You are on page 1of 7

A Scenario-based Network Reconfiguration Framework under

Uncertainties
C. Zhang1, Z. Z. Lin1, F. S. Wen*1, J. S. Huang2

1. School of Electrical Engineering, Zhejiang University, Hangzhou, China (*fushuan.wen@gmail.com)


2. School of Electrical Engineering, University of Western Sydney, Sydney, Australia (j.huang@uws.edu.au)

Keywords: power system restoration; network selecting restoration paths could also be guided by a
reconfiguration; risk; scenario analysis; uncertainty betweenness index, a measure on the pivotability of a
complex weighted graph, i.e., an index reflecting the
Abstract importance degree of each edge in network interconnection
[9]. The requirements for unit start-up are considered in [10],
and in [11], the system-partitioning strategy as well as the
The main issues of network reconfiguration after a complete
restoration path and sequence of the nodes in each
blackout are to restore the generating units and to establish a
subsystem are all taken into account in the restoration. A
stable network. This paper presents a framework containing
method for network reconfiguration and the sequence
three objectives of network reconfiguration. Maximizing the
optimization based on goal programming is proposed in [12],
total number of restored generation nodes is the first objective
in which the duration of reconfiguration and the restoration
to be optimized, so as to accelerate the reconfiguration
of units output are taken as the main issues in evaluating the
process. The second objective is to minimize the total
effect of the reconfiguration. In [13], the number of nodes of
charging capacitor of the restoration paths to decrease the
the restarted generators and the benefit from the restored
reactive power generated by the transmission lines. The third
loads are taken as the return of the restoration schemes, and
and final objective is to reduce as much as possible the risks
then the network reconfiguration scheme with the maximum
of the strategy. The uncertainties of transmission lines are
return is searched in the process of the network
also considered in the framework developed, with the use of
reconfiguration. References [7-13] convert multi-objective
the scenario-based method to address the uncertainty. Next,
restoration problem into a single-objective one. The reduced
the well-established differential evolution algorithm and the
model, nevertheless, is inadequate in taking into account the
Dijkstra algorithm are used to seek an optimal strategy
restrictive relationship among the multi-objectives. In [14],
through solving the multi-objective optimization problem.
the fast and elitist non-dominated sorting genetic algorithm
Finally, the New England 10-unit 39-bus power system is
is proposed for solving the multi-objective problem of
employed to demonstrate the proposed method.
network reconfiguration in order to avoid objective
preference.
1 Introduction The above methods, as observed, have all neglected the
uncertainties associated with transmission lines and the
Power system restoration after a complete blackout
consequential impacts on the restoration process as well as
includes three key phases in a row: black-start, network
the final result of network reconfiguration. This paper is to
reconfiguration and load pick-up. The main issues of the address such a challenge. Risks caused by these
network reconfiguration are to restore the generating units
uncertainties are fully considered in the proposed framework.
and establish a stable network first, and then to prepare for
Three objectives are included in the developed network
the comprehensive load pick-up in the following stage [1, 2]. reconfiguration model: maximizing the total number of the
To date, extensive research has been engaged on power
restored generation nodes, minimizing the total capacitance
system restoration from a blackout. In [3], a discrete particle of the path, and minimizing potential risks [15, 16]. The
swarm optimization algorithm is used to seek the optimal well-established scenario-based method is used to solve the
strategy of the network reconfiguration and scale-free
multi-objective optimization model, with the Dijkstra
networks are applied to assess the structure of the network. algorithm and differential evolution algorithms being
In [4-6], a multi-agent approach is used to seek the optimal
applied to find the optimal strategies. The New England
strategy of network reconfiguration. The method is however
10-unit 39-bus power system is served to demonstrate the
infeasible for large-scale power systems. The network essential features of the proposed method.
reconfiguration is formulated in [7] as a nonlinear
optimization problem for the restoration of important loads,
taking the maximum percentage of the total restored load as 2 Multi-objective optimization model of
the objective. In [8], power sources and loads are ranked network reconfiguration problem
quantitatively according to the network cohesion degree, and
The main issues of network reconfiguration are to
the effectiveness of the reconfiguration is assessed by the
restore the generating units and to establish a stable network
efficiency of reconfiguring the network. The optimization of
