You are on page 1of 5

2017 7th International Conference on Power Systems (ICPS)

College of Engineering Pune, India. Dec 21-23, 2017

Optimal Tie Line Placement of Distribution


System Incorporating Performance Based Rates
Aprajay Verma K. Shanti Swarup
Department of Electrical Engineering Department of Electrical Engineering
Indian Institute of Technology Indian Institute of Technology
Chennai,India Chennai,India
Email:ee16s034@smail.iitm.ac.in Email:swarup@ee.iitm.ac.in

Abstract—In this paper optimal tie line placement problem A. Literature Review
is solved focusing on minimization of penalties incurred
Tie line planning problem using probabilistic load flow
by Distribution System Operator(DSO) for not meeting
reliability standardVA discrete function of penalty rate model for transmission systems in [7] ,reliability and
was taken which varies increasingly with interruption time. availability analysis of transmission network was done
The risk profiles of each load was simulated by Monte [8], a hybrid method of fuzzy set and Monte Carlo
Carlo simulation method considering failure outage data of simulation for reliability evaluation has been proposed
the feeders and lateral transformers.A graph theory based
in [9],optimization of substation assets has been done
distribution load flow has been conducted and optimal tie line
location was found out by fundamental loop based algorithm. using petri net model support for Monte Carlo simulation
in [10]. A method for power deliveries for companies
Keywords—Reward and penalty schemes, Tie line place- to integrate performance based rates in overall business
ment, Performance based rates, Monte carlo simulation, model was proposed in [11] , a reliability index probability
Fundamental loops distributions are used to establish an appropriate balance
between reliability improvement and cost saving strategies
in [12], a tool for assessing the uncertainties of system
I. I NTRODUCTION reliability and the information to manage PBR risk has
been developed in [13]. A graph theory based network
Distribution System Operators(DSO) in deregulated flow analysis has been used to determine the maximum
market systems these days have to meet with certain reli- flow that can be transported between two nodes within a
ability standards , the motivation to do so comes from the directed graph [14]. A novel graph theory based approach
reward and penalty scheme(RPS) [1] imposed by system has been used for restoring large scale power systems
regulators. These schemes have certain target levels for affected by complete blackout [15].A power load flow
reliability standards and regulator usually imposes reward solution has been solved by a directed graph with edges
and penalty based on whether DSO has met those target represented by a transmission line in [16]. A fast method
levels or not. The development method for the regulator has been proposed which can decide downstream and
to enforce such a regulation was studied in [2] ,the study upstream power flow in [17].
was conducted for three scenarios with planning method This paper includes formulation based on risk levels of
made from only regulator’s point of view, another study various loads, the risk levels were obtained by time se-
[3] used Monte-Carlo simulation to get average number of quential Monte-Carlo simulation [18] simulated in Python
worst case scenario . It has been seen that the quality of and C. The reliability level of loads is strongly dependent
supply can be improved considerably if the optimization is on the reliability of components feeding it, in [19] it was
done in planning stage itself. Hence ,there was a need of a found that failure rate of breakers and bus bars is very
study that incorporated RPS scheme at the planning stage less as compared to transformers and lines hence in this
itself.The key element of planning stage of distribution study only failure rates of transformers and lines was con-
system is placement of tie line ,the reliability of system sidered. Failure outage data was obtained from [19] and
is severely impacted by the position of tie lines which are reward/penalty rates were referred from [1]. An algorithm
used in system to restore power in a bus when it has to based on fundamental loop was used to find the optimal
be isolated due to occurrence of faults in particular feeder. buses for tie line placement. The contents of paper are
The tie line placement problem in past has been solved by summarized as follows section II describes implementation
various formulation [4] [5] but all of them include usage of Monte Carlo simulation, section III describes reward
of mean failure rate which cannot give the probability and penalty schemes, section IV describes the applied
of various interruption level in a system which makes algorithm and section V results and conclusion.
impossible for DSO to properly estimate the penalty or
reward that a particular placement of tie line will give as II. M ONTE C ARLO S IMULATION
Reward/Penalty rate changes with average customer cost Monte Carlo simulation for power system has been in
and moreover, traditionally N-1 criteria is used which has use for quite sometime.It involves generation of random
not been proven to be cost effective [6] as it does not take number following a particular probability curve,in this pa-
cognizance of probobality of each failure. per inverse transformation method has been used [20] to do

