See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/362592355

Heuristic Approach to Ring Main Unit Placement within a Self-Healing

Distribution Network

Conference Paper · August 2022

DOI: 10.1109/PowerAfrica53997.2022.9905271

CITATIONS READS

2 146

3 authors, including:

Chua Liang Su
General Electric
10 PUBLICATIONS 499 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Saudi Electricity Company 500 MW Peak Load Shaving Scheme View project

All content following this page was uploaded by Chua Liang Su on 10 August 2022.

The user has requested enhancement of the downloaded file.

 Heuristic Approach to Ring Main Unit Placement
within a Self-Healing Distribution Network
Ibrahim Adaji Jerry Chua-Liang Su Suresh Gautam
GE Energy Consulting GE Energy Consulting GE Energy Consulting
General Electric Company General Electric Company General Electric Company
Abuja, Nigeria Dubai, United Arab Emirates New York, USA
[email protected] [email protected] [email protected]

Abstract— Active management and protection of automatically as part of a FLISR scheme in a DMS
distribution networks are beginning to take a foothold in utilities framework.
within Africa. In this regard, Fault Location, Isolation, and
Service Restoration (FLISR) scheme is a prevalent smart grid Every utility has varying needs for automation, depending
functionality for enhancing the reliability performance of a on its aspiration for quality of supply, demographics, and
distribution system. At the high level, FLISR facilitates faster budgetary needs. In this regard, feeder automation is a
sectionalization of a distribution feeder during a failure event challenging endeavor as feeders are numerous and are spread
and reliability improvement is achieved by limiting the exposure over large geographic areas, making capital expenditure
to or duration of outage experienced by customers connected to (CAPEX) and operating expense (OPEX) an expensive
the feeder i.e., Self-Healing. This paper introduces a Heuristic proposition. Therefore, feeder automation can be considered a
Method suitable for FLISR scheme implementation: select combinatorial optimization problem [7] and is often limited to
manually operated medium voltage isolation switches are the worst performing feeders. This paper proposes a Heuristic
upgraded to automated and/or tele-controllable Ring Main Method for identification of RMU placement as part of a
Units (RMU). The method assigns a Reliability Value that can FLISR scheme by concentrating on problematic feeders;
be readily calculated for any manual switches under economic effects in terms of constraints on the number of
consideration. A ranking system is then introduced to rank the RMUs that can be deployed will be considered in conjunction
manual switches where ones with the highest Reliability Values
with conventional reliability aspects of a distribution network
are prime candidates for automation. A case study will be
such as failure rate, repair time and switching time of the
presented based on an actual distribution network in a utility in
North Africa, using a predictive reliability model developed in
components. As the method inherently involves sampling of
CYMDIST software to compute and compare the reliability candidate feeders for further in-depth study, it is particularly
performance (SAIDI) of selected feeders with RMUs. The result useful when a distribution network has numerous potential
shows that the reliability performance with RMU is most manual switches for upgrade and/or when reliability and
sensitive to feeders with longest length attributes and least network data of the underlying distribution system lack
sensitive to the feeders with lowest capacity attributes in the accuracy, thus making the RMU placement problem more
network. tractable and less computationally intensive. As utilities spend
less money, the automation investment is more likely to be
Keywords—Power System Reliability, FLISR, RMU, Self- paid off with reduced frequency and duration of feeder
Healing System, Smart Grid, SAIDI outages.
I. INTRODUCTION Section II describes the proposed Heuristic Method for
optimal RMU placement, including the concept of Reliability
Outages in a distribution system can occur when a fuse or
Value which is the core of the proposed method. Numerical
a switchgear operates to clear a fault and downstream
results illustrating the effectiveness of the proposed method
customers lose power. This may be due to the sudden failure
on RMU placement are presented and discussed in Section III
of a component such as an overhead line, transformer and so
and followed by a summary in Section IV.
on. Concerning power switchgear devices [1], the most
suitable solutions for self-healing implementation comprises II. METHODOLOGY
switchgears, reclosers [2], or combined solutions. In a
conventional system, the information about the failure is This section presents the heuristic approach used in
typically relayed to the Distribution Network Operator via achieving the implementation of the FLISR scheme. The
trouble calls from the customers, and patrol crew may detect steps for the Heuristic Method are as follows:
the fault or outage via on-site inspections. As such failure A. Feeder Classification
detections require human involvement, the associated
corrective action will require more time to complete, thus Considering the large number of feeders in a distribution
compromising quality of supply performance. On the other network, feeders must be arranged in groups for easy analyses
hand, a Distribution Management System (DMS) [3] usually and identification necessary for upgrade. Classification of the
involves automation at feeder level which typically entails population of the utility’s feeder circuits is given first priority.
installation of remotely controllable sectionalizing devices This classification should be done based on a relevant set of
such as RMUs along the feeder: when there is a problem with physical and electrical attributes which may include but not
the feeder, data will be relayed to the control center for limited to:
analysis [4]-[6]. Once the problem has been identified, a • Service area characteristics (urban, suburban),
control engineer can remotely operate the switch/RMU to temperature, relevant geographical features
isolate the section causing the problem and restore supply to
either side of the isolated section. This process may be done

