IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO.

2, MAY 2013 1355

Transmission Expansion Planning of Systems

With Increasing Wind Power Integration
George A. Orfanos, Pavlos S. Georgilakis, Senior Member, IEEE, and Nikos D. Hatziargyriou, Fellow, IEEE

Abstract—This paper proposes an efficient approach for proba- Branch-node incidence matrix when a line in the
bilistic transmission expansion planning (TEP) that considers load right of way is out of service.
and wind power generation uncertainties. The Benders decomposi-
Vector of active power generation with elements
tion algorithm in conjunction with Monte Carlo simulation is used
to tackle the proposed probabilistic TEP. An upper bound on total (generation in bus ).
load shedding is introduced in order to obtain network solutions Vector of active power generation when a line in
that have an acceptable probability of load curtailment. The pro- the right of way is out of service with
posed approach is applied on Garver six-bus test system and on elements .
IEEE 24-bus reliability test system. The effect of contingency anal-
ysis, load and mainly wind production uncertainties on network Vector of maximum generator capacity.
expansion configurations and costs is investigated. It is shown that Vector of the predicted load.
the method presented can be used effectively to study the effect of
increasing wind power integration on TEP of systems with high Vector of load curtailment with elements .
wind generation uncertainties. Vector of load curtailment when a line in the
Index Terms—Benders decomposition, Monte Carlo simulation, right of way is out of service with
probabilistic contingency analysis, transmission expansion plan- elements .
ning, wind power generation.
Load penalty factor ($/MW).
Load penalty factor for the probabilistic
NOMENCLATURE approach ($/MW).
Cost of a line added to the right of way ($). Total load curtailment cost in normal operation
without contingencies ($).
Susceptance of the line between buses and .
Total load curtailment cost in single contingency
Conductance of the line between buses and . situation ($).
Number of new lines added to the right Total load curtailment cost for all
of way. situations ($).
Initial number of lines between buses and . Upper bound of load curtailment for the planning
Maximum number of lines that can be added to horizon (MW).
the right of way. Probability of events.
Active power flow in the right of way E(y) Expected value of variable y.
(MW). Set of load buses.
Active power flow in the right of way
Set of all existing and new right-of-ways.
when a line in the right of way is out of
service (MW). Set of selected contingencies.
Active power flow limit on the right of
way (MW). I. INTRODUCTION
Phase angle in bus .

T
Phase angle in bus when a line in the HE main objective for transmission expansion plan-
right of way is out of service. ning (TEP) in deregulated power markets is to provide
Branch-node incidence matrix. nondiscriminatory and competitive market conditions to all
stakeholders, while maintaining power system reliability [1].
There is a number of uncertainties that have to be taken
Manuscript received February 09, 2012; revised May 16, 2012 and July 04, into account that can be classified into two categories [2]: 1)
2012; accepted August 14, 2012. Date of publication September 27, 2012; random uncertainties, such as load development, generators’
date of current version April 18, 2013. This work was supported in part by the
operating costs, availability and bidding prices, availability of
European Commission under contract FP7-ENERGY-2007-2-TREN-218903
(IRENE-40 project). Paper no. TPWRS-00120-2012. transmission lines, renewables production and 2) nonrandom
The authors are with the School of Electrical and Computer Engi- uncertainties, such as location of new generators, available
neering, National Technical University of Athens (NTUA), Athens,
transmission expansion investment budget, etc. The statistics
Greece (e-mail: gorfanos@power.ece.ntua.gr; pgeorg@power.ece.ntua.gr;
nh@power.ece.ntua.gr). of random uncertainties can be derived from past observations,
Digital Object Identifier 10.1109/TPWRS.2012.2214242 but nonrandom uncertainties are not repeatable and cannot be