that is a prerequisite for the following comprehensive load 0”T0”TH (1)
pick-up. In network reconfiguration, it is required to first where T0 is the time when the generator obtains the restart
determine the sequence of node restoration and then that of power.
the paths supplying power to these nodes. However, there For T0”TH, the generator will restart and provide power
are many uncertainties associated with transmission line to the system quickly, which would make a great
restoration, and any failures of the transmission lines in contribution to network reconfiguration. Otherwise, there
connecting to the power system directly influence the could be a significant delay in restarting the generator.
restoration process and the result of network reconfiguration Therefore, it is important to ensure as many as possible
[17]. Should the information of all the transmission lines generators obtaining the restart power within the critical
have been acquired prior to the restoration, the lines causing maximum interval, which is one of the objectives in the
connection failures could be left out in setting up the developed framework. Meanwhile, it should also restore as
strategy for network configuration. It is unable, however, to many generating nodes as possible [22]. If a generator can
acquire the information completely, and some lines should be restarted quickly, it will provide power to the system and
not be excluded due to their importance in network contribute to the network reconfiguration process. So it
reconfiguration, even though they may be subject to could be regarded that the generator receives income. When
uncertainties [18]. A feasible solution is to investigate missing the critical maximum interval, on the contrary, it
potential uncertainties and their impacts, and minimize the will take a long time for a generator to restart and thus has
economic losses. no contribution in the network configuration phase. The
Uncertainties relate to connecting states of transmission income of this generator is considered as zero. Define the
lines and possess discrete features. The scenario analysis income of generator i is Vi. If equation (1) is satisfied, the
method can be employed to deal with this issue, with a generator could be restarted within the critical maximum
scenario representing a result caused by the uncertainties interval and Vi=1; otherwise, Vi=0. In the process of the
[19]. Hence, the underline philosophy of the scenario network reconfiguration, the total income of generators is
analysis method is to convert the uncertainties into a variety calculated by [13]:
of identified scenarios [20]. Though all the scenarios could f1 ¦Vi (2)
occur, only one of them takes place actually. The final i\
optimal solution could be obtained through analyzing where Ȍ is the set of generating nodes. Maximizing f1 is the
different scenarios and accompanied optimal solutions, and objective to restore as many generating nodes as possible.
considering the relative importance of different scenarios By taking into account the uncertainties, equation (2) is
[21]. modified as
Energizing the lines without the loads is one of the
f1 ¦ ( ps ¦ Vi ) (3)
important operations in the process of the power system s: i\
restoration. The success of the line energizing depends on
where s is a scenario of the network reconfiguration; ȍ is the
not only the overvoltage switching capacity, but also the
set of all the scenarios; ps is the weight of scenario s, and
success operation rate of switchers, environment and
weather, to name a few [17]. The uncertainty in network ¦ ps 1 . Because of the uncertainties, the actual condition
s:
reconfiguration is primarily associated with the probability
of successfully energizing transmission lines. Under a could not be acquired exactly. Each possible condition is
certain scenario, by using the scenario analysis method, lines regarded as one scenario, in which, the information of the
that cannot be reconnected are identified whereas all other transmission lines is defined. Maximizing equation (3) is the
lines are deemed to have no problem in the reconnection. first objective to be optimized.
For those important lines, scenarios can be divided based on
the successful rate of connecting the lines. 2.2 The charging capacitor of lines considering
The network reconfiguration strategy should be set so uncertainties
that as many as possible generating units are to be restored The overvoltage is one of the most important problems
for accelerating the reconfiguration process. The strategy in the network reconfiguration phase. The overvoltage is
with less charging capacitor is preferred to prevent against related to the reactive power generated by transmission lines
overvoltage of the transmission lines. Furthermore, risks with no-load or light-load. So, it is reasonable to choose the
under various scenarios should be fully considered to ensure restoration path with less charging capacitor for decreasing
the robustness of the strategy. the reactive power generated in the path. The total charging
capacitor could be represented as:
2.1 Generating units restoration considering
uncertainties f 2 ¦ cLij (4)
Lij )