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 • Circuit length, topology, construction type industry has defined a set of reliability metrics to quantify
(underground or overhead) such impact. These reliability matrices are statistical
representation of reliability data for a specific system or
• Circuit historical reliability experience and past failure feeder with clearly identified components [10] and can be
rates used to quantify the reliability of overall distribution system.
• Capacity of feeder, loading profile, length of feeder, The following Reliability Metrics [11] are proposed to
and, quantify the reliability of each feeder circuit considering a
year as duration:
• Customer density, breakdown of customer type
(residential, commercial, industrial, and so on.) • System Average Interruption Frequency Index
(SAIFI): The average number of interruptions per year
The classification is intended to serve mainly two that a customer experience
purposes. First, it will put similar feeders in one group
allowing for the selection of sample feeders for model • System Average Interruption Duration Index (SAIDI):
development in a study software. Second, the classification The average number of outage hours or minutes per
will enable extrapolation of the result to other feeders in the year that a customer experience
same group. • Customer Average Interruption Duration Index
B. Feeder Sampling (CAIDI): The average length of an interruption during
a year
Feeders are sampled based on feeder classification
attributes with the use of a percentile. With the Percentile, PR, • Average Service Availability Index (ASAI): The
each attribute of a feeder can be ranked in relation to all fraction of time that a customer has received power
feeders, using the following equation: during a given time period. The index is usually
𝑃𝑅𝛼,𝜑 = 100 ∙ (𝑁𝑏 − 1)/𝑇 (1) expressed as a percentage of the year that the average
customer receives power
Where:
𝛼 = index of feeder attributes A number of feeders from each attribute groups is selected
𝜑 = index of feeders for modelling in a standard software with reliability module.
𝑇 = total number of feeders For the reliability model, data is only needed for components
𝑁𝑏 = number of feeder(s) where attribute value for 𝛼 has rank below that, when they fail or mis-operate, will interrupt the flow of
feeder 𝜑 power, or escalate the interruption [12]. The key pieces of data
critical to the study are shown in the TABLE I. below. Other
The feeder sampling can subsequently be performed for component data such as power transformers and busbars can
each attribute based on its Percentile e.g., top 10%-20% also be incorporated in the model especially if they will result
sample. Alternatively, statistical method such as student’s t- in meaningful effect on the results of reliability assessment.
test can be used to determine a suitable sampling size to meet
a specific level of confidence interval [8]. Historical reliability data for components and circuits will
be derived from several sources prioritized as follows:
C. Reliability Value Calculations
1) Operating/archived data from the utility
The reliability improvement of replacing an existing 2) Country reliability standards data or statistics
manual switch to an RMU can be measured as the number of 3) Reliability data from other utilities in the region
customer faults saved by the device [9]. The improvement
4) Industry data from IEEE, CEA, CIGRE, IEC, JEA and
can be estimated by the number of upstream customers
similar sources
protected by the device and the length of line exposed for
5) Best engineering judgement
fault downstream to the device. This estimated improvement
is termed as Reliability Value (RV) and acts as a proxy to the
potential improvement in reliability realized by upgrade of an The modelling of the selected sampled feeders in a
existing manual switch to an RMU. A higher value of RV for software is for confirmation of the effectiveness of the
Heuristic Method in predicting the manual switches as prime
a switch indicates a better candidate for the upgrade. The RV
for each switch in a feeder can be calculated using the candidates for automation as well as the attributes that give
following expression: better reliability improvement. It is expected that the switch
with the best reliability improvement corresponds to that with
𝑅𝑉𝛿,𝜑 = 𝐶 ∙ 𝐿𝑥 ∙ 𝐹𝜑 (2) the highest ranked RV calculated using (2).