0885-8950/$31.00 © 2012 IEEE

1356 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

statistically represented. The large integration of renewable Section III. The method is applied to Garver six-bus test system
energy sources (RES) into modern power systems has made and to IEEE 24-bus reliability test system and the results are
the TEP problem even more challenging, because the greatly analyzed and compared in Section IV. Conclusion are drawn in
increased uncertainties introduced often require new transmis- Section V.
sion lines in order to maintain a satisfactory level of power
system security and adequacy [3], [4]. II. PROBABILISTIC TEP PROBLEM
Independent of the market conditions, the power system
should always be operated in a way that no contingency A. TEP Problem
triggers cascading outages or causes any form of instability.
In a power system represented by the DC load flow model,
Since securing the system against all possible contingencies is
the mathematical formulation for the static deterministic TEP
practically impossible, the system operator only checks a set
model is
of credible contingencies [5]. Most security rules therefore call
for the system to be able to withstand the loss of any single
component, thus being “ secure” and sometimes the loss (1)
of a selected combination of two components, i.e., “
secure”. In transmission expansion problems, the steady state
security analysis aims at satisfying the nodal power balance (2)
with no violations of the transmission lines maximum flow
under normal and contingency situations. (3)
TEP can be classified as static or dynamic according to the (4)
study period. For static planning, the developer searches for the (5)
suitable circuits that should be added in the current transmission
(6)
system. If multiple years are considered and an optimal expan-
sion along the whole planning horizon is searched, planning (7)
is classified as dynamic. Various methods have been applied (8)
to solve the transmission expansion planning problem, such (9)
as linear programming [6], dynamic programming [7], branch
and bound [8], mixed integer programming [9], decomposi- In the above TEP formulation, also known as adequacy TEP, the
tion techniques [10], simulated annealing [11], tabu search objective is to find an optimal transmission structure to meet
[12], genetic algorithm [13], and differential evolution [14]. the peak load demand with minimum investment and loss of
Probabilistic methods for the solution of TEP problem include load cost, while satisfying operational limitations. Equation (2)
probabilistic reliability criteria [15], risk assessment methods stands for the power nodal balance equation; (3) is the DC power
[16], and chance constrained programming [17]. flow model, while (4)–(6) specify the operational limits of the
In this paper, the static TEP problem is solved with the Ben- system. Constraint (7) defines the range of the investment vari-
ders decomposition technique incorporating security anal- ables. In the above formulation, re-dispatching of generators is
ysis. The deterministic TEP method is described first. More considered.
specifically, secure and adequate transmission solutions are de- In order to include the N-1 security criterion in the TEP for-
termined by assuming different possible future dispatches of mulation, problem (1) is modified as follows [18]:
generators for the peak demand along with generation re-dis-
patching, in order to investigate the impact of generation expan-
sion and bidding behavior to security constrained static trans- (10)
mission planning cost. Generation costs and bidding strategies
can significantly change the outcome of TEP since they can af-
fect lines utilization. However, the system planner can make (11)
a good assumption of these costs and incorporate them into (12)
the TEP problem, while the congestion cost reduction resulting
(13)
from new lines can be compared to the new lines cost for a cer-
tain period of time in a cost-benefit analysis. (14)
The impact of wind power integration on TEP investment (15)
cost while maintaining system security at a satisfactory level (16)
is investigated next. In the proposed probabilistic TEP method,
load and wind power generation uncertainties as well as forced (17)
outage rates of the transmission lines are included and an ac- (18)
ceptable upper bound of possible load shedding is set, in order (19)
to find secure transmission solutions with minimum investment
cost. The DC power flow with the incorporation of losses is Parameters with superscript denote the modified variables
employed for the network representation. The proposed prob- when a line on the right of way is outaged. The con-
abilistic TEP approach is formulated in Section II and solved in straints that should be met in problem (10) are constraints
ORFANOS et al.: TRANSMISSION EXPANSION PLANNING OF SYSTEMS WITH INCREASING WIND POWER INTEGRATION 1357

