Indexing terms: Power systems and plant, Linear programming, Transmission network expansion planning
Introduction
0IEE, 1993
Paper 9607C (W),
first received 28th April 1992 and in revised form
26th March 1993
The authors are with the Institute of Power Engineering and Electronics, School of Engineering Sciences, University of Novi Sad, 21000
N o n Sad, Serbia
516
additional operation costs originating from the transmission network are completely neglected.
(b) Interactive transmission expansion planning can be
a very tiring and time-consuming process, when only the
sensitivity-based expansion planning criterion [l] is used.
Moreover, the real optimal solution is never known, and
the problem cannot be well visualised. In some cases,
there are a large number of network reinforcement candidates with very little difference between the expansion
planning criterion values, which, on the other hand,
always favour the construction of longer (more expensive)
lines.
(c) Reliability evaluation using the model of capacity
state classification by subsets from Reference 2 seems to
be inappropriate for meshed transmission networks, as
the Kirchhoff voltage law (KVL) is not encompassed in
this model. This means that the calculated bulk power
system reliability indices are better than the real ones.
This paper proposes a new methodology for optimum
transmission expansion planning, with the goal of
improving the shortcomings listed above. The overall
expansion planning task is decomposed into two independent entities: the investment problem gives the
answer to the best network reinforcements, and the
additional operation expenses and corresponding bulk
power-system reliability indices are calculated by the
operation model. Both models are defined in such a way,
that their aggregation into a unique entity is very simple
and compatible with the generation expansion planning
methodology Wasp [3]. This feature avoids drawback
(a). The investment problem is defined as a linearprogramming model, and, besides the sensitivity-based
interactive expansion planning policy, it considers the
economy of the solution and produces optimal expansion
plans. In this way, drawback (6) is eliminated. The operation problem is defined through an independent Monte
Carlo simulation, based on the DC load flow [4, 51. As
the DC load flow and real-life control and operation
policies are included in this model, drawback (c) is
removed.
This paper is exclusively devoted to the consideration
of the investment problem. It represents recent research
results of improvements on the model proposed in Reference 6. The development of this investment model was
motivated by some possible shortcomings of the original
model, as cited in Reference 6. Namely, it was mentioned
that, in some situations, in the presence of small load curtailments in the existing network (obtained by the
minimum load curtailment (MLC) model [l]), a zerovalued objective function can be obtained as the optimal
solution of the marginal network (MN) model [SI. This
IEE PROCEEDINGS-C, Vol. 140, No. 6, NOVEMBER 1993
II
solve the investment problem
i
I
I
year= year.1
'r'
year=base year
Fig. 1
517
-BO
+ G + Mlf
ITG = ITD
G" Q G Q
cH
IS196
f Qi:
where
G"' Q Go Q GM
(2)
l@lQ6
f"=O
where the superscript designates the corresponding
initial quantities. The physical meaning of the above conditions is that there are nonzero power flows over elements of the existing network, whereas they are all equal
IEE PROCEEDINGS-C, Vol. IN, No. 6, NOVEMBER 1993
Linear-programming
marginal network
associated
constraints
(model of man. 2)
variable cost
When using the DC load flow, the MLC model is specified by the following linear-programming problem:
min z = ITR
+G +R =D
IT(G + R) = ITD
associated
constraints
-BO
min z =
1 Kk-,,,fL-"
(34
- B A O + AG' + Mlf
I T AG = IT(R - G"
0 < A G Q G'-
=R
- G"'+ Go
+ Go)
G"
(3b)
(34
(34
(34
f <f
(3f 1
(4b)
(4)
G"<G<GM
(44
OQRQD
(44
I*l
to zero in new elements. The nonzero power flows satisfy
all imposed operation constraints, but the delivered loads
are on the level the existing network is capable of supplying ( D O ) .
Two different cases of the initial conditions in eqn. 2
should be discussed. The situations where the delivered
loads Do are equal to the required loads D are, from the
planning standpoint, trivial cases, as no network
reinforcements are necessary. Then, there is no need to
solve the investment model of eqn. 1. In the opposite
case, when load curtailments exist in the original network
Do < D), it is necessary to solve the reformulated investment model, in which the initial load flow pattern in eqn.
2 is also incorporated. By substituting eqn. 2 into the
model of eqn. 1 and transforming all the variables to be
non-negative, the following equivalent model is obtained:
(44
(4fl
zero-valued. For this reason, it is proposed here to introduce 'good' reinforcement candidates X - , only into the
LPMN model. This is decided upon the initial loading of
a branch, obtained by the MLC model (eqn. 4). Only in
cases where a relative branch loading exceeds a prespecified limit (decided upon the user-defined factor y, which
indicates the critical zone of the branch loading), a properly directed variablef,-, is set into the LPMN model
(Table 1, item 1). Of all possible new rights-of-way, only a
few are placed into the LPMN model. The corresponding
requirements are listed in Table 1, item 2.
The solution of the LPMN model in eqn. 3 gives
optimum network reinforcements to the planning engineer. In addition, the whole transmission investment
problem can be very well visualised, as the only demands
in the LPMN model are the curtailed loads. Thus, the
treatment of all power flows that can be supplied by the
existing network is avoided. The only load flows participating in the LPMN model are those supplying the
curtailed loads and clearly distinguishing the system
bottlenecks.
Fig. 2
A p=
- 8"
(5)
next outage
I1
verloading exists
I
N
100
LPMN model
MN model
MLC model
LPNM model
136.5
150
150
LPMN model
MLC model
200
another new line should be constructed in the same comdor 4-6 (Fig. 34,giving finally the network that satisfies
all imposed operation constraints (Fig. 34. It is inter-
50
100
LPMN model
MLC model
245.3
15.3
150
MLC model
276.8
Fig. 3 Transm'ssion expanswn planning applied to the Gmwr test system [IO]
(I The &at iteration (1 branch 2-6 ioitiaUy constructed)
b Thc m n d iteration ( I branch 2 - 6 1 brnnch 3-5 conatructcd)
E Thc third iteration (2 bmn&cs 24.1 bnmh 3-5 Fonstructcd)
d The fourth itclation (2 branched 2-6, I branch 3-$1 branch 4-6 constructed)
c
Find solution
SR Mirovico2
Srbobran
N Sad
nodes
length critl
20 41
30km -003
-0brenovac A
967
-L+R
1 P.hjim
Basta
Rhe
Bajlna Bosto
004162 2
He
k j i n o Basta--
--
-- tent A
Vardiste 9 3 9
Pozega
Conclusion
522
References
115-120
6 LEVI, V.A., and CALOVIC, M.S.:A new decomposition based
method for optimal expamion planning of large transmission networks, IEEE Tram., 1991, PWE3-6, pp. 937-943
7 LASDON, L.S.: Optimization theory for large systnns (Macmillan
Co.,New York,1970)
8 GARVER, L.L.: Tmnamision metwork estimation wing h e a r programming, IEEE Trans., 1970,P M ,pp. 1688-1697