, ~ PAS ~ PAS (7)
g,mm g– g,max

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 394
0-7803-6674-3/00/$10.00 (C) 2000 IEEE
 The above described corrective measures to avoid
congestion just only change the configuration of the
The taps of the transformers and shunt elements have con- transmission system. Due to the possible manipulation of the
stmctional boundaries, which should be observed. competitive generation market, it could be very problematic,
to modify the generations. But if all the declared measures do
#R(QRST) ~ ~IR(QRST) ~ ~IR(QR,ST) not suffice to avoid the congestion the transmission system
r(r,e),min — r(r,e) r(r,e),max
(9]
operator can redispatch the generation to get a secure system.
The constraints of the voltage at the reference node are the In this case the constraints of the generation units and the
upper and lower voltage limits. balance of the whole system is to consider.

V:,;F < VREF < V;:F (10] (14)

Corrective switching is a measure to avoid congestion, (15)

which is used in practice [10]. Purpose of the switching is to
reduce branch overloads, solving voltage problems and to
The redispatch of generation takes place in minimum wise
reduce short-circuit currents [5]. That can be done in different
and only if necessary. The grid operator has to pay for the
manner. The discrete vector ~ describes the network topol- increase of the generation; also the operator pays or gets
ogy. A typical representation [5] of the topology is shown in money (it depends on the market system) for the reduction of
Fig. 1. generation. This assesses the redispatch of the generation, so
that can be formulated as

1[
1 + sgn(AP~N )
q . Zjpy * Cg,pos
2
(16)
sgn(@gGN ) – 1
+ Cg,neg
2
1

The sum of the assessed redispatch of all generations must be

minimized.

2 activepower
.g
Fig. 1.Representationof the networktopology

%P.J
In a) the discrete topological variable s~# represents the state
of a switch at a line (resp. at a transformer) which connects a
line 1 with a busbar n. The lines or transformers can switch on
or off and a changing over of lines to another bus is possible. ➤ PgGN
I I I
That can also be modeled with “pseudoparallel” lines [5] PgG:iH Pg’’:in Pg’DPgA:ax PgG:a
which is shown in b). In both cases a coupling of the bus bars
by a line or transformer is to beware. So we have to formulate reactive power
for every line 1
A&
(11)

That means that the line is switched off or on. In the last case
the line can only be connected with one node. The constraints
of the switches at generations Sn
‘N and loads s~D are similar.

(12)
I
But compared with the switches at lines, they cannot put off, Fig. 2. Generation unit which provides ancillary services
because a break of the load or generation is not permitted.
Shunt elements and bus couplers can put off, so we have to
Because of the assessed redispatch(16) and the constraints (7),
formulate
(8), (14) the generation units which have concluded a contract
to the transmission system operator are defined as shown in
(13)

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 395
 Fig. 2. There is no assessed redispatch of the reactive power.
The transmission system operator can induce values between
the minimal and maximal limit of reactive power, which is
similar with the limits of the reactive power of the ancillary
services,

C. Objective function

Aim of the congestion management system is to avoid

congestion and to suggest corrective measures to reach a se-
cure system state. That means the requirements (1), (2), (3)
and (4) should be fulfilled. If it is not possible by the control-
lable elements of the transmission system a minimized
redispatch of the generation is permitted (16). The objective
function can be written as follows:

o= The approach iterates till one of the following criteria is

fulfilled:
+ a) The objective function value is equal zero. That means the
congestion is cancelled and corrective measures are
+
Qi’k:=x.) suggested. A corrective redispatch of the generation is not
necessary.
This objective fimction must be minimized. If the value of b) The objective function value of the best individual of each
the objective function is zero, no requirement is violated and generation converges or the maximal number of generations
no redispatch is necessary. In the case that c,.~ is significant is reached. In this case the congestion is cancelled and
greater than CM and O is not equal zero, the result of the mini- corrective measures inclusive the minimal necessary
mized objective function is a state without congestion and corrective redispatch of generation are suggested.
with a minimal mandatory redispatch.
B. Discrete decision variables

IV. SOLUTION TECHNIQUE The handling with discrete decision variables is one prob-
lem of Evolution Strategies. Due to that a special modeling of
A. Base algorithm the network topology is necessary. Differing from the
modeling shown in Fig. 1 a new representation of the topology
The proposed approach is based on an Evolution Strategy. is chosen. The known decision variables s are elements of O
The Evolution Strategies are suitable to solve high and 1 and the constraints (11,12,13) should be fidtilled. In Fig.
dimensional non-linear problems and they have no requisition 4 the decision variables are represented by a rotary selector.
on an analytical description of the problem. The typically ES
occupy only with continuos decision variables. To solve the
described problem of the CMS a specially adapted ES is
developed. The base algorithm is given in Fig. 3. The
Evolution Strategy starts from an initial set of p parents. The
assessing of this start generation is based on the objective
fimction (17) which includes the known approaches of the
security analysis: Load flow calculation (LF), contingency
switch at lines, bus switch at generations
analysis (CA) and short-circuit calculation (SC). The best
couplers or shunt elements or loads
individuals of this population are selected; these are the
parents of the next generation. If one parent with a value of the
Fig. 4. Representation of network topology with rotary selectors
objective function equal zero exists, the approach is finished
because a state without congestion is found. In the case of
congestion a new generation of L children is generated by Because of that the discrete decision variables are no longer
recombination. In the next step the decision variables are only elements of O and 1.
slightly modified by mutation. After that every child is
assessed by the named approaches of the security analysis. On ‘1
‘N, s:crs~ e {O,..., N1} (18)