(2)–(9) with the addition of constraints (11)–(18). If producers’ A. Probabilistic Operation Subproblem
operating costs, reflected in their bids, are taken into account
The probabilistic adequacy subproblem (22)–(28) is used for
or the dispatch of the generators at the specific time of the
the formulation in (21). In the stochastic resource planning,
planning horizon is known, then in problems (1) and (10)
each possible system state is represented by a scenario. The
vector is constant and constraints (5) and (16) are excluded.
Monte Carlo simulation technique [20] is applied to simulate the
This formulation reveals the difficulty of the network planner
random output of wind power generation, load demand and un-
in the new deregulated environment where congestion cost and
certainties of system component outages. For each system state,
the need for new lines are closely related with the generators
the objective of the adequacy operation subproblem is to mini-
bids and cost functions.
mize load shedding under normal operation by applying a gen-
eration dispatch:
B. Probabilistic TEP Problem Formulation
In order to consider the uncertainties in future load demand
and wind power output, a probabilistic formulation of the prob- (22)
lems (1) and (10) is needed. The expected loss of load replaces
the curtailed load under normal operation and the wind power
output together with the outage rate of the transmission lines (23)
and/or uncertainties in future load growth are considered using (24)
Monte Carlo simulation: (25)
(26)
(20) (27)
(28)
The probability of load curtailment exceeding a specified
limit is included in the problem formulation as follows: where is the solution obtained at the th Benders iteration.
Generators lower limits can also be included, representing the
units that must run continuously for security reasons and/or due
to long term bilateral energy contracts. Estimated losses on each
branch are calculated using (29):
(21)
where is a percentage of the total peak load. This upper
bound is used to find solutions that minimize the probability of (29)
having load curtailment over this threshold at the peak load of
the planning horizon. Active losses are then allocated to the loads in the extreme
nodes of each branch and (22)–(28) are solved again. This leads
III. SOLUTION OF THE TEP PROBLEM to an iterative process that terminates when phase angles at all
buses do not change significantly from the previous iteration.
The Benders decomposition technique [19] is used to solve For fixed generation dispatch, vector is constant and constraint
this mixed integer nonlinear problem in both deterministic and (26) is excluded.
probabilistic expressions. The original problem is separated into The power output of a wind turbine as a function of wind
several subproblems that are solved iteratively: 1) the master speed (m/s), is
problem, which is a binary integer programming problem that
identifies the candidate investments and 2) the operation sub-
problems that are linear problems with fixed integer variables
that check whether the scheme selected from the previous
master problem can meet system operation constraints. If any
constraint becomes active in any of the operation subproblems,
a Benders cut is formulated, based on the linear programming
(30)
duality theory, and added cumulatively in the master problem
for solving the next iteration of the algorithm. Benders cuts are is the cut-in and is the cut-out wind speed and is
influenced by the line limits constraints, since when congestion the rated power of the wind turbine. From this linear wind power
occurs, excessive generation cannot be transferred by the production approximation, the power output can be calculated if
congested lines and load in some buses may be curtailed. In the the wind speed and the wind turbine’s characteristics are known.
probabilistic formulation the probability of load curtailment It should be noted that according to current operating practices
above a threshold affects the upper bound of the Benders’ de- in many countries, wind power is priority dispatched. Thus, in
composition method. This helps the convergence of the method this formulation it is assumed that the production of a wind park
to solutions that have a small probability of load curtailment can be curtailed, only if the operation subproblem is infeasible.
and at the same time almost zero probability of load curtailment If line outages rates are not considered in the Monte Carlo
above this threshold. simulations and a deterministic method is followed for the N-1
1358 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

security analysis, problem (22) is solved for each of the secu- B. Investment Subproblem (Master Problem)
rity subproblems. The objective is to minimize load curtailment
The investment subproblem takes as input the Benders cuts
under a single contingency by applying a generation dispatch:
formulated from the operation subproblem(s) and finds the new
lines added at each iteration. This problem is a binary integer
(31) problem, which seeks for the minimum cost of new added lines
with constraints provided by the corresponding subproblem(s).
(32) The formulation of the investment problem is as follows [22]:

The constraints that should be met in problem (31) are

(23)–(29) and (32), with the difference that for each contin- (34)
gency a new branch node incidence matrix is created and the
number of lines in the examined right of way is reduced by one.
The security subproblem is solved for all credible contingen- (35)
cies included in set . After all contingencies have been
considered, the total load curtailment is computed by (33): (36)
(37)
(33)
(38)