978-1-5386-1789-2/17/$31.00 ©2017 IEEE 591


Start
Line 1
Bus 3
Read database
Line 0
Bus 0 Bus 2 Line 2
Line 3 Count=0
Bus 4 Bus 1

Get Branch and Transformer Switch Status


Figure 1: Directed graph describing the downstream nodes of bus2

the same. The analyses was done at 11KV level and only
permanent outages were considered. The probability of Check if switch No
failure of component was modeled by exponential distribu- status have change Count+=1
from last iteration
tion and the probability of time to repair was modeled by
log-normal distribution [21].The simulation for one year
was conducted by dividing time sequence into blocks of
15 minutes. Each sample had an independent set of switch
Yes
statuses for each feeder and transformer.Moreover voltage
profile of each load was also observed by conducting Condition incident matrix according to switch status
distribution load flow.
In order to calculate the net load faced by each bus a Detect leaf nodes
graph theory based algorithm was followed. The network
was represented by directed graph with feeder always
Prepare tree using all the information collected above
pointing towards source nodes.The loads faced by each
bus were calculated by eq. 1.
 Do Distribution Load Flow to get Network Data

Sj = Si,j + Si,b (1)


bGj Collect the relevant statistics

Sj :- Total Load at Bus j


Si,j :- Actual Load being drawn at Bus j Executed in C
Gj :- Set of all downstream nodes w.r.t. bus j
Count
Fig. 1 explains the directed graph used to calculate ≤ 35040
Yes
Executed in Python
the loads of bus 2 in the given 5 bus network.As seen
in this figure the load faced by bus 2 will be loads of
No
downstream buses (in red) in addition to the load at bus 2 Stop
itself. Now the current in each feeder is calculated similar
to the forward sweep of forward backward algorithm but to Figure 2: Flow chart describing the Monte Carlo simulation for one year
calculate the bus voltages we went forward again and use
the potential drop in lines obtained from current values in
first forward sweep to calculate bus voltages recursively. level of 95% . The inequality in eq. 2 has to be transformed
The following flowchart Fig. 2 explains the simulation to an equality by taking the greatest bound as in eq. 3 and
for one year the number of years for which simulation was solved by non linear one dimensional root finding
has to be conducted to get a statistically valid result algorithm .
is determined by convergence criteria. Convergence of  inf
2 −y 2 /2 2E|α|3
Monte Carlo simulation was obtained by application of √ √ e dy + √ √ = 0.05
Bikeliss theorem and central limit theorem as given in 2π 10−5σ N N (σ + 10−5 N )3
[20]. The final equation that came was (3)
 inf III. R EWARD AND P ENALTY S CHEMES
2 2 2AE|α|3
√ √ e−y /2 dy + √ √ ≤ρ (2) In deregulated market the utility offering minimum price
2π ξ σN N (σ + ξ N )3
to customer prevails, this has regulated in reduction of the
α :- Set of I.I.D. variable cost invested by utility on maintenance which has caused
σ :- Standard deviation of α decrease in customer reliability. Therefore, a regulating
N :- Number of values in α mechanism was required that should give incentive to the
ξ :- Required level of accuracy level utility for increasing its reliability standard. In order to
ρ :- Confidence level solve this problem system regulator started implementing
A :- constant between 1 and √12π reward and penalty schemes or performance based rates. A
System Average Interruption Frequency Index(SAIFI) method has been develop in [2] for regulators to prescribe
was taken as convergence parameter , mean and standard rates using customer interruption costs, variation in the
deviation was estimated by taking one year worth of sam- DSOs financial risk and to limit the DSOs financial risk. In
ples the accuracy was taken to be 10−5 having confidence this paper a discrete pricing scheme is taken, discretization