Where:
𝛿 =index of manual switches TABLE I. SUMMARY OF COMPONENT RELIABILITY DATA
C= Number of Customers Upstream to switch 𝛿 Breakers, Fuses,
F= Average faults per unit length of feeder per year Lines and Cables Load Break
Lx= Length of Line Exposed downstream to switch 𝛿 Switches, etc.
Failure rate Failurs/year Failures/years
D. Reliability Assessment Mean time to repair In minutes/hours In minutes/hours
Reliability Assessment will quantify the overall impact of
Mean time to switch N.A In minutes/hours
the equipment reliability, protection system dependability, the
system layout and automation choices on the flow of power
from substation to load through the feeder circuit. The power
 A reliability assessment is run for each sampled feeder
following the steps (a to f) outlined below:
a) Run the reliability analysis without RMUs in the feeder
b) Calculate the feeder level Reliability Metric e.g., using
SAIDI
c) Identify a switch in each feeder based on the highest
Reliability Value
d) Upgrade the selected switch by installing an RMU
e) Re-run the reliability analysis and calculate the feeder
level Reliability Metric values Fig. 1. Simplified schematic of study network
f) Calculate the Reliability Metric improvement B. Assumptions on Feeder Classification and Sampling
E. Extrapolation of Result For purpose of illustration, the 353 feeders were classified
Upgrading each manual switch on a feeder across the into six attributes (and abbreviations) as follows:
distribution network to improve reliability is not a feasible • Highest Feeder Capacity (HFC)
solution.
This extrapolation process is repeated until a target • Lowest Feeder Capacity (LFC)
number of manual switches have been identified which can • Longest Feeder Length (LFL)
stem from the budgetary allowance of the utility. With this
approach, a limited number of automated switches can be • Shortest Feeder Length (SFL)
identified in a straightforward manner and subsequently be • Highest Customer Density (HCD)
deployed across the network thus improving the feeder
reliability as well as the overall network reliability. • Lowest Customer Density (LCD)
These feeder attributes were sampled and ranked based on
III. CASE STUDY
the Percentile using (1). Three random feeders were sampled
The proposed framework for the placement of RMU in a for each of these attributes between 85th to 100th Percentile
distribution network utilizing the Heuristic Method was (i.e., top 15%), which gives a total of 18 unique samples to be
presented in the previous section. In this section, a case study modelled (approximately 5% of all feeders). Note that higher
is presented. number of samples per the guideline described earlier in
A. Study Network Characteristics Section II.B can be used for better prediction of sensitivity;
limited sample size is used here for brevity.
For evaluating the effectiveness of the proposed model, an
electricity distribution company which serves a large city in C. Assumptions on Component Reliability Data
the Northern Africa is used as a case study. The characteristic The failure rate of the components was calculated based
of the network is given in TABLE II. and an overview of the on past failure data. The time to restore upstream customers
schematic is given in Fig. 1. has no effect on the case study being considered, nevertheless,
All the feeders are radially operated: each feeder emanates an appropriate value was used as input to the computation
from a Primary Substation (SS) or a Distribution Point (DP) software.
and ends with an open point(s), as depicted in Fig. 1. For the D. Assumptions on Reliability Value and Usage of
purpose of this paper, a Kiosk is defined as a protection device Spreadsheet
with multiple (normally 3) manual load break switches and
The Reliability Value for each Kiosk can be calculated
one circuit breaker or fuse (for protecting a distribution
using (2). The associated number of customers affected by an
transformer) while a DP is a switch room with more than one
outage component is required to compute the SAIDI, which is
Kiosk. There are a total of 2,921 Kiosks, each of which are
the main Reliability Metric used for this analysis. Since the
considered as a potential candidate for RMU placement in this
number of customers on feeder basis is not readily available,