Another way of performing the security analysis is to relax where and are the solutions of the adequacy and security
constraint (25), considering vector as the generation dispatch subproblems of the previous iteration and is the sensitivity
of the adequacy subproblem and minimize line overloads [21]. of the optimum values and with respect to the decision
Both methods result in the same final transmission investment variable. The Benders cuts are represented by (35), and sensi-
solutions. It should be noted that re-dispatching of generation is tivity factor is given by
not based on generators’ bids or costs but on minimizing total
load curtailment. (39)
The algorithm of computing the expected loss of load and
the expected dual variables in the operation subproblem of each
where p is the operations subproblem solved ( for the
Benders iteration, follows the following ten steps, where load
adequacy subproblem and for the th security
demand, wind speed and transmission lines availability are as-
subproblem for each contingency ). are the dual
sumed independent:
variables (Lagrange multipliers) of (23) for the adequacy and
1) Determine forced outage rate (FOR) of each transmission
the security subproblems. When node or node is not con-
line and assign a standard uniform probability density
nected to the system, the sensitivity factor for the right of
function (pdf).
way is
2) Determine the pdf of the peak load of the planning horizon.
3) Given the mean value and the standard devi-
(40)
ation of the wind speed, determine the parameters
of the Weibull distribution function that represent the wind
speed at the location of the wind turbines. The total cost for Benders th iteration is the sum of the
4) Generate a number from the standard uniform pdf of the new added lines, and the curtailed load costs, and ,
FOR of each line determined in step 1, and compare it computed from the th investment and operation subproblems,
with its unavailability. If this number is less than its un- respectively.
availability, then the line is on outage, otherwise it is in In the probabilistic formulation of the investment sub-
operation. problem, after all the generated scenarios have been solved
5) Generate a number from the pdf of the peak load deter- for the operation subproblems of the previous iteration, the
mined in step 2, and compute the peak load at each bus expected values of loss of load and the expected values of
accordingly. the sensitivity factors, replace the corresponding deterministic
6) Generate a number from the Weibull distribution of the values in (35). The Benders cuts added to the investment
wind speed determined in step 3, and calculate the power problem of each iteration are as follows:
output of the wind turbine using (30).
7) Solve the adequacy subproblem (22) for the network con-
figuration of step 4, load of step 5 and wind power gener-
ation of step 6 and save the outputs ( , Lagrange multi- (41)
pliers, etc.). The total cost for Benders th iteration is calculated by adding
8) Compute the current expected values of and the dual the cost of the new added lines and the expected curtailed load
variables needed in the next investment subproblem. cost multiplied by the probability
9) Repeat steps 4 to 8, until sufficient accuracy is obtained. computed from the th investment and probabilistic operation
10) Find the probability . subproblems.
ORFANOS et al.: TRANSMISSION EXPANSION PLANNING OF SYSTEMS WITH INCREASING WIND POWER INTEGRATION 1359

IV. RESULTS AND DISCUSSION TABLE I

TEP SCHEMES FOR VARIOUS PEAK LOAD UNCERTAINTIES AND LOAD
A. Introduction SHEDDING UPPER BOUNDS FOR GARVER 6-BUS TEST SYSTEM