592
·104
Line 1
1 Bus 3
Line 0
Bus 0 Bus 2 Line 2
0.5 Line 3
Bus 4 Bus 1
Price(£/hr)

0 Figure 4: 5 bus sytem taken under consideration

Table I: Table showing expected penalty obtained


−0.5
Sno. Bus no. Expectation of Penalty/Reward(£/hr)
1 0 -3022.1349
−1 2 1 -11019.27
3 2 -4928.596
0 5 7 12 14
4 3 -8548.732
Number of time samples(x15min) 5 4 -8150.66

Figure 3: Reward and Penalty rate considered per customer

Carlo Simulation was carried out in intel i7 2.3 GHz 4 core


has been done based on prices in [1] . The cost function processor in ubuntu-debian environment, multiprocessing
C(x) in end looks like Fig. 3 .It can be seen that this was used to fasten the process and data was handled by
function has 5 levels. The expectation of reward/penalty SQLite database. For calculating eq. 3 brent method was
of each bus is dependent on the probabilities of cumulative used which was implemented in python. On solving eq. 3
probability distribution of its CAIDI(Customer Average for N it was found out that it will take 2840 years of worth
Interruption Duration Index) calculated by Monte Carlo data for simulation to converge taking System Average
Simulation. Interruption Frequency Index as convergence parameter
. Hence, Simulation corresponding to one year was ran
IV. F UNDAMENTAL L OOP BASED A LGORITHM 2840 times. The loads of all the bus were taken as
Fundamental Loop is a closed path with all the 0.534+j0.308p.u. and failure rate of lateral transformer is
branches except one belonging to the cut-sets of taken as 0.0150 and average repair time as 10hr.The base
graph.This situation corresponds to the one in distribution voltage is taken as 11KV and base power as 1MW. After
systems, where the network is already radial in nature Monte Carlo simulations following CAIDI were obtained
which has topology similar to tree and addition of tie Fig. 7-11.It can be seen that the probability distribution
line will form a fundamental loop.The fundamental loop curve of bus 0 is different in shape as compared to others,
can be characterized uniquely by buses it traverses. Our the reason being that as bus 0 is source node the cause
objective is to find location of tie line that gives us of failure of a load is dependent only on failure of lateral
maximum reduction in utility’s penalty due to outages. transformers, while for other buses the major cause of load
Hence, the algorithm in this paper checks for the sum outage is the failure in lines coming in the path containing
of expectation of rewards/penalties in each buses in all load and source buses. The Expected penalty of each nodes
the possible fundamental loops and select the loop which in Tab. I were calculated using these CAIDI plots and cost
gives us the minimum value. curve obtained in Fig. 3
All the fundamental loops of the network were tabulated
and the one which showed minimum value of Cj was
A. Problem Formulation chosen for installation of tie line. In this case it can be
The problem can be seen as a minimization problem seen in the Tab. II that fundamental loop 7 has the least
in which minimum value is to be picked from a discrete value of Cj and the tie line can be installed between 3-1
set of numbers, now if we define Ωj being a set of all Fig. 5 as all other edges of loop are already the part of
fundamental loops present in a network and j being an the initial network.
element of Ωj than our problem corresponds to
Select j such that Cj gives minimum value, where

Cj = Ei (penalty/reward) jΩj (4)
iFj
Line 1 Tie-line
Fj :- Set of all the buses in fundamental loop j
Bus 3
Ei :- Expectation of Penalty/reward of bus i Line 0
Bus 0 Bus 2 Line 2
V. R ESULTS Line 3
A five bus system shown in Fig. 4 was taken for study Bus 4 Bus 1
the data for the system is given in Tab. III, it was assumed
that only one tie line is available for dispatch .The Monte Figure 5: Final System showing tie-line location

