0% found this document useful (0 votes)
143 views11 pages

A New Approach To Advanced Analytics in Utility Asset Management

Uploaded by

ahconsultancy5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
143 views11 pages

A New Approach To Advanced Analytics in Utility Asset Management

Uploaded by

ahconsultancy5
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 11

Electric Power & Natural Gas Practice

A new approach to
advanced analytics in
utility asset management
Studying how one North American transmission and distribution
utility’s implementation of advanced analytics in asset management
can help other organizations embark on similar journeys.

This article is a collaborative effort by Anjan Asthana, Pranjal Dubey, Alfonso Encinas, Anand Mohanrangan,
Aditya Pande, Luis Fernando Rios Siliceo, Jesús Rodríguez González, and Willem van Schalkwyk,
representing views from McKinsey’s Electric Power & Natural Gas Practice.

© Ed Freeman/Getty Images

September 2022
Asset management can account for a significant UtilityCo: An overview
percentage of a transmission and distribution UtilityCo faced a number of key challenges that
(T&D) company’s operating expenses and capital are common in the industry. For example, the
expenditures, with optimized operations and utility didn’t take a risk-based approach when
investments key to generating savings. New making asset replacement decisions or prioritizing
technologies can enable companies to capture preventive-maintenance activities, and it had
these efficiencies. In fact, a recent McKinsey decentralized asset management operations, with
article explained how T&D utilities can leverage each operating company taking a distinct approach
advanced analytics in their asset management and methodology. In addition, although UtilityCo
strategies to unlock 10 to 20 percent in savings was able to collect valuable data, the data were
while improving overall reliability and performance underused and stored in multiple systems. Finally,
of their networks.1 UtilityCo relied on rules that oversimplified asset
management decisions—for example, the “three
This article builds on that thinking and takes a close strikes” rule, which called for replacing cables after
look at a North American T&D utility, which we they experienced three outages.
refer to as UtilityCo. In 2021, UtilityCo leveraged
advanced analytics in asset management to unlock
savings of 20 to 25 percent in operating expenses The results from advanced analytics
and 40 to 60 percent in capital expenditures, UtilityCo was able to effectively use advanced
which could then flow as savings into the profit- analytics in asset management in four ways. First, it
and-loss (P&L) statement or be reinvested to optimized capital expenditures either by maintaining
deliver significant reliability improvement. These current risk and spending less—and letting the
savings and increased investment capacity are excess capital expenditures flow into the P&L or
particularly relevant given today’s increasing be reinvested to deliver more reliability—or by
constraints, including pressure from customers spending the same amount and achieving higher
on affordability, inflation growth, supply chain reliability through replacing the riskiest assets.
bottlenecks, and the growing need for investments Second, it lowered preventive-maintenance (PM)
in the energy transition, such as renewable-energy operating expenses by optimizing PM activities.
solutions, electric-vehicle charging infrastructure, When successful, this optimization can deliver
and cybersecurity. Based on the success of the similar or better reliability at lower cost. Third, it
initial model, UtilityCo developed a road map lowered corrective-maintenance (CM) operating
that scales the asset management risk–based expenses by lowering spending on CM after those
approach to two-thirds of the capital portfolio over riskiest assets had been replaced. And fourth, it
two years. replaced the riskiest assets to help achieve higher
reliability (measured as lower SAIDI and SAIFI2
The following case study highlights the results performance) due to fewer failures.
of implementing advanced analytics at UtilityCo,
including the approach taken, the lessons Regarding capital expenditures for UtilityCo’s
learned, and the best practices to adopt for others transmission transformers, the company underwent
embarking on a similar journey. Although this a paradigm shift, collecting data about each dollar’s
article is presented as a stand-alone example, our impact on interrupted customer minutes. With this
experience shows that the results from applying new perspective, UtilityCo determined it could
advanced analytics to asset management are reduce risk approximately two to three times over
accelerated when deployed as part of a broader while spending the same amount, maintaining the
organizational transformation. same capital expenditures, and reducing customer

1
Rui de Sousa, David González Fernández, Jesús Rodríguez González, and Humayun Tai, “Harnessing the power of advanced analytics in
transmission and distribution asset management,” McKinsey, April 9, 2018.
2
System average interruption duration index and system average interruption frequency index.