The proposed algorithm was implemented in MATLAB 7,

was tested on Garver 6-bus test system [6] and IEEE 24-bus
reliability test system [23] using a PC with Core 2 Duo CPU
clocking at 3.0 GHz and 2 GB of RAM. First, the deterministic
TEP problems (1) and (10) are investigated for both a given gen-
erator dispatch and generation re-dispatching. Then, the proba-
bilistic TEP problem (21) is solved for various peak load uncer-
tainties, upper load shedding bounds and wind charac-
teristics. Transmission lines outages are modeled using a failure
rate of 1% in order to “capture” as many as possible of the
contingency situations in the probabilistic framework, since a
lower value could underestimate the effect of the contingency
conditions in the investment plans [2]. The load penalty factor
is assumed equal to while the load penalty factor calculated in a cost-benefit analysis of the considered transmis-
for the probabilistic approach is set equal to , in sion plans using annual simulation. From the solutions of the
order to find the most secure expansion solutions. It should be fixed generation simulations, it is concluded that the final in-
noted that up to four and up to three circuits could be added vestment plans at Garver’s test system strongly depend on the
per right of way for Garver 6-bus and IEEE 24-bus test system, generation at bus 3, which is closer to the high demand buses 2
respectively. and 5.
2) Probabilistic TEP: In the probabilistic approach, five se-
B. Garver 6-Bus Test System ries of tests are simulated for Garver’s test system. The pdf of
1) Deterministic TEP: For the deterministic TEP approach, the peak load at each bus follows a normal distribution with a
problems (1) and (10) are solved for both re-dispatched and mean equal to the base case data. Only re-dispatched generation
fixed generation. The optimum solution for the adequacy is considered.
problem (1) including losses requires the addition of 3 new Without Wind Generation: In the first three series of simula-
lines between buses 4 and 6 , 1 new line between tions, no wind power is assumed and a different standard devi-
buses 3 and 5 and 1 more line between buses 2 and 3 ation is assigned to the load pdf of each bus (0%, 5%, and 10%)
, while the total investment cost is 130 k$ with zero representing the impact of the peak load forecast uncertainty on
curtailed load. The solution of the security expansion problem TEP. For each standard deviation, different upper bounds for the
(10) requires the addition of 7 new lines ( , , maximum load shedding parameter are examined. The
and ), while the total cost is 180 k$ achieving values assigned to these bounds are calculated as percentages
zero load shedding for all the configurations of the of the total peak load of the system. The corresponding results
system. for each simulation are given in Table I. These results demon-
In order to investigate the impact of different generation dis- strate that the final planning schemes are significantly affected
patches on transmission planning, total generation is kept con- by peak load uncertainty (standard deviation over peak load,
stant (760 MW), but it is dispatched differently to the three gen- ) and load shedding upper bound . As
erators. If generation at buses 1, 3 and 6 is fixed at , expected, for the same load shedding, higher investment costs
and , then for the security expan- are needed when peak load uncertainty increases. On the other
sion problem (10) the optimum solution requires ten new lines hand, when load shedding tolerance increases, less expensive
( , , and ) and the investment configurations are selected. In this way, the proposed method
cost is 298 k$. If the generation dispatch is , provides the flexibility to the planner to balance the objectives
and , the solution obtained for the of minimizing total investment cost and risk. It should be noted
security problem requires eleven new lines ( , , that for zero peak load uncertainty and zero load shedding, the
, and ) and the total cost increases results of the probabilistic and deterministic TEP are the same
to 318 k$. Finally, if the generation dispatch is 50 MW at bus which justifies the selected failure rate for transmission lines
1, 265 MW at bus 3 and 445 MW at bus 6, ten lines are added outages.
for a secure system ( , , ) with a total With Wind Generation: In order to simulate the wind power
investment cost of 270 k$. uncertainty, it is considered that the installed generation at bus 3
From the previous analysis it is concluded that different se- is composed of 120 MW of wind power capacity plus 240 MW
cure and adequate configurations of the network are obtained of conventional thermal power capacity. The generation output
when generation is considered fixed or when it can be re-dis- of all the wind turbines follows (30), with parameters set at
patched to avoid load curtailment. In power system operation, , and . The wind
generation production depends on generators bids, however in speed characteristics for the base scenario are
static TEP formulation, re-dispatched generation leads to less and . For 10% standard deviation of the
expensive investment plans. The cost of re-dispatching can be peak load and for various maximum load-shedding values, the
1360 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

TABLE II TABLE IV
TEP SCHEMES FOR AND WIND TEP SCHEMES FOR IEEE 24-BUS RELIABILITY TEST
POWER GENERATION CONSIDERATION FOR GARVER 6-BUS TEST SYSTEM SYSTEM USING THE DETERMINISTIC TEP METHOD