593
Table II: Table showing Cj values of fundamental loops
location of tie line, cost analysis can also be done to
Sno. Fundamental loop Value of Cj (£/hr)
calculate the return of investment by subtracting the cost
1 0-2-0 -14041.4137 of installing the tie line by the amount of money saved
2 0-2-4-0 -16101.391 due to decrease in penalties.
3 0-2-4-1-0 -27120.6609
4 0-2-3-0 -16499.46 R EFERENCES
5 2-4-2 -13079.256
6 2-3-2 -13474.328 [1] H. Mohammadnezhad-Shourkaei, A. Abiri-Jahromi, and M. Fotuhi-
7 2-3-1-4-2 -32467.258 Firuzabad, “Incorporating service quality regulation in distribution
8 2-4-1-2 -24098.526 system maintenance strategy,” IEEE Transactions on Power Deliv-
9 3-4-2-3 -21627.988 ery, vol. 26, no. 4, pp. 2495–2504, Oct 2011.
[2] K. Alvehag and K. Awodele, “Impact of reward and penalty scheme
10 1-4-1 -19169.93
on the incentives for distribution system reliability,” in 2014 IEEE
PES General Meeting — Conference Exposition, July 2014, pp.
1–1.
[3] A. M. L. da Silva, A. M. Cassula, L. C. Nascimento, J. C.
The placement of tie line is dependent on lot of external Freire, C. E. Sacramento, and A. C. R. Guimaraes, “Chronological
factors such as location of two buses,distance between monte carlo-based assessment of distribution system reliability,” in
2006 International Conference on Probabilistic Methods Applied
them and feasibility of laying new line between them. to Power Systems, June 2006, pp. 1–7.
Hence, it makes sense to seek next optimal solution if the [4] N. G. Paterakis, A. Mazza, S. F. Santos, O. Erdinc, G. Chicco, A. G.
most optimal solution is infeasible to realize practically, as Bakirtzis, and J. Catalao, “Multi-objective reconfiguration of radial
distribution systems using reliability indices,” in 2016 IEEE/PES
we can see Tab. II also gives out the priority list of all the Transmission and Distribution Conference and Exposition (T D),
fundamental loops as it gives value of penalties explicitly, May 2016, pp. 1–1.
hence it can be observed that second most optimal location [5] B. Khorshid-Ghazani, H. Seyedi, B. Mohammadi-ivatloo, K. Zare,
and S. Shargh, “Reconfiguration of distribution networks con-
for tie line can be derived from fundamental loop no 3 , sidering coordination of the protective devices,” IET Generation,
which gives a tie line between 0 and 1 Fig. 6. Transmission Distribution, vol. 11, no. 1, pp. 82–92, 2017.
[6] Z. Limei, L. Wei, W. Tao, C. Yanyan, and S. Jian, “Research on
reliability-based feeder system planning methods,” in 2016 China
International Conference on Electricity Distribution (CICED), Aug
Line 1 2016, pp. 1–6.
Bus 3 [7] S. Baidya, A. Walid, Nahid-Al, and A. H. Chowdhury, “Tie line
Line 0 planning using probabilistic load model,” in 2012 7th International
Bus 0 Bus 2 Line 2 Conference on Electrical and Computer Engineering, Dec 2012,
Tie-line Line 3 pp. 197–200.
Bus 4 Bus 1
[8] E. Al-Wafi, A. Al-Subhi, and M. Al-Muhaini, “Reliability assess-
ment of a practical power system using monte carlo simulation,”
in 2015 Saudi Arabia Smart Grid (SASG), Dec 2015, pp. 1–6.
Figure 6: Figure showing 2nd best tie-line location [9] M. M. Mahmoodi, E. Afjei, F. Nadimi, and A. Siadatan, “A
new fuzzy-monte carlo simulation approach and its application in
reliability evaluation of power systems protection,” in 2014 22nd
Iranian Conference on Electrical Engineering (ICEE), May 2014,
A. Inferences pp. 831–835.
It is important to observe that this simulation gives [10] O. Fouathia, J. C. Maun, P. E. Labeau, and D. Wiot, “Petri net based
monte carlo simulation for modeling and analyzing the reliability
a restoration strategy for a first order permanent outage of distribution substation facilities in belgian power system,” in
simultaneously, as it can tell which fundamental loop a CIRED 2005 - 18th International Conference and Exhibition on
branch belongs implying which tie switch to operate when Electricity Distribution, June 2005, pp. 1–5.
[11] R. E. Brown, “System reliability and power quality: performance-
a fault occurs in that branch.This would save operation based rates and guarantees,” in IEEE Power Engineering Society
time considerably when a fault occurs as system operator Summer Meeting,, vol. 2, July 2002, pp. 784–787 vol.2.
will not have to decide the appropriate tie line to be closed [12] R. Billinton, L. Cui, and Z. Pan, “Quantitative reliability consid-
erations in the determination of performance-based rates and cus-
such that it restores power optimally .The execution time tomer service disruption payments,” IEE Proceedings - Generation,
of simulation was 50 min which can be justified as we are Transmission and Distribution, vol. 149, no. 6, pp. 640–644, Nov
solving planning problem and simulations had to be done 2002.
off line. [13] R. E. Brown and J. J. Burke, “Managing the risk of performance
based rates,” IEEE Transactions on Power Systems, vol. 15, no. 2,
pp. 893–898, May 2000.
VI. C ONCLUSION
[14] T. Werho, V. Vittal, S. Kolluri, and S. M. Wong, “Power system
A Monte Carlo based reliability network was formed connectivity monitoring using a graph theory network flow algo-
using Python and C . Multiprocessing was used to decrease rithm,” IEEE Transactions on Power Systems, vol. 31, no. 6, pp.
4945–4952, Nov 2016.
the time required for simulation, improvement of factor of [15] J. Quirs-Torts and V. Terzija, “A graph theory based new approach
3 was observed courtesy of that. A probability distribution for power system restoration,” in 2013 IEEE Grenoble Conference,
curve of customer average interruption duration index June 2013, pp. 1–6.
[16] S.-K. Chai and A. Sekar, “Graph theory application to deregulated
was derived from the simulations. A discrete Performance power system,” in Proceedings of the 33rd Southeastern Symposium
Based Rate was formed and a method to find optimal on System Theory (Cat. No.01EX460), Mar 2001, pp. 117–121.
location of tie line was proposed that aims to minimize the [17] F. F. Wu, Y. Ni, and P. Wei, “Power transfer allocation for open
access using graph theory-fundamentals and applications in systems
penalty imposed by system regulator for not meeting the without loopflow,” IEEE Transactions on Power Systems, vol. 15,
reliability standards. It was found that the given method no. 3, pp. 923–929, Aug 2000.
also simultaneously solves restoration problem for the [18] C. Marantes, G. Strbac, and R. Allan, “Sequential monte carlo
simulation assessment of the quality of supply in distribution net-
single layered permanent outage faults. A 5 bus test system works,” in 2006 International Conference on Probabilistic Methods
was used to analyze the method. Given the geographical Applied to Power Systems, June 2006, pp. 1–6.