2 A new approach to advanced analytics in utility asset management


interruptions. Alternatively, UtilityCo could maintain level of risk as determined by the current plan while
the same level of risk as determined by the current spending 20 to 25 percent less on PM.
plan while spending 40 to 60 percent less, thus
creating capital headroom for reinvestment, For the underground-cables asset class, UtilityCo
maintaining the same level of customer interruptions was able to avoid up to 70 percent more outages
but reducing capital expenditures (Exhibit 1). as compared with the baseline by replacing its
Another option was to select a pathway that both riskiest cables. The optimization model also gave
reduced risk and required less spending. UtilityCo the flexibility to achieve either higher
reliability at current spending levels or P&L savings
On operating expenses, UtilityCo had the option of at current reliability.
spending the same amount on PM—removing 1.5 to
2.0 times the level of risk from the system compared Finally, UtilityCo developed a visualization platform
with the current baseline—or maintaining the same that displayed the results produced from advanced

Web <2022>
<EPNG UtilityX>
Exhibit
Exhibit <1>1 of <7>

The
The savings
savings potential
potential for UtilityCo’s
UtilityCo’s transmission
transmission transformers
transformers improved
by
by 40
40 to 60 percent
to 60 percent for
for capital
capitalexpenditures
expendituresand
andbyby 20
20to
to25
25percent
percentfor
for
operating expenses.
operating expenses.
Baseline and optimized capital expenditure Baseline and optimized operating expenditure
spend, $ millions preventative maintenance (PM) spend,
$ millions
Capital expenditure scenarios Operating High and medium health or criticality
expenditure Low health or criticality
scenarios

15 15 –40–60% –20–25%

1.0
1.2
1.0
7

0.5
0.3 0.2

Baseline A: Constant B: Reduced Baseline A: Constant B: Reduced


capital capital operating operating
expenditure, expenditure, expenditure, expenditure,
reduced risk constant risk reduced risk constant risk

Options for unlocking value: Options for unlocking value:


A: Reduce risk1 by 2–3× compared to baseline while A: Remove 1.5–2× more risk1 compared to operating
keeping capital expenditure spend constant expenditure baseline while keeping operating
B: Alternatively, reduce capital expenditure by expenditure spend constant
40–60% while maintaining same risk reduction as B: Alternatively, reduce operating expenditure by
in baseline 20–25% while maintaining same risk reduction as
in baseline

1
Risk is a calculation of health score times criticality.
Source: McKinsey UtilityX analysis

A new approach to advanced analytics in utility asset management 3


analytics (see sidebar “Unlocking value as part of a The approach
broader transformation”). This dashboard allowed UtilityCo’s advanced analytics–led approach
UtilityCo to visualize, prioritize, and implement new entailed leveraging both internal and external
maintenance activities (Exhibit 2). asset data to calculate a health score (the
probability of failure) and criticality (the cost of
failure) of a given asset. From there, it used the
health score and criticality to estimate asset risk

Web <2022>
<EPNG UtilityX>
Exhibit 2of <7>
Exhibit <2>

A visualization
visualizationplatform
platformallowed
allowedUtilityCo
UtilityCototovisualize,
visualize,prioritize,
prioritize,and
and
implement new maintenance activities.
implement new maintenance activities.

Clickable dashboards
provide high
usability and
responsiveness

Maintenance-
actions dashboard
Risk map allows provides an
users to review overview of fleet
the risk level for maintenance and
each region and repair activities
identify geographic
patterns

Fleetwide charts provide a Deep dives on single assets with


single, high-level overview on the details on asset-specific indicators (eg,
distribution of health, criticality, transformer gas analysis) help inform
and risk decision making at the asset level

Unlocking value as part of a broader transformation

UtilityCo implemented the asset manageable. Second, cross-functional needed to fill new technical roles, such as
management advanced-analytics teams were already working on various data scientists and engineers, and upgrade
solution as part of a broader business initiatives—for example, the IT and its IT infrastructure via cloud migration and
transformation. While the broader transmission and distribution (T&D) teams machine-learning platforms, among other
transformation wasn’t essential to were focused on workforce management— options. This was easier to do as part of
implement the advanced-analytics solution, and were thus able to build on these the transformation because UtilityCo was
it accelerated the adoption process. First, relationships to drive the advanced- already in the process of building a digital
UtilityCo had already adopted a change analytics initiative. Finally, to successfully center of excellence, which was able to
mindset, so the implementation of a new scale asset management to other assets manage these new functions.
asset management strategy was more and operating companies, UtilityCo