TABLE III
LOAD SHEDDING PROBABILITY FOR DIFFERENT WIND
SPEED CHARACTERISTICS AND
FOR GARVER 6-BUS TEST SYSTEM

results for the optimum plans are shown in Table II. Moreover,
Table III provides the probability of load shedding for the less
expensive configuration for various wind speed characteristics.
For the investment plan presented in Table III, the addition
of only one new line compared to the solution obtained in
the deterministic security TEP problem, can maintain the prob-
ability of load curtailment at satisfactory level, although part
of conventional generation capacity has been replaced by wind
power capacity at node 3. The uncertainty of wind power gen-
Fig. 1. Probability distribution function f(x) of generation output of WF7 and
eration raises the planning risk and produces more expensive WF22 wind farms located at buses 7 and 22, respectively.
network configurations to meet the same standards of system
adequacy and security. The probability of having load curtail-
ment is increasing when lower wind speeds are considered. functions. If the same random number is used for both sites, the
power outputs of the two wind parks become fully correlated.
C. IEEE 24-Bus Reliability Test System However, in this paper some correlation between the two sites is
1) Deterministic TEP: Results from the deterministic anal- assumed using the cross-correlation index [24]. Fig. 1 shows the
ysis considering only re-dispatched generators are presented in probability distribution function of the two wind farms output
Table IV. The loads and generation levels are assumed to be when and , assuming a
three times higher than their original values [23]. From the re- cross-correlation index between the two wind sites of 0.75.
sults it is concluded that the security constraints increase The results for 5% peak load uncertainty and for various load-
significantly the number of new added lines, especially at the shedding limits and different wind speed characteristics are pre-
lower parts of the network where buses are not so strongly con- sented in Table V. The probability of load curtailment exceeding
nected, and raises the total investment cost by $3.59 million for the upper bound is less than 0.05 for all investment plans.
the case of considering line losses. A sensitivity analysis that The solutions obtained for have a greater risk of
takes into consideration multiple load levels, including genera- load curtailment compared to the cases, but lead to
tors cost and bid curves and outage rates is needed, to perform a less expensive expansion schemes with rather small probability
cost benefit analysis for the suggested transmission investments. of load curtailment. The number of iterations needed for the
It is possible that a less expanded, but secure grid burdened by convergence of Benders’ decomposition method is very much
lines congestion costs requiring generator re-dispatching might dependent on the number of new lines installed for each TEP
be more cost effective for the transmission assets owner and op- scheme, while the CPU time needed for the probabilistic TEP
erator than an overbuild network. method to converge depends also on the probabilistic adequacy
2) Probabilistic TEP: In order to simulate wind generation, subproblems solved for each Benders’ algorithm iteration. The
half of the conventional generation capacity connected at buses solution for the first TEP scheme of Table V, was the most time
7 and 22 is assumed to be replaced by wind generators with wind demanding. For this case, the algorithm converged after 61 iter-
power parameters equal to , ations and 57 min of CPU time.
and . If the two wind sites (of buses 7 and 22) The cumulative distribution function (cdf), F(x), of load
are assumed uncorrelated, random wind speed numbers for the curtailment for zero actual wind generation production for
two sites are generated from two different Weibull distribution the six plans of Table V is presented in Fig. 2. If, during
ORFANOS et al.: TRANSMISSION EXPANSION PLANNING OF SYSTEMS WITH INCREASING WIND POWER INTEGRATION 1361

TABLE V
TEP SCHEMES FOR 5% PEAK LOAD UNCERTAINTY AND CORRELATION INDEX
0.75 BETWEEN THE TWO WIND SITES FOR IEEE 24-BUS TEST SYSTEM

Fig. 3. TEP investment cost for various percentages of conventional capacity

replacement by wind power.

the same wind generation capacity as before (450 MW at bus

7 and 450 MW at bus 22) replaces a predefined percentage of
the rest 900 MW of conventional generation capacity located at
the same buses. The results show that for wind generation inte-
gration up to 450 MW, i.e., lower than 50% of conventional ca-
pacity replacement, the resulting TEP investment cost remains
the same. Overall, it can be concluded that the effect of high
wind power integration on TEP depends critically on the level
of penetration and cannot be straightforward determined. The
availability of probabilistic, flexible tools, as the one presented
in this paper, can effectively solve this problem.

