Distribution System Restoration With Renewable

Resources for Reliability Improvement Under

System Uncertainties
PRIYANSHU GUPTA ABHINAVRAJ P K
November 8, 2023

1 Introduction
Integration of renewable distributed generation units, such as solar photovoltaic
systems and wind turbine generators , into existing power systems because they
can offer several benefits to power systems, including support during normal
operation and enhanced reliability during contingencies. However, system op-
erating constraints, uncertainties in generation availability for renewable DG
systems, and time-varying load demand are major barriers in their effective
integration.
The reliability assessment of distribution systems embedded with renewable
DG sources has been carried out giving importance to system uncertainties(such
as variation of load demand,wind speed,and solar radiation) and optimal the
systems restoration plan.. The optimization of restoration processes involves
minimising customer interruption duration, which can improve customer-side
supply reliability by reducing the customer interruption duration of each fault
event, thus reducing the system average interruption duration index (SAIDI).
Conventional distribution systems typically balance supply capacity with
maximum load demand to meet customer needs without considering the need
for DG unit support.Constant load demand and DG output are often used
for short-term restoration, but this can mask system uncertainties and cause
unrealistic solutions, resulting in fewer customers being restored or additional
interruptions due to mismatch between actual generation and load demand.

2 Mathematical Formulation for The Restora-

tion Process
The associated function f, which will result in optimal switch configuration, can
be expressed as
XT X N
f = min (1 − Bt,n )Ln
t=1 n=1

1
The total interruption hours due to a system fault event is denoted by T.The
total number of system nodes in out-of-service areas is denoted by N.The binary
status of load point n at time t is represented by Bt,n, with 1 representing the
restored load point.The number of customers at load point n is denoted by Ln.
The total number of switches in the out-of-service areas can be represented by
M. SW is a binary matrix representing the status of switches in an M x T format
at each time step
system Constraints to be considered are modelled as :
loadpoint, also known as the child node, may only be restored in conjunction
with the restoration of the related candidate node, also known as the parent
node, and the closure of the switch between the loa dpoint and the candidate
node. The constraint mentioned above can be written as
Bj,t ≤ SWj,t Bi,t
J and I are the child and parent nodes respectively. where SWj,t is the binary
state of the jth switch at time t. Bj,t and Bi,t represent the binary status of
the jth and ith node( restored or not restored).
A DG unit’s total loads restored shouldn’t be greater than its capacity at
any given time step (t).
X
PDGt ≥ Bt,n Pt,n
nϵßt

Pt,n (kW) is the required capacity at load point n, including its associated loss
component in time t. And Bt,n is the status of load point n at time t.
The minimum time tmin required to operate the corresponding manual
switches shall not be less than the time difference between two consecutive
manual switching operations:
SWm,t1 = SWm,t2 ∀ |t1 − t2 | < tmin
It is anticipated that the load point that was reenergized earlier won’t be
de-energized later on.This constraint can be modelled using binary variables as
Bt,n ≥ Bt−1,n

3 Encoding Process For Optimization

If a binary encoding strategy is applied, the restoration region under examina-
tion, comprising M switches with an interruption period of T hours, will have
a solution space of 2M T potential possibilities. However, because of the lim-
itations of the problem, only a limited number of combinations may produce
workable solutions.This article proposes a dynamic selection technique as an
alternative to the conventional binary encoding scheme, in which it is possible
to translate the problem variables from the matrix SW into a vector with two
sets of integer variables.
X = { {x1 ...., xi ..., xT } , {xT +1 , ..., xT +N }}

2
 The choice to either restore or not restore new customers is represented by
T integers in the first set. N numbers in the second set represent the choice
of the potential restoration nodes. After the initial issue variables are trans-
formed into a vector containing T + N integers, the numbers may be translated
into decisions that have real significance by using the suggested transformation
function provided as
 
xi
yi = 1 + xi − f loor · ai
ai
The number of candidate nodes, or alternative possibilities, is denoted by ai ,
and the priority index of the nodes that need to be restored is indicated by yi .
Xi is the ith integer input from the set X.

3.1 Selection Procedure from first part of set X

It is clear from the initial set of vectors X1 that there will always be two options
( ai = 2 ∀i = 1,...,T) related to the decision of whether or not to restore more
customers.Using equation given above and ai = 2, it is possible to determine
that yi will always correspond to the value of xi in the first instance by either
1 or 2. Where yi = 1 denotes restoration of additional customers, and yi = 2
means maintaining current system status.

3.2 Selection Procedure from second part of set X

The solution method selects eight numbers at random to create the second set
of the vector X2 = [5 8 10 12 15 18 28 77]. To restore the nodes consecutively,
these integers must be dynamically translated to a selection choice.It can be
seen that node 1 will be the first candidate node if the DG starts restoring the
interrupted area. By applying the equation shown above with x = 5 (i.e., first
element of the second set, X2) and a = 1, the solution y = 1 can be established.
Likewise ,node 2 can be added to the restoration sequence list. Subsequently,
nodes 3, 5, and 7 are the potential candidate nodes that can be added to the
restoration sequence list ranked based on the higher to lower load demand. The
relevant load points will be added to the candidate node list and the node itself
will be removed from the list once it is added to the restoration sequence list.
Ultimately, using the provided numbers, a unique restoration sequence of [1 2 7
3 4 8 5 6] may be produced; the matching selection guide is [1 1 2 1 1 1 1 1].

Figure 1: Example feeder with two DG

3
3.3 Selection Guide for a Multi Generator System
When there are more than one DG, the suggested dynamic selection method
is used on each one to find a unique restoration process. If more than one
restoration sequence overlaps, the restoration sequence is chosen by treating
these DG units as a single unit and picking the largest DG node as a potential
node.

4 CONCLUSION
In this article, we have discussed the different system constraints which are re-
quired to be taken into consideration for reliability assessment of a power system,
and how to model those constraints using variables and equations. Secondly ,
we have undergone the study of optimization algorithms for system restoration
after a failure , to ensure lesser customer interruption and to ensure the already
restored networks are not interrupted again. A more intelligent TRIBE PSO
method is used in the paper that we have referred to , to optimise the restoration
procedure. For reliability assessment under different scenarios a time sequential
Monte Carlo methods were used and the reliability indices SAIDI and SAIFI
were used to show the customer side reliability .