4 A new approach to advanced analytics in utility asset management


and prioritize asset replacement and maintenance the concept of “lead” versus “follower” (Exhibit 4).
activities based on risk (Exhibit 3). For example, it ensured that poles (lead) were
modeled before cross arms (follower).
Early on, UtilityCo had a clear plan to scale
advanced analytics across all its assets and Health score: Estimating the probability
operating companies. It prioritized assets based on of failure
impact, including operating expenses and capital Depending on the asset, UtilityCo considered more
expenditures; time to impact; and feasibility, such than 100 variables to estimate the probability of
as quantity and quality of data and the technical failure. A machine-learning model was trained on
difficulty of building models. In addition, when internal data (such as the age of the asset, work
sequencing distribution assets, UtilityCo followed orders, and failure history) as well as on external

Web <2022>
<EPNG UtilityX>
Exhibit 3of <7>
Exhibit <3>

UtilityCo’s
UtilityCo’s approach
approach for leveraging advanced analytics was based on the
based on the
probability and cost of failure as well as on building an optimization engine.
probability and cost of failure as well as on building an optimization engine.
Description Outputs

1 Calculated the probability


of failure for each asset
Probability of
failure for each
Decision-making risked-based model

Compute based on the following: asset in the Least critical/ Most critical/
probability of • historic outage data following year most healthy least healthy
failure (health) • structural data (eg, age
and manufacturer)
• environmental data
(eg, soil temperature and
water volume)

Criticality
2 Calculated the cost of
failure for each asset based
Cost of failure
for each
of asset Proposed
shift in
Define impact on the following: asset for the spending
of failure • repair following year
(criticality) • service
• safety
• environmental impact Health of asset

3 Grouped clusters of assets


to optimize operational
Clusters
of assets Maintenance optimization

High-level efficiency based on


Operating expenditures
clustering for business rules:
the asset • eg, grouped assets for
replacement at the Capital expenditures
protective-device level

Risk/reliability
4 Generated multiple
scenarios to optimize the
Optimized list
of assets to be
Optimize schedule for the inspection replaced over the Long-term decision support
replacements or replacement of assets next year and
and inspection based on certain constraints; prioritized project-
eg, geography and age management
activities

A new approach to advanced analytics in utility asset management 5


Web <2022>
<EPNG UtilityX>
Exhibit 4of <7>
Exhibit <4>

UtilityCo’s
UtilityCo’s sequence for distribution
sequence for distribution assets wasbased
assets was basedon
onimpact
impactand
andfeasibility.
feasibility.

HIGH

Cables and wires


(eg, underground
cables, wires)

Pole-top
equipment (eg,
poles, cross arms,
line transformers

Feasibility
Switches and
fuses (eg, overhead
and underground
switches, cutouts
and fuses)

Multi-assets
(eg, reclosers,
cross-asset
optimization)
LOW
LOW Impact HIGH

data (such as weather data, which stretched back number. Next, the data were divided into a training
a few years). A holdout data set3 was used to test set, which was used to train the machine-learning
the model performance. For example, when looking model, and a test set, which was used to help test
at the transmission transformers asset class, the the performance of the model after training (for
model was able to predict approximately 45 percent example, testing how often the model correctly
of failures in approximately 20 percent of the data predicted asset failure). Because UtilityCo was
(Exhibit 5). interested in going beyond a standard machine-
learning algorithm, it also incorporated a failure-
To build the health model, UtilityCo aggregated data mode analysis into the transmission-transformer
from several different systems, such as geographic models to support detailed assessments of
information and outage-management systems. probability of failure by component and to help with
The utility then cleaned and unified the data in the prioritization of condition-based maintenance.
preparation for the machine-learning model and
identified prediction targets. In some cases, the Finally, UtilityCo combined the outcomes from
process was straightforward, such as labeling a a previous engineering health model with the
transformer that had suffered an outage. In other machine-learning model to calculate the probability
cases, it was more difficult, such as when labeling of failure and improve performance.
a failed cable that was missing a serial or part

3
Holdout data, or test data, are historical, labeled data used for validating machine-learning models.

6 A new approach to advanced analytics in utility asset management


Web <2022>
<EPNG UtilityX>
Exhibit <5> of <7>
Exhibit 5
Estimating the probability of failure for transmission transformers relies on
Estimating the probability of failure for transmission transformers relies on
key input data, important variables, and sample output.
key input data, important variables, and sample output.
Step one: Step two: Variables and Step three: Sample output. Targeting the
Key input data their relevance to the 20 percent of transmission transformers with the
health index of target, % worst predicted health (bottom two deciles) can
capture approximately 45 percent of outages