V. CONCLUSIONS

The increase of wind power penetration into power systems

has increased uncertainties and introduced new challenges to
the transmission planners. This paper proposes an efficient
approach to study the effect of wind power integration on trans-
mission expansion planning. A probabilistic framework for the
static TEP problem is presented that considers N-1 security
analysis, as well as uncertainties in wind power generation and
load demand. The Benders decomposition algorithm is adopted
Fig. 2. Cumulative distribution function F(x) of load curtailment for zero wind to solve the proposed TEP problem. The algorithm is first
power production for the six plans of Table V. evaluated for the deterministic TEP problem considering the
impact of re-dispatched generation on transmission investment
cost. Monte Carlo simulation is used to apply the considered
the planning process the wind speed is overestimated probabilistic approach, investigating also the introduction of an
, then for peak load conditions, if upper load shedding limit on the probabilistic TEP formulation.
the actual wind power production is zero, there is a high prob- The proposed method is tested on Garver 6-bus and on IEEE
ability of load curtailment even for . On the other 24-bus reliability test systems. The results show that the incor-
hand, if wind speed is underestimated , poration of a variety of uncertainties increases transmission in-
the network is overinvested and the probability of load cur- vestment cost, which can be mitigated by the introduced upper
tailment is low for zero wind power generation. For average limit in load shedding. The effect of wind integration in trans-
wind speeds , the probability of load mission planning is evaluated for different wind speed charac-
curtailment for the corresponding network configuration is teristics and for replacement of various amounts of conventional
rather low and comparable to the low average wind speed case, generation. Transmission investments with increased security
while the cost of the investment needed is considerably lower. should be decided based on a probabilistic approach that takes
Fig. 3 shows the TEP investment cost for different fractions into consideration load demand and wind power generation un-
of conventional generation replacement by wind power at buses certainties. The probabilistic method presented in this paper can
7 and 22, considering and , be effectively used to provide effective transmission expansion
5% peak load uncertainty and . It is assumed that plans with increased flexibility.
1362 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 2, MAY 2013

REFERENCES [17] N. Yang and F. Wen, “A chance constrained programming approach to