Generally, the generating units restored in the network where ĭ is the set of restoration paths; Lij is the line from
reconfiguration phase are mainly thermal power generators. node i to j. Minimizing f2 is taken as an objective for
If the generators could get the restart power within the curtailing the total charging capacitor. With uncertainties,
critical maximum interval TH, it would need less time to whether a transmission line could be reconnected
restart. So the time constraint is as follows: successfully is not determined and can vary in different
scenarios. The restoration path is selected from lines that make the selected strategy approach the optimal strategies in
could be successfully reconnected. The selected restoration different scenarios, The min-max problem defined by (7) is
paths are thus different in different scenarios, and so is the thus selected as the third objective of network
charging capacitor. Uncertainties in reconnecting the lines reconfiguration.
results in the uncertainty of in selecting the restoration path.
Based on the consideration of this, equation (4) can be 2.4 The multi-objective optimization framework
modified as:
The multi-objective optimization framework could be
f 2 ¦ ( ps ¦ cLij ) (5)
s: Lij )
acquired by synthesizing the three identified objectives:
­max f
° 1 ¦ ( ps ¦
s: i\
Vi )
2.3 Risk of network reconfiguration strategy °°
Not only the income but also the relative risks are the ®min f 2 ¦ ( ps ¦ cLij ) (8)
° s: Lij )
important factors that need to be reflected in the process of °min f max(T ( s, x )  T ( s, x))
decision making with uncertainties. One strategy to deal °̄ 3
s:
with the risk is to invest the scheme which is more robust, The main difference between the multi-objective
i.e., the scheme that performs well in all scenarios. Many problem and the single objective one is that the solutions of
methods studying robustness of schemes have been the multi-objective problem are not unique. There is a set of
proposed. In [20], the method based on the concept of regret optimal solutions, such as Pareto optimal solutions or
in conforming the robust objective is employed, and the non-dominated solutions. The Pareto optimal solution is
objective of min-max regret is formulated [20, 23]. In a defined as a solution that may be better than the others in
certain scenario, the regret of a strategy is defined as its one objective and be worse in some other objectives at the
distance to the optimal one [20]. The maximal regret of a same time. In other words, it is impossible to optimize part
strategy could be found from all the scenarios, and then, by of the objectives without sacrificing the others. All the
comparing the strategy with the maximal regret against all solutions in the set of Pareto optimal solutions are not
the others, the strategy with minimal regret could be comparable in terms of all the objectives [25].
identified. Here min-max regret objective is employed to The operation constrains, such as the node voltage and
ensure that the finally selected strategy performs well in all the line capacity, should be complied in solving the
scenarios, which is achieved by selecting a strategy that is multi-objective problem. In general, wide allowable ranges
close to the optimal one in different scenarios. Alternative of voltage and line capacity imply a quick restoration in the
objectives could also be adopted. One of them is the distance network reconfiguration phase. When the constraints are
metric objective constructed in accordance with the violated, the optimal power flow program could be used to
preference of decision makers [24]. adjust the output of the generators and/or the load shedding.
A non-optimal strategy and the optimal one in scenario
s are denoted by x and x . The number of scenarios is k, and
3 Algorithms for solving the multi-objective
ɗh (h=1, 2, ···, k) is the set of strategies in scenario h. The
regret of strategy x in scenario s is: optimization framework
regret ( s, x) T ( s, x )  T ( s, x) (6)
3.1 Differential evolution algorithm
where T ( s, x ) max(¦ Vi  P ¦) c Lij ),
xɗh
i\ Lij  Differential evolution (DE) algorithm is a simple and
efficient imitating biological evolution algorithm, which has
T ( s, x) ¦\ V  P ¦) c