Health status of transmissions transformers


Capture, clean, and
process data by decile, % of total outages within given
decile (n = 89)

Gas tests
49.6
(eg, DGA1) Bottom decile
(predicted 26.1
worst health)
Structural data
(eg, age,
manufacturer) Corrective 18.8
maintenance 30.8

10.1

Failures (eg, type Transformer age 11.0


of failure, date)
13.0

Preventative 8.7
maintenance 7.0

External data
(eg, weather)
18×
7.2
High-side voltage
1.4
8.7

Manufacturer 2.9
> 1.0

2.9
Network
configuration Top decile
0.1 (predicted 1.4
best health)

Dissolved gas analysis.


1

Source: McKinsey UtilityX analysis

A new approach to advanced analytics in utility asset management 7


Criticality: Estimating the cost of failure wasn’t necessary to cluster the replacement for
UtilityCo estimated the cost of failure across transformers. Typically, simple business rules—such
several dimensions, including repair, service, safety, as clustering cables to be repaired based on the
and the environment (Exhibit 6). Repair costs are protective device they are connected to—can be
those related to bringing the asset back online used to cluster assets.
after a failure, service costs are estimated based
on lost revenue and other factors related to the Building an optimization engine
“importance” of the customer (for example, a hospital UtilityCo used an optimization engine to prioritize
is considered “more important” than a single asset replacement and PM activities based on the
household), and safety and environmental costs are risk of the asset. To build the optimization engine,
dependent on location and asset type. UtilityCo first estimated the risk of each asset by
multiplying its health score by its criticality. Asset
Depending on the asset, utilities will need to group replacement was then prioritized based on the
the replacement of assets for operational efficiency. resulting risk score and the asset’s replacement
For example, UtilityCo found that grouping the cost. For example, a high-risk transformer with a
replacement of underground cables and applying lower replacement cost was prioritized over the
operational constraints made sense, but it same type of transformer with a higher cost. To

Web <2022>
<EPNG UtilityX>
Exhibit <6> of <7>
Exhibit 6
Key
Key dimensions are used
dimensions are used in
in calculating
calculating the
the cost
cost of
offailure
failure for
for aa typical
typical asset.
asset.
Type of impact Key indicators

Lost revenue due to outages

Customer-service costs

Service Cost based on types of customer, such as hospitals versus homes

Area type (crowded, residential)

Wildfire risk area


Environment
Cost of failure
(criticality)

Population density

Installation location
Safety

Length of cable

Number of phases
Repairs

8 A new approach to advanced analytics in utility asset management


To optimize PM, a detailed failure-mode
analysis was incorporated into the
optimization engine to enable estimates
of how much risk was removed by each
PM activity.

optimize PM, a detailed failure-mode analysis incomplete or duplicated. For example, data from
was incorporated into the optimization engine to one asset class were missing installation dates. As
enable estimates of how much risk was removed a workaround, the manufacturing date was used
by each PM activity. In addition, the optimization instead. Developing data architecture and putting
engine factored in the cost required to perform processes in place to capture and perform quality
each of these activities and prioritized activities that control checks on the right types of data were key to
reduced the most risk at given costs. addressing this issue going forward.

Although the artificial-intelligence and machine-


Lessons learned and the recipe learning space is still emerging, it is growing quickly.
for success The data scientists and engineers who are key
Implementing an advanced analytics–led asset to building solutions are scarce. Thus, UtilityCo
management program resulted in several lessons built a digital center of excellence and leveraged
learned, particularly with regard to getting started, it to manage the pipeline of talent and to develop
data quality, talent and capabilities, change processes and trainings to drive consistency across
management, and implementation and governance. the organization.

First, the team faced internal resistance when During the advanced-analytics implementation,
getting started, including concerns about not having UtilityCo asset managers were asked to make
enough data, or the right data, to address regulatory changes to their management processes (see
considerations. The first key step to addressing this sidebar “Driving additional value from cross-asset
resistance was to implement a proof of concept, optimization”). The key to addressing this issue was
identifying assets that had good enough data to get to engage the asset managers early on and bring
started and developing a solution that was better them along as the solution was being developed.
than the current state. Success with the proof of
concept gave UtilityCo the confidence to proceed Finally, incorporating advanced analytics into the
with rolling out the solution across multiple assets processes for selecting assets for replacement
and operating companies. and updating maintenance processes and
policies based on model recommendation was
Another lesson involved data that were either siloed, an implementation and governance challenge.
scattered across several different systems, or Engaging subject matter experts (SMEs) early and