[1] P. Maghouli, S. H. Hosseini, M. O. Buygi, and M. Shahidehpour, “A transmission system expansion planning,” Elect. Power Syst. Res., vol.
multi-objective framework for transmission expansion planning in 75, no. 2–3, pp. 171–177, 2005.
deregulated environment,” IEEE Trans. Power Syst., vol. 24, no. 2, [18] I. de J. Silva, M. J. Rider, R. Romero, A. V. Garcia, and C. A. Murari,
pp. 1051–1061, May 2009. “Transmission network expansion planning with security constraints,”
[2] M. O. Buygi, G. Balzer, H. M. Shanechi, and M. Shahidehpour, Proc. Inst. Elect. Eng., Gen., Transm., Distrib., vol. 152, no. 6, pp.
“Market-based transmission expansion planning,” IEEE Trans. Power 828–836, Nov. 2005.
Syst., vol. 23, no. 4, pp. 2060–2067, Nov. 2004. [19] M. Shahidehpour, H. Yamin, and Z. Li, Market Operations in Elec-
[3] F. Bouffard and F. D. Galiana, “Stochastic security for operations tric Power Systems: Forecasting, Scheduling and Risk Management.
planning with significant wind power generation,” IEEE Trans. Power New York: Wiley, 2002.
Syst., vol. 22, no. 1, pp. 306–316, Feb. 2008. [20] W. Li, Risk Assessment of Power Systems: Models, Methods, and Ap-
[4] R. Billinton and W. Wangdee, “Reliability-based transmission rein- plications. New York: Wiley, 2005.
forcement planning associated with large-scale wind farms,” IEEE [21] H. Yu, C. Y. Chung, K. P. Qong, and J. H. Zhang, “A chance con-
Trans. Power Syst., vol. 22, no. 1, pp. 34–41, Feb. 2007. strained transmission network expansion planning method with con-
[5] T. Van Cutsem and C. Vournas, Voltage Stability of Electric Power sideration of load and wind farm uncertainties,” IEEE Trans. Power
Systems. Norwell, MA: Kluwer, 1998. Syst., vol. 24, no. 3, pp. 1568–1576, Aug. 2009.
[6] R. Villasana, L. L. Garver, and S. L. Salon, “Transmission network [22] R. Romero and A. Monticelli, “A hierarchical decomposition approach
planning using linear programming,” IEEE Trans. Power App. Syst., for transmission network expansion planning,” IEEE Trans. Power
vol. PAS-104, no. 2, pp. 349–356, Feb. 1985. Syst., vol. 9, no. 1, pp. 373–380, Feb. 1994.
[7] Y. P. Dusonchet and A. H. El-Abiad, “Transmission planning using [23] IEEE Reliability test system task force of the application of probability
discrete dynamic optimization,” IEEE Trans. Power App. Syst., vol. methods subcommittee, “Reliability test system,” IEEE Trans. Power
PAS-92, no. 4, pp. 1358–1371, Jul. 1973. App. Syst., vol. PAS-98, no. 6, pp. 2047–2054, Nov.–Dec. 1979.
[8] J. Choi, A. A. El-Keib, and T. Tran, “A fuzzy branch and bound-based [24] W. Wangdee and R. Billinton, “Considering load carrying capability
transmission system expansion planning for the highest satisfaction and wind speed correlation of WECS in generation adequacy assess-
level of the decision maker,” IEEE Trans. Power Syst., vol. 20, no. 1, ment,” IEEE Trans. Energy Convers., vol. 21, no. 3, pp. 734–741, Sep.
pp. 476–484, Feb. 2005. 2006.
[9] N. Alguacil, A. L. Motto, and A. J. Conejo, “Transmission expansion
planning: A mixed-integer LP approach,” IEEE Trans. Power Syst.,
vol. 18, no. 3, pp. 1070–1077, Aug. 2003. George A. Orfanos is currently pursuing the Ph.D. degree at the School of
[10] S. Binato, M. V. F. Pereira, and S. Granville, “A new Benders decom- Electrical and Computer Engineering of the National Technical University of
position approach to solve power transmission network design prob- Athens (NTUA), Athens, Greece.
lems,” IEEE Trans. Power Syst., vol. 16, no. 2, pp. 235–240, May 2001. His research interests include power system planning, electricity markets, and
[11] A. S. D. Braga and J. T. Saraiva, “A multiyear dynamic approach for distributed generation.
transmission expansion planning and long-term marginal costs compu-
tation,” IEEE Trans. Power Syst., vol. 20, no. 3, pp. 1631–1639, Aug.
2005.
[12] E. L. da Silva, J. M. A. Ortiz, G. C. de Oliveira, and S. Binato, “Trans-
mission network expansion planning under a tabu search approach,” Pavlos S. Georgilakis (M’01–SM’11) is currently a Lecturer at the School of
IEEE Trans. Power Syst., vol. 16, no. 1, pp. 62–68, Feb. 2001. Electrical and Computer Engineering of the National Technical University of
[13] E. L. da Silva, H. A. Gil, and J. M. Areiza, “Transmission network ex- Athens (NTUA), Athens, Greece.
pansion planning under an improved genetic algorithm,” IEEE Trans. His current research interests include power systems optimization, renewable
Power Syst., vol. 15, no. 3, pp. 1168–1175, Aug. 2000. energy sources, and distributed generation.
[14] T. Sum-Im, G. A. Taylor, M. R. Irving, and Y. H. Song, “Differen-
tial evolution algorithm for static and multistage transmission expan-
sion planning,” IET Gen., Transm., Distrib., vol. 3, no. 4, pp. 365–384,
2009. Nikos D. Hatziargyriou (SM’90–F’09) is currently a Professor at the School
[15] J. Choi, T. Tran, A. A. El-Keib, R. Thomas, H. S. Oh, and R. Billinton, of Electrical and Computer Engineering of the National Technical University of
“A method for transmission system expansion planning considering Athens (NTUA), Athens, Greece, and Deputy CEO of the Public Power Corpo-
probabilistic reliability criteria,” IEEE Trans. Power Syst., vol. 29, no. ration, the Electricity Utility of Greece. His research interests include dispersed
3, pp. 1606–1615, Aug. 2005. and renewable generation, dynamic security assessment, and application of ar-
[16] M. O. Buygi, H. M. Shanechi, G. Balzer, and M. Shahidehpour, “Trans- tificial intelligence techniques to power systems.
mission planning approaches in restructured power systems,” in Proc. Prof. Hatziargyriou is convener of CIGRE Study Committee C6 “Dispersed
IEEE Power Tech Conf., Bologna, Italy, 2003. Generation” and a member of the Technical Chamber of Greece.