the basis of the assessing the L best children are selected. They
are the new parents of the next generation. sl‘N, s;D e {O,...N1 –1} (19)

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 396
 By this representation fewer discrete decision variables exist V. CASE STUDY
(in a three-bus-station just one decision variable instead of
three). Furthermore this kind of modeling observe the Within the scope of a case study, the presented approach
constraints (11, 12, 13) implicitly. has successfidly been adapted to several various power
systems. To demonstrate that the proposed Evolution Strategy
C. Discrete recombination and mutation is able to handle the complex task of congestion management
a simply structured test system is chosen. The studied test
Two essential parts of the Evolution Strategy are system is a EHV power system (380-kV) with two substations,
recombination and mutation of the children. These parts are each with two bus bars. (Fig. 5).
straightened on continuous decision variables. To recombine
and mutate the discrete vector ; a special strategy is
developed which is adapted on the typical features of the
network topology. Just slightly changes of the network GN1
topology can generate a totally new system state. As a result of [F ~
this only one parent of the p parents is randomly chosen (this BC1

parent is also base of recombination of the continuous

variables), so that there is no mixing between the topology of
two parents by recombination. The first step of the discrete
GN2
mutation is the randomly generation of the step size 6, whose
value is between the limits O and t$max (in the studies dna = substation A substation B

5). 3 is the amount of elements of the vector ~, which are

Fig. 5. EHV test system
changed. After that, 6 elements are chosen and modified. The
modification represents a random position of the rotary In the shown topology the load LD 1 is supplied by line LN1
selector. and load LD2 by line LN2 respectively. The two bus bars in
substation A are coupled by the bus coupler BC 1.
D. Continuous recombination and mutation As a result of the trading contracts are concluded, which are
listed in table 1. The generation unit GN3 supplies the power
The parent i 1is chosen by the recombination of the discrete losses.
Table 1.Contractsfor onetime step
decision vector. A second parent 22 is randomly chosen out of
the ~ individuals. With the information of these parents a child Generation Active power
is generated. The step size vector of the child is built as fol- Contract Load
unit [MW]
lows: 1 GN1 LD 1 200
2 GN2 LD2 450
(20)
~’=+(~’+~2) 3 GN3 LD2 50

Each continuous decision variable of the child is equal to the

congruent element of one of the parents. For these contracts a congestion on line LN2 occurs, caused
by contract 2. The current requirement (2) is not equal zero.
(21) With the described test system the approach is started. To
reach a better convergence a few start individuals are chosen
After recombination the step size vector and the vector of which represent a system state without congestion. This means
decision variables are mutated. First the elements of the step a system state with a redispatch of the generations as shown in
size vector are slightly modified. table 2.
= ~; *e(ndf{O,l}) Table 2. Redispatch of the start generation
(22)
‘x

Whereby ndf (O,1) is a normal distribution timction with an Generation unit Redispatch [MW]
expected value of O and a standard deviation of 1. With this GN2 -250
new step size the mutation of the decision variables is done. GN3 +250

xx =x; + ax * ndf {0,1) (23)

Now this redispatch has to be minimized without violating the
Because of (23) all children are different from their parents requirements. For the strategy parameters (p, h) 8 parents and
with a typical difference of C . The step size vector is adapted 56 children are chosen. In a first case the changing of the
during the generation, so that the Evolution Strategy navigates topology was blocked.
itself to the global optimum. Figure 6 shows the results. After 50 generations the
objective function converges towards the minimal corrective

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 397

0-7803-6674-3/00/$10.00 (C) 2000 IEEE

 redispatch of the active power. At this system state the Compared with other researches, the new method allows to
generation unit GN2 injects the most power which is possible use all known corrective measures (especially the corrective
without an overload of line LN2 (see table 3). The switching) to avoid congestion. The approach uses an
requirements are fidfilled during the whole operation. Evolution Strategy to determine a secure state of the power
system without congestion and with a minimized corrective
Table 3. Minimized redispatch of the solution generation redispatch of the generation units. First numerical experiments
with a chosen test system demonstrate the efficiency of the
Generation unit Redispatch [MW] proposed solution method.
GN2 -191
GN3 +191
VII. REFERENCES