case study. Each of the MV feeder has at least one Kiosk and
the kVA rating of transformers was used in place of customer
can have up to 22 Kiosks. Further, the open points of all the
number. This is performed by using a conversion factor of
sampled feeders are taken to be static, i.e., they remain at the
0.829 i.e., the Number of Customers in the entire network is
same location and open throughout the study.
divided by the Total kVA Capacity of all the Distribution
TABLE II. CHARACTERISTICS OF THE DISTRIBUTION NETWORK
Transformers (see TABLE II. for individual values), with a
unit of Customer/kVA. The parameter “Length of Line
Item Stats Exposed” in (2) can be calculated based on the total length (in
Number of Regions 10 km) of backbone main line and major branches of the feeder.
Number of Customers 1,188,200 Since an average failure rate is applied to all feeders, the
Number of Bulk Supply Substations 11
variable 𝐹𝜑 in (2) can be omitted from all calculations.
Stemming from the above, (2) can be expressed simply as:
Number of Kiosks 2,921
Number of Distribution Transformers 2,155 𝑅𝑉𝛿,𝜑 = 0.829 ∙ 𝑈 ∙ 𝐷 (3)
Number of Feeders 353 Where:
Total Cable Length in kilometres 2,758 U = Total kVA capacity of transformers upstream to δ
D = Total length of line Exposed downstream to switch δ
Total Distribution Transformer KVA 1,433,000
Equation (3) was then used to calculate the RV of each Kiosk.
Due to the large number of feeders, a template was created in
a spreadsheet to automatically populate the RV as well as rank
each Kiosk in relation to all Kiosks within the same feeder.
The template uses the kVA of the distribution transformers
(upstream to the Kiosk under consideration) and the
downstream feeder length as input. An example of the results
of RV calculated for each Kiosk of a sample feeder using this
template is shown in TABLE III.
E. Study Results
The CYMDIST software was used to model the sampled
feeders following the steps outlined in Section II and the
assumptions described in Section III as relates to the
Reliability Assessment. The results for the average SAIDI for
the selected 6 attributes are shown in TABLE IV. It is evident
from TABLE IV., that SAIDI improvement is most sensitive Fig. 2. SAIDI Performance of sampled feeder: With and Without RMU
to the ‘Longest Feeder Length’ (LFL) and least sensitive to the
‘Lowest Feeder Capacity’ (LFC) attribute. Each feeder Further analysis was carried out on the modelled feeders
represents an attribute group thus the improvement in for the purpose of comparing the SAIDI with respect to the
reliability of the feeder is representative of the group that proposed RV method; this essentially enumerates all possible
feeder belongs to. candidates for upgrade with RMU and checks whether each
candidate Kiosk provides the lowest SAIDI value (highest
TABLE III. RV CALCULATIONS FOR A SAMPLE FEEDER improvement). This method of upgrading each Kiosk with an
RMU is termed “Exhaustive Search” (ES). The comparison of
Feeder Name LFC_#3 RV in relation to the ES method is presented in Fig. 2. The
Feeder Cap. 8,420 kVA followings can be deduced from the Fig. 2:
Feeder Len. 21.305 km
Estimated Cust. 6,981 Nos.
• The upgrade of Kiosks with RMU always improves
Customer Dens. 327.7 Nos./km
SAIDI irrespective of the method deployed to select
the location
Upstr. Cap. Downstr.
Kiosk#
(kVA) Len. (km)
RV Rank
• The estimated SAIDI improvement provided by both
2977 500 20.31 8,416 12 methods align closely for the feeders with the
2859 8,120 1.36 9,155 11 “Shortest Feeder Length” (SFL) and “Highest
3604 7,220 0.61 3,651 14 Customer Density” (HCD) attributes
4311 7,720 0.34 2,144 15
2878 3,400 16.89 47,592 1 • There is some discrepancy on the estimated
2803 3,660 14.09 42,736 3 improvement between the two methods when used
2900 3,760 13.09 40,786 5 on feeders with other attributes. These differences
3904 4,460 12.66 46,790 2 are expected to be “worst case” since the sampling
338 4,760 10.53 41,532 4 was performed near to “top end” of Percentile
1459 5,260 8.03 34,993 6 • Both methods predicted the same Kiosk as the best
1093 5,360 7.70 34,214 7 candidate for upgrade for 50% of the feeders
1094 6,960 3.50 20,194 8
3017 7,760 2.40 15,439 10 It is worth stating that the ES method is limited in its
3462 7,920 0.70 4,596 13 application for RMU placement since it involves modelling
