Report Study on Emerging Technologies Organization - Duke Energy

Description
Duke Energy has been involved in data mining its smart grid system and device data for the last two years. One major issue that emerged through this work was extracting data from the various siloes it resided in for analysis and new value development.

Final Report
Duke Energy
Emerging Technologies Organization

Data Modeling and Analytics Initiative

April 2014

Data Modeling and Analytics Final Report
Duke Energy Data Modeling and Analytics Initiative
Table of Contents
Table of Contents
List of Tables
List of Figures
Executive Summary
Section 1 OVERVIEW OF THE DATA MODELING AND ANALYTICS INITIATIVE 1-1
1.1 Initiative Overview ............................................................................................... 1-1
1.2 Dataset Overview ................................................................................................. 1-2
1.3 Participation Opportunities during the Initiative .................................................... 1-3
1.4 Information Included in Participant Responses ..................................................... 1-4
1.5 Vendors Invited to the DMAI ............................................................................... 1-4
1.6 Vendor Responses Overview ................................................................................ 1-4
1.7 Data Quality Issues ............................................................................................... 1-5
Section 2 USE CASES OVERVIEW AND VALUE ESTIMATION .................................. 2-1
2.1 Vendor Responses Overview ................................................................................ 2-1
2.2 Observations and Insights ..................................................................................... 2-2
Section 3 TOOLS, SYSTEMS AND ANALYTICS USED .................................................. 3-1
3.1 Overview .............................................................................................................. 3-1
3.2 Data Analytics Tools used by Vendors.................................................................. 3-1
3.3 Analytics Used by Vendors ................................................................................... 3-4
3.4 Insights and Observations ..................................................................................... 3-4
Section 4 RECOMMENDATIONS AND NEXT STEPS ..................................................... 4-1
4.1 Next Steps ............................................................................................................ 4-1

List of Appendices
A List of Use Cases

Table of Contents
ii Duke Energy Data Modeling and Analytics Final Report 6/16/14
List of Tables

Table 1-1 DMAI Core Team Members .................................................................................... 1-2
Table 1-3 Data Quality Issues.................................................................................................. 1-5
Table 2-1 Value Estimates for Use Case Categories ................................................................ 2-3
Table 2-2 Use Case Value-Consolidated Responses ................................................................ 2-4
Table 3-1 Summary List of Tools by Big Data Component ..................................................... 3-2

List of Figures

Figure 1-1. Use Cases by Category and Vendor Coverage ....................................................... 1-5

Data Modeling and Analytics Final Report
EXECUTIVE SUMMARY
Duke Energy has been involved in data mining its smart grid system and device data for the last
two years. One major issue that emerged through this work was extracting data from the various
siloes it resided in for analysis and new value development. Therefore Duke Energy developed a
“Sandbox” – a data model and dataset that combined a finite set of data elements from a variety
of systems – to analyze and identify new value opportunities from utilizing the smart grid
system.
To accelerate this process, Duke Energy sponsored the Data Modeling and Analytics Initiative
(DMAI), an innovative forum by which big data experts were given a slice of the dataset to
analyze for opportunities and insights that it could incorporate into its big data analytics strategy
and activities.
Seventeen vendors submitted final reports that discussed data issues, models, and tools used to
analyze the data, and use cases that could be developed for new value opportunities. Responses
varied considerably based on the skills and expertise of each vendor. Vendors provided over
150 unique use cases for consideration. Duke Energy asked for general financial information
regarding the potential benefits of the use cases; however, these results were limited. Therefore,
a follow-up exercise and interview was implemented that allowed the vendors to provide value
scores to major use case categories, and provided qualitative input as to where the value may be
discovered within each category.
Significant insights were generated from information on how vendors constructed and applied
systems and analytics to develop and model the use cases. This information can be incorporated
into Duke Energy’s big data and analytics strategy activities. Below are key observations and
insights that came out of the DMAI:

There is significant potential value that can come from implementing a big data platform
across a variety of use cases areas.

We encountered a variety of problems extracting data from Duke Energy’s systems.
Consolidation and integration of data elements are required to perform the analytics necessary
to identify and realize the value identified above. Issues with data include missing data, no
common information model, problems linking data sets from different systems, and
challenges from extracting data.

Vendors had few problems ingesting Duke Energy data. While the systems and tools for data
ingestion varied among vendors, the results for ingestion across vendor platforms was
generally consistent.