i
Lij 
Lij , x  ɗh , and ȝ is the weighted
been successfully employed to solve single- and
multi-objective problems [26]. With the DE algorithm, an
conversion coefficient. The restoration of generators and the initial population is first generated randomly, the crossover
charging capacitor of restoration paths are the two main and mutation operation similar to conventional genetic
factors considered for seeking the optimal strategy in all the algorithms are then used to generate the filial generation,
scenarios. Other factors subject to consideration depend on and finally the selection operation is performed to populate
the preference of decision makers. the new generation. In the DE algorithm, the size of the
population is maintained.
The min-max regret objective is defined as:
min f 3 max(regret ( s, x)) (7) A new individual is generated by applying the mutation
s: operation as given below:
To solve the problem of (7), the optimal strategies in
different scenarios should be sought separately to obtain C Pr1  D ( Pr 2  Pr 3 ) (9)
T ( s, x ) , s  :. For each scenario s, the difference between
where C is the new individual; Pr1, Pr2 and Pr3 are three
all non-optimal strategies and the optimal one is first different individuals of the parent; Į is a parameter in [0.5,
evaluated to seek for the largest value 1]. To increase the diversity of the population, the crossover
max(T ( s, x )  T ( s, x)) . The min-max optimization will operation is employed. A random parameter z in [0,1] is
s:
generated for each element of the parent’s individuals. restoration time by employing the Dijkstra algorithm for
Comparing z with Į, the relevant element of the parent’s each parent’s individual, and calculate the T.
individual is replaced by the element of the new individual if Step4: Generate filial individuals by the mutation and
z is less than Į; otherwise, the relevant element of the crossover operation.
parent’s individuals is maintained. Each filial individual Pr' Step5: Seek the restoration path with the minimal
restoration time by employing the Dijkstra algorithm for
is generated by using such a crossover operation. By
each filial individual, and calculate the T.
comparing the filial and the parent individuals, the parent is
Step6: Compare the T value of the parent’s individual
replaced if the filial is better. In this way, the next population
with the filial individual, and get the next generation by the
is generated.
selection operation.
In the DE algorithm, the selection operation is done by Step7: Repeat Step4, Step5 and Step6 until the iteration
adopting the avarice strategy. Only when the filial individual number is reached, and generate the optimal strategies in
is better than the parent’s one, the parent’s one is replaced. different scenarios.
In the operation of (9), the new individual is generated from Step8: Generate the initial population for the
several individuals of the parent; in other words, each multi-objective problem, and set the calculation
element is generated from relevant elements of other parameters.
individuals. However, this method neglects the influence of Step9: Seek the restoration paths with the minimal
different elements in the same individual. In this paper, the restoration time in different scenarios by employing the
new elements are generated from the other ones in the same Dijkstra algorithm for each parent’s individual, and calculate
individual, so the mutation operation is as: the three objectives in equation (8).
Step10: Generate filial individuals with the mutation
ci xi  D ( x j  xk ) (10) and crossover operation.
Step11: Seek the restoration paths with the minimal
where ci is an element of the new individual; xi, xj and xk are restoration time in different scenarios by employing DE
the different elements of the same individual. The crossover algorithm for each filial individual, and calculate the three
and selection operation are the same as that discussed earlier. objectives in equation (8).
In the process of solving the network reconfiguration Step12: Compare the three objectives of the parent’s
problem, each individual in the population is a strategy individual with the filial individual, get the next generation
which is the restoration sequence of the important nodes. by the selection operation.
The elements of an individual represent the important nodes, Step13: Repeat Step10, Step11 and Step12 until the
and the sequence of the elements represents the restoration iteration number is reached.
sequence of the important nodes. Step14: Compare the individuals of the population with
each other, and get the Pareto optimal solutions.
3.2 Dijkstra algorithm In the above process, the optimal strategies in different
At the beginning, the length da of the starting node a is scenarios should be first found, and it will be used to
set as zero, and the length of other nodes are set as infinity. calculate the robustness of the strategy.
If there is a connecting line from node u to node v, and the
length of the line is w(u, v), supposing du to be the shortest 4 Case study