1095 7,820 2.65 17,179 9 each feeder in the network and test running each Kiosk for
upgrade which is time consuming and not feasible for a large
TABLE IV. AVERAGE SAIDI: WITHOUT VS WITH RMU distribution network especially when the network model does
Average SAIDI Average SAIDI Average SAIDI
not exist in appropriate format.
Without RMU With RMU Improvement
Attributes IV. SUMMARY
[Mins/Cust./Yr] [Mins/Cust./Yr] [Mins/Cust./Yr]
(A) (B) (A-B)
This paper has introduced a Heuristic Method suitable for
LFL 166.9 147.6 19.3 placement of RMUs as part of a FLISR scheme. The method
LCD 79.2 71.4 7.7 provided a means of upgrading manual switch to an
automated/tele-controllable one which results in improved
HFC 55.5 51.8 3.7
fault restoration time and subsequently reliability performance
HCD 20.6 19.3 1.3 of a distribution network. The value propositions of the
LFC 17.5 16.5 1.0 method are that it is relatively straightforward to implement
and does not require modelling of the utility’s entire network,
SFL 16.5 14.1 2.4 thus saving on time resources. This was made possible by the
Average 59.4 53.5 5.9 usage of feeder sampling and assigning Reliability Value to
any manual switch of a feeder considered for upgrade. A study
case based on an actual distribution network of a utility in
North Africa was used to test and compare the reliability [4] E. M. Davidson, M. J. Dolan, G. W. Ault and S. D. J. McArthur,
"AuRA-NMS: An autonomous regional active network management
performance of this method to the Exhaustive Search method: system for EDF energy and SP energy networks," IEEE PES General
the comparison did not converge for all the sampled feeders in Meeting, 2010, pp. 1-6, doi: 10.1109/PES.2010.5590045.
predicting the same kiosk as best candidate for upgrade; this [5] J. Heckel, "Smart substation and feeder automation for a smart
limitation is inherently due to the fundamental differences of distribution grid." CIRED 2009-20th International Conference and
both methods in predicting the likelihood of a certain Exhibition on Electricity Distribution-Part 1. IET, 2009.
component to have a failure. Other reasons may include the [6] C. L. Su, “Evaluating the Potential Improvements in Reliability from
the Implementation of AURA-NMS.” Final Report submitted to
usage of Distribution Transformer kVA rating instead of the Engineering and Physical Sciences Research Council on 2 December
actual customer numbers for the estimation of RV. Further, 2008
the study network is characterized by feeders with customers [7] N.S Efacec, H.D Leite, E. Rodrigues and I. Miranda, “Reclosers to
that are sparsely distributed instead of being concentrated in a Self-Healing Schemes in Distribution Networks: A techno-Economic
Assessment ” Conference Paper April 2016; doi:
specific area. Suggestions for further research include testing 10.1109/ENERGYCON.2016. 7513983
such hypotheses in a more rigorous manner, developing [8] C.L. Su, “Analysis of Factors Leading to Load Shedding Using
algorithm that takes both customer number as well as the Correlation Coefficients”. Presented in International Conference on
Distribution Transformer kVA into consideration and Power System Technology (POWERCON 2010), 24-28 October 2010,
incorporating other relevant reliability data that have not been Hangzhou, China.
explicitly considered in this paper. [9] L. Taylor, “Economic Value and Justification for Sectionalization”,
Presentation to IEEE WG on Switching and Overcurrent, Calgary, 28
July 2009
REFERENCES
[10] R.E. Brown, “Electrical Power Distribution Reliability,” Second
[1] IEEE Standard Definitions for Power Switchgear, IEEE Std. Edition, CRC Press, 2009.
C37.1001992,Oct. 1992.
[11] "IEEE Guide for Electric Power Distribution Reliability Indices," in
[2] IEC and IEEE International Standard: High-voltage switchgear and IEEE Std 1366-2012 (Revision of IEEE Std 1366-2003) , vol., no.,
Control gear – Part 111: Automatic circuit reclosers and fault pp.1-43, 31 May 2012, doi: 10.1109/IEEESTD.2012.6209381.
interrupters for alternating current systems up to 38 kV, IEC Std. [12] R. Billington and R. N. Allan, “Reliability evaluation of power
62271-11, IEEE Std. C37.60-2012, Sep. 2012. systems”, Plenum Press, New York and London, 1984
[3] S. Thomas, and J.D. McDonald, “Power system SCADA and smart
grids,” CRC press, 2017.

View publication stats