A new approach to advanced analytics in utility asset management 9


Driving additional value from cross-asset optimization

One of the key advantages of asset Because risk measurement is the same more on one asset over another based on
analytics is cross-asset optimization. across all assets, the risk-based approach estimated system average interruption
Once UtilityCo implemented advanced makes it easier to decide where to duration index (SAIDI) improvement
analytics for multiple assets, it was able to invest the next million dollars of capital per million dollars spent, resulting in
take a risk-based approach that allowed expenditures to achieve company- approximately 94 percent improvement to
replacement capital expenditures to wide objectives. For example, with the reliability (exhibit).
be invested in the assets that reduced objective of improving reliability, UtilityCo
risk and increased reliability the most. determined it was better off spending

Web <2022>
<EPNG UtilityX>
Exhibit
Exhibit <Sidebar> of <7>

Everything
Everythingbeing
beingequal,
equal,investing
investingreplacement
replacement capital expenditures in asset
asset 11 instead of asset 1 will lead to an improvement
11 instead of asset 1 will lead to an improvement in reliability
in reliability of
of approximately
approximately
94 percent. 94 percent.

Estimated SAIDI1 improvement, $ millions/SAIDI-minute improved2

7.5

6.2

4.6
–94%

2.9

2.1
1.8 1.7 1.6 1.6

0.9
0.5
Asset Asset Asset Asset Asset Asset Asset Asset Asset Asset Asset
1 2 3 4 5 6 7 8 9 10 11

Note: Data sanitized by taking 90 percent of the original value.


System average interruption duration index.
1

Numbers have been rounded.


2

Source: McKinsey UtilityX analysis

10 A new approach to advanced analytics in utility asset management


running pilots to test new processes gave UtilityCo impact than to get everything perfect (or build
confidence in its models’ abilities to meet its needs. a big data lake) before starting. For example,
UtilityCo started with proof of concept for two
Several key ingredients contributed to the success assets to demonstrate the value before scaling
of UtilityCo’s asset management transformation: to more assets.

— Top-down leadership buy-in and push. — Change management. Asset managers


Because implementing such a solution usually typically have years of experience managing
involves changes in processes across different assets and follow specific procedures
departments, entailing new technologies, and policies. As a result, UtilityCo found it
UtilityCo discovered that leadership buy-in and difficult to convince its managers to take
push from the top was critical for advanced- recommendations from machine-learning
analytics adoption. For example, the vice models. For example, when it came to replacing
presidents of T&D and IT joined biweekly sprint underground cables, UtilityCo asset managers
reviews to take stock of progress, encourage were accustomed to the three-strike rule—in
the team, and communicate the importance of which a cable is replaced after three failures
the effort. within a 24-month period—and it was a difficult
task to convince them to replace a cable that
— Agile approach to working. By setting one- to had a high probability of failure but no previous
two-week sprints and working collaboratively to failures. One successful strategy was to involve
achieve the goals for each sprint, the team had the team from the beginning and bring them
time to cocreate, bring along team members, along in the process of building the models.
and course-correct as needed without falling
behind. As part of this, a cross-functional
team of business and IT was key to making the
implementation successful. For example, in a UtilityCo transformed its asset management
joint meeting with IT and business, a subject strategy from a manual process that used
matter expert was able to provide guidance on very limited data to make asset replacement
which corrective work orders should be included decisions and a preventative, one-size-fits-all
in the data used to calculate the probability of maintenance approach into a strategy that
failure for a given transmission transformer. leverages extensive data and advanced analytics
to make replacement and repair decisions
— Not letting perfection be the enemy of based on asset risk. In making this change,
getting started. It was more important for UtilityCo unlocked significant savings in capital
UtilityCo to get started with the data that were expenditures and operating expenses while
available and demonstrate some economic increasing reliability for customers and regulators.

Anjan Asthana is a senior partner in McKinsey’s Miami office; Pranjal Dubey is an engagement manager in the Bay Area
office, where Anand Mohanrangan and Aditya Pande are partners and Willem van Schalkwyk is an associate partner;
Alfonso Encinas is a partner in the Washington, DC, office; Luis Fernando Rios Siliceo is a consultant in the Monterrey office;
and Jesús Rodríguez González is a senior partner in the Madrid office.

Designed by McKinsey Global Publishing


Copyright © 2022 McKinsey & Company. All rights reserved.

A new approach to advanced analytics in utility asset management 11

You might also like