path from node a to node u, the shortest path from node a to
node v could be found by adding line (u, v) to the end of The New England 10-unit 39-bus power system shown
node u, and the length of the path is du+w(u, v). If the length in Fig. 1 is employed to demonstrate the proposed method.
is less than the previous value dv, the previous one is The black start generator located at node 33. Nodes 30,
replaced by du+w(u, v), which will be the new dv. This 31, 32, 33, 34, 35, 36, 37, 38 and 39 are important
process is repeated until all the dv represent the shortest generating nodes to be restored, and the critical maximum
length from node a to node v [13, 20]. In the process of intervals of the generators are 15, 15, 15, 20, 15, 40, 30 and
solving the network reconfiguration problem, the restoration 20 minutes, respectively. The node number of the important
sequence of the important nodes is determined in each load nodes and the important substation nodes are 4, 6, 8, 15,
strategy, and the Dijkstra algorithm is used to seek the path 16 and 26. Suppose that there is only one generator at each
with the minimal restoration time. generating node. The important lines with uncertainties to be
studied are line 5-6 and 25-26. Four scenarios are identified
accordingly:
3.3 Implementation of the proposed algorithms
Scenario 1: all the lines can be reconnected
The process seeking for the optimal strategy is successfully.
following the steps given below: Scenario 2: only line 5-6 can not be reconnected.
Scenario 3: only line 25-26 can not be reconnected.
Step1: Input data, and set the calculation parameters, Scenario 4: both line 5-6 and line 25-26 can not be
such as the iteration number and the size of population. reconnected.
Step2: Generate the initial population. The parameters of the lines are shown in Appendix A.
Step3: Seek the restoration path with the minimal The calculation parameters are set as follows:
1) For seeking the optimal strategies in different line 33-19, 19-20 and 20-34, and the initial charged region is
scenarios, the number of the population size is set shown as Fig. 1. In this phase, not much load is restored due
as 100, and the maximum number of iteration is set to the limitation of load increasing rates of generators and
as 100. the other operation constraints. It is assumed that all the
2) For seeking the solution of the multi-objective generators could obtain enough power to restart.
problem, the number of the population size is set as The optimal strategies for four scenarios are shown in
100, the maximum number of iteration is set as Table 1. The restoration paths of the optimal strategies are
100. shown in Table 2. The results of the multi-objective
3) The parameter of Į in the mutation operation is set optimization are shown in Table 3.
as 0.7. The T ( s, x ) value of Scenario 1 is better than
4) The weighted conversion coefficient ȝ is set as 1. Scenario 2 and Scenario 3 as shown in Table 1, because all
5) The Ps of four scenarios are set as 0.5, 0.2, 0.2 and the lines can be reconnected successfully in Scenario1 while
0.1. there is one line that cannot be reconnected in Scenario2 and
Scenario3, with less number of lines available in the
restoration. In Table 2, the optimal restoration path of
Scenario 1 selects line 5-6 and 25-26. Line 5-6 can not be
reconnected in Scenario 2, so other lines are selected to
replace line 5-6. However, line 25-26 is still in the
restoration path of Scenario 2. In Scenario 3, line 25-26 can
not be reconnected, so line 27-26 is selected to restore node
26 with line 5-6 staying in the restoration path. In Table 1,
the T ( s, x ) value of Scenario 2 is worse than Scenario 3,
showing that line 5-6 is more important than line 25-26 in
the network reconfiguration phase. Failure in reconnecting
line 5-6, as a result, will bring about more losses in the
restoration phase. Scenario 4 is the worst scenario because
both line 5-6 and 25-26 can not be reconnected, which might
cause even more economic losses.
The Pareto optimal strategies are acquired as shown in
Table 3. The three strategies given in Table 3 are Pareto
optimal, which means that the improvement of one objective
is achieved by taking the cost of sacrificing the others. The
final strategy can be selected from the set of Pareto optimal
Fig.1: The New England 10-unit 39-bus power system solutions in the next step. The preference of the decision
makers or other factors can also be considered in the process
The strategy represents the restoration sequence of of selecting the final strategy, or by considering only one
important nodes, such as generating nodes, important load objective. Compared with single objective optimal methods,
nodes and important substation nodes. The Dijkstra the developed framework provides more options for decision
algorithm is employed to seek the path with the minimal makers in selecting the final network reconfiguration
restoration time for each strategy. Before the network strategy.
reconfiguration phase, the generator located at node 34 has
been restarted by the generator located at node 33 through