[1] R.S. Fang, A.K, David, “Transmission Congestion Management in an

Electricity Market”, IEEE Trans. Power Systems, vol. 14, no. 3, August
1999, pp. 877-883.
[2] T.W. Gerda, “On Transmission Congestion and Pricing”, IEEE Trans.
Power Systems, vol. 14, no. 1, February 1999, pp. 241-248.
[3] H. Glavitsch, F, Alvarado, “Management of Multiple Congested
Conditions in Unbundled Operation of a Power System”, IEEE Trans.
Power Systems, vol. 13, no. 3, August 1998, pp. 1013-1019.
[4] H.-J. Koglin, H. Miiller, “Overload Reduction Through Corrective
o~
Switching Actions”, IEE International Conference on Power System
o 5 10 15 20 25 30 35 40 45
Monitoring and Control, 1980, pp. 159-164
generation
[5] J.G. Rolim, L.J.B. Machado, “A Study of the use of Corrective
Switching in Transmission Systems”, IEEE Trans. Power Systems, vol.
Fig. 6. Evolution Process without topology changing 14, no, 1, February 1999, pp. 336-341,
[6] K.F. Schafer, J.F. Verstege, “Adaptive Procedure for Masking Effect
Compensation in Contingency Selection Algorithms”, IEEE Trans.
In a next step the changing of the topology is possible, so Power Systems, vol. 5, no. 2, May 1990, pp. 539-546.
that the Evolution Strategy can use the enlarged state space [7] G. Schnyder, H. Glavitsch, “Security Enhancement Using an Optimal
with the discrete decision variables. Switching Power Flow”, IEEE Trans. Power Systems, vol. 5, no. 2, May
1990, pp. 674-680.
[8] D. Shirmohammadi, B. Wollenberg, A. Vojdani, P. Sandrin, M. Pereira,
F. Rahimi, T. Schneider, B. Stott, “Transmission Dispatch and
Congestion Management in the Emerging Energy Market Structures”,
IEEE Trans. Power Systems, vol. 13, no, 4, November 1998, pp. 1466-
1474.
[9] J.OF. Silva, L.J.B. Machado, “Switching Lines Selection to integrate
the Network Topology Optimization with the usual Overload Control
Actions of Electric Power Systems”, 11‘h Power Systems Computation
Conference (PSCC) 1993, pp. 233-239.
[10] J.N. Wrubel, P.S. Rapcienski, K.L. Lee, B.S. Gisin, G.W. Woodzell,
0 5 10 15 20 25 30 35 40 45 “Practical Experience with Corrective Switching Algorithm for On-line
generation Applications”, IEEE Trans. Power Systems, vol. 11, no. 1, February
1996, pp. 415-421.

Fig. 7. Evolution Process with topology changing

VIII. BIOGRAPHIES

Figure 7 shows the value of the objective fimction, which

converges towards zero. That means, no congestion exists and Markus Doll received his DipL-Ing. degree in Electrical Engineering from
the University of Wuppertal in 1995. From 1996 to 1998 he was with PSI AG,
no corrective redispatch of the generation units is necessary.
Berlin. Since 1998 he has worked as a research assistant to Prof. Dr. Verstege
The Evolution Strategy has found a system state, which fidfills at the University of Wuppertal. His areas of interest are congestion
all requirements of a secure power system. The solution of the management, power system control and application of Evolution Strategies in
problem is to switch the bus coupler BC2 on, so that the power power system operation. M. Doll is member of IEEE and member of VDI,
Germany.
flow is splitted into two parts on the lines LN1 and LN2. If the
test system is modeled without the bus coupler BC2, the
Johannes F. Verstege received his Dipl.-Ing. degree in Electrical
Evolution Strategy switches all generation units and loads at Engineering and Dr.-Ing. from Aachen University of Technology in 1970 and
substation B at one bus bar. The result is nearly the same as 1975, respectively. From 1970 to 1979 he worked at the Institute of Power
with BC2. Systems of Aachen University of Technology as a research assistant and later
as a senior researcher. From 1979 to 1982 he was with STEAG, Essen, In
1982 he was appointed full professor for Power System Engineering at the
University of Wuppertal. His research activities have been focnsed on power
VI. CONCLUSIONS
system control, security assessment and enhancement and on optimization
problems. J, F. Verstege is senior member of IEEE and member of VDI and
This paper presents a new approach for a congestion VDE, Germany.
management system. To detect multiple congestion the known
approaches of the security analysis are implemented.

0-7803-6674-3/00/$10.00
0-7803-6672-7/01/$10.00
(C)(C)
2000
2001
IEEE
IEEE 398
0-7803-6674-3/00/$10.00 (C) 2000 IEEE