594
[19] R. N. Allan, R. Billinton, I. Sjarief, L. Goel, and K. S. So, “A
reliability test system for educational purposes-basic distribution
system data and results,” IEEE Transactions on Power Systems,
vol. 6, no. 2, pp. 813–820, May 1991.
[20] D. T. Carl Graham, Stochastic Simulation and Monte Carlo Meth-
ods. Springer, 2013.
[21] W. L. Roy Billinton, Reliability Assessment of Electric Power
Systems Using Monte Carlo Methods. Springer, 1994.

VII. A PPENDIX
This Appendix includes the branch data of system taken
into consideration and the probability distribution function
observed of all the buses.
Table III: Branch data of system taken

Sno. Impedance(p.u./km) Length(km) Failure rate(occ./km-year) Repair time(hr)


0 0.0120+j0.003 1 0.0650 5
1 0.0120+j0.003 2 0.0650 5
2 0.0120+j0.003 3 0.0650 5
3 0.0120+j0.003 4 0.0650 5 Figure 9: Probability distribution function of CAIDI of bus2

Figure 10: Probability distribution function of CAIDI of bus3


Figure 7: Probability distribution function of CAIDI of bus 0

Figure 8: Probability distribution function of CAIDI of bus 1


Figure 11: Probability distribution function of CAIDI of bus4

595

You might also like