Number of the
Optimal strategies
generators restarted Charging
Scenarios (restoration sequence of important T ( s, x )
within critical capacitor/P.U.
nodes)
maximum interval
Scenario 1 16,39,36,30,35,26,38,4,6,15,31,8,32,37 6.0 7.5313 -1.5313
Scenario 2 16,30,39,36,35,15,26,38,4,8,6,31,32,37 6.0 7.6789 -1.6789
Scenario 3 16,36,30,35,39,4,26,38,8,37,15,6,31,32 6.0 7.5795 -1.5795
Scenario 4 16,30,36,35,39,4,15,26,38,8,6,31,37,32 6.0 7.7271 -1.7271
Table 1: Optimal strategies for the four scenarios
Scenarios Restoration paths of the optimal strategies (Restoration sequence of nodes)
(19),16,17,18,3,2,1,39,(16),24,23,36,(2),30,(23),22,35,(2),25,26,29,38,(3),4,5,6,(16),15,(6
Scenario 1
),31,(5),8,(6),11,10,32,(25),37
(19),16,17,18,3,2,30,(2),1,39,(16),24,23,36,(23),22,35,(16),15,(2),25,26,29,38,(3),4,5,8,7,
Scenario 2
6,31,(6),11,10,32,(25),37
(19),16,24,23,36,(16),17,18,3,2,30,(23),22,35,(2),1,39,(3),4,(17),27,26,29,38,(4),5,8,(2),2
Scenario 3
5,37,(16),15,(5),6,31,(6),11,10,32
(19),16,17,18,3,2,30,(16),24,23,36,(23),22,35,(2),1,39,(3),4,(16),15,(17),27,26,29,38,(4),5
Scenario 4
,8,7,6,31,(2),25,37,(6),11,10,32
Table 2: Restoration paths of the optimal strategies
Note: the bold nodes are the important nodes; the nodes in the brackets are the nodes restored in the previous
steps; and the next node is restored by these nodes

Number of the generators Charging


Robustness
Pareto optimal strategies restarted within critical capacitor
(f3)
maximum interval (f1) (f2)/P.U
Strategy 1 16,36,30,35,39,15,26,38,4,32,8,37,6,31 6.0 7.7616 0.1716
Strategy 2 16,36,39,30,35,15,37,4,26,38,8,6,32,31 5.7 7.5901 1.000
Strategy 3 16,4,30,36,15,35,26,38,8,6,31,37,39,32 5.0 7.5900 1.000
Table 3: Results of the multi-objective optimization

Power Systems, 2, pp. 271-277, (1987).


5 Conclusion [3] Y. Liu, X. Gu. “Skeleton-network reconfiguration
based on topological characteristics of scale-free
In this paper, uncertainties associated with networks and discrete particle Swarm optimization”,
transmission lines are considered in a multi-objective IEEE Trans on Power Systems, 22, pp. 1267-1274,
optimization framework for power grid reconfiguration. (2007).
Three objectives are included in the proposed model to [4] T. Nagata, H. Watanabe, M. Ohno, et al. “A
maximize the total number of the restored generation nodes, multi-agent approach to power system restoration”,
minimize the total capacitance of the path, and reduce the IEEE Trans on Power Systems, 17, pp. 452-457,
associated risks of the strategy to the least level. The (2002).
well-established scenario-based method is used to solve the [5] J. Yen, Y. Yan, J. Contreras, et al. “Multi-agent
multi-objective optimization model, and the differential approach to the planning of power transmission
evolution and the Dijkstra algorithm are employed to find expansion”, Decision Support System, 28, pp.
the optimal strategies. Finally, the New England 10-unit 279-290, (2000).
39-bus power system is served for demonstrating the [6] D. Liu, Y. Chen, G. Shen, et al. “A multi-agent based
essential features of the proposed method. The framework approach for modeling and simulation of bulk power
presented in the paper provides a new way for optimizing system restoration”. // Proceedings of 2005 IEEE/PES
the multi-objective network reconfiguration strategy. Transmission & Distribution Conference &
Exposition: Asia and Pacific, Dalian, China, 2005.
Acknowledgements [7] Z. Wei, Y. Liu, X. Gu. “DPSO algorithm based
network reconfiguration of power systems for
This work is jointly supported by National High maximizing load recovery efficiency”, Automation of
Technology Research and Develop Program (863 Program) Electric Power Systems, 31, pp. 38-42, (2007).
(2011AA05A105), National Natural Science Foundation of [8] Y. Liu, X. Gu. “Node importance assessment based
China (51007080), the Fundamental Research Funds for skeleton network reconfiguration”, Proceedings of the
the Central Universities (2012QNA4011), and a key project CSEE, 27, pp. 20-27, (2007).
from Zhejiang Power Company. [9] Z.Z. Lin, F.S. Wen. “A new optimization method for
determining restoration paths based on weighted
References complex network model”, Automation of Electric
Power Systems, 33, pp. 11-15, (2009).
[1] Y. Xue, S. Fei, F. Bu. “Upgrading the blackout [10] Z. Han, X. Gu, Y. Liu. “Optimization of restoration
defense scheme against extreme disasters: Part II paths considering unit start-up time requirements at
tasks and prospects”, Automation of Electric Power early stage of power system restoration”, Proceedings
Systems, 32, pp. 1-5, (2008). of the CSEE, 29, pp. 21-26, (2009).
[2] M. M. Adibi, P. Clelland, L. Fink, et al. “Power [11] X. Gu, Z. Han, H. Liang. “Optimization of parallel
system restoration-a task force report”, IEEE Trans on restoration through power system partitioning after
blackout”, Proceedings of the CSEE, 29, pp. 8-15, The starting The end The charging The restoration
(2009). node of the node of capacitor of the operation time of
[12] Y. Liu, Q. Gao, X. Gu. “Optimization of restoration line the line line/P.U the line/min
sequence of network based on goal programming”, 2 1 0.6987 1
Automation of Electric Power Systems, 34, pp. 33-37, 39 1 0.75 1
(2010). 3 2 0.2572 2
[13] S.Q. Zeng, F.S. Wen, Y.X. Xue, et al. “Optimization
25 2 0.146 1
of network reconfiguration strategy for power systems
considering operating time uncertainty”, Automation 4 3 0.2214 1
of Electric Power Systems, 35, pp. 16-21, (2011). 18 3 0.2138 1
[14] H. Wang, Y. Liu. “Multi-objective optimization of 5 4 0.1342 1
power system reconstruction based on NSGA-II”, 14 4 0.1382 1
Automation of Electric Power Systems, 33, pp. 14-18, 6 5 0.0434 1
(2009). 8 5 0.1476 1
[15] W. J. Burke, H. M. Merrill, F. C. Scheppe, et al. 7 6 0.113 1
“Trade off methods in system planning”, IEEE Trans
11 6 0.1389 2
on Power Systems, 3, pp. 1284-1290, (1988).
8 7 0.078 1
[16] C. J. Andrews. “Evaluating risk management
strategies in resource planning”, IEEE Trans on 9 8 0.3804 1
Power Systems, 10, pp. 420-426, (1995). 39 9 1.2 2
[17] Y. Liu, G. Liu, X. Gu. “Line risk assessment in 11 10 0.0729 1
transmission network restoration schemes”, 13 10 0.0729 1
Automation of Electric Power Systems, 35, pp. 12-16, 14 13 0.1723 2
(2011). 15 14 0.366 2
[18] G. Liu, Y. Liu, X. Gu. “Identification of critical lines 16 15 0.171 2
in restoration scheme for transmission network”,
17 16 0.1342 1
Automation of Electric Power Systems, 35, pp. 23-28,
(2011). 19 16 0.304 2
[19] V. Miranda, L. M. Proenca. “Why risk analysis 21 16 0.2548 2
outperforms probabilistic choice as the effective 24 16 0.068 2
decision support paradigm for power system 18 17 0.1319 2
planning”, IEEE Trans on Power Systems, 13, pp. 27 17 0.3216 2
643-648, (1998). 22 21 0.2565 1
[20] P. Maghouli, S. H. Hosseini, M. O. Buygi, et al. “A 23 22 0.1846 2
scenario-based multi-objective model for multi-stage 24 23 0.361 1
transmission expansion planning”, IEEE Trans on
26 25 0.513 2
Power Systems, 26, pp. 470-478, (2011).
[21] R. T. Rockafellar, R. J. B. Wets. “Scenarios and policy 27 26 0.2396 2
aggregation in optimization under uncertainty”, 28 26 0.7802 2
Mathematics of Operations Research, 16, pp. 119-147, 29 26 1.029 1
(1991). 29 28 0.249 1
[22] Y. Xue. “Space-time cooperative framework for 12 11 0.249 1
defending blackouts (Part III): optimization and 12 13 0.249 2
coordination of defense-lines”, Automation of Electric 6 31 0.1319 1
Power Systems, 30, pp. 1-10, (2006). 10 32 0.3216 1
[23] M. O. Buygi, H. M. Shanechi, G. Balzer, et al.
19 33 0.2565 1
“Network planning in unbundled power system”,
IEEE Trans on Power System, 21, pp. 1379-1387, 20 34 0.1846 1
(2006). 22 35 0.361 2
[24] P. Maghouli, S. H. Hosseimi, M. O. Buygi, et al. “A 23 36 0.249 1
multi-objective framework for transmission expansion 25 37 0.249 2
planning in deregulated environments”, IEEE Trans 2 30 0.249 1
on Power Systems, 24, pp. 1051-1061, (2009). 29 38 0.249 2
[25] T. Xie, H. Chen, L. Kang. “Evolutionary algorithms 19 20 0.249 2
of multi-objective optimization problems”, Chinese
Journal of Computers, 26, pp. 997-1003, (2003).
[26] H. Meng, X. Zhang, S. Liu. “A differential evolution
based on double population for constrained
multi-objective optimization problem”, Chinese
Journal of Computers, 31, pp. 228-235, (2008).

Appendix A Parameters of the lines of the New


England system

You might also like