To understand big data implications for other areas of Duke Energy, more data is needed than
what was included in the initial data set. The vendors provided numerous examples of use
cases that could be implemented with the inclusion of additional data. Social media, asset
attributes (age, type) and event alerts were the three most common data elements most often
identified by vendors.

EXECUTIVE SUMMARY
ES-2 Duke Energy Data Modeling and Analytics Final Report 6/16/14

The initiative provided significant insights into the different tools and systems used to
manage big data. Development of technical and functional specifications, along with
development of overall solution architecture should be Duke Energy’s first priorities upon
finalizing its strategy.

Many new models and analytics were introduced to Duke Energy by the vendors in their final
reports. Furthermore, they demonstrated the use of these models using the Duke Energy
dataset.

Vendor responses and capabilities typically fell into three categories: (1) industry experience,
(2) analytics experience, and (3) IT systems. No one vendor has all three areas completely
covered.

There are resource and skill gaps between what is available at Duke Energy today versus the
potential capabilities identified from the vendor final reports. The most important issues to
address are: availability of data, a comprehensive analytics strategy, and overcoming the silo-
based structure of our data and systems.

The following are recommendations and next steps for Duke Energy to implement in 2014:

Expand the list of collaborative vendors beyond the initial participants, as they continue to
provide valuable insights and feedback with respect to their offerings and experiences. Meet
with them individually to discuss the technical and analytic results of their research.

Continue work on refining and forecasting the potential value of having a big data and
analytics platform at Duke Energy. Prioritize the use case categories to pursue, identify the
top use cases, and develop a base case forecast for realizing the value.

Utilize the results of the IT systems review to start development of technical and functional
specifications that will become part of Sandbox 2.0 and the big data platform.

Present final results to Analytics Leadership Team and make the resources, research, and
documentation available to the team for input into their 2014 activities.

Prepare an big data analytics organizational structure, resource, and skills gap analysis to
identify areas that Duke Energy may want to supplement with future employees or external
resources. Develop job descriptions for the key positions.

Address the issue that consolidation and integration of data elements are required to perform
the analytics necessary to identify and realize the value identified above. Issues with data
include missing data, no common information model, problems linking data sets from
different systems, and challenges from extracting data.

Data Modeling and Analytics Final Report
Section 1
OVERVIEW OF THE DATA MODELING AND ANALYTICS INITIATIVE
1.1 Initiative Overview
Duke Energy has undertaken a series of data mining and analytics projects to reduce costs and
improve operational efficiencies by monitoring, collecting, and cleansing data from its smart grid
test area. As a result, these inter-related projects were launched to construct a data-rich
environment of sensors, implement a dedicated data warehouse (known colloquially as “the
Sandbox”), mine the data for new insights, create new analytical tools, and develop a
comprehensive “big data” strategy for the corporation.
The Sandbox is a stable IT environment used to collect a variety of data beyond metering-based
information from legacy and new systems, as it ensures development activities do not disturb
production systems. These coordinated activities capture a wide range of data from the smart
grid that can be translated into a variety of new, value-creating activities that reduce costs such
as operations and maintenance as well as capital expenses, improve reliability, and anticipate
hardware and software requirements for future technology development. Additionally, this
project mapped each of the traditional and non-traditional indicators of distribution grid health,
reliability, and connectivity; which can provide insights on how to correlate, aggregate, and scale
data elements more effectively across all operational systems. This process was extremely
complex and required inter-departmental collaboration across a variety of traditional utility
“silos.”
The first project that utilized this data was the Duke Energy Modeling and Analytics Initiative
(DMAI). After internal collaboration, we invited qualified analytics vendors to participate in the
initiative, and gave them the opportunity to analyze a cross section of data and provide
Duke Energy with insights and recommendations for further analysis. We believed that there was
significant value in working with vendors on this initiative as it would help Duke Energy
accelerate its big data analytics process and capabilities. In summary, this project would be used
to identify and quantify the value to Duke of a variety of use cases at each stage of a smart grid
rollout. This would also allow us to quantify the value of these projects to the company,
prioritize the work, and determine the value for building out smart grid-enabled infrastructure.
The DMAI was undertaken in a collaborative spirit, with several departments within Duke
participating. Weekly update calls were held, and the team was involved in all aspects of the
process, from selection of vendors to participate, to review and sign off of the data elements and
files that would be given to the participants, to ongoing review of information coming in during
the Initiative. Table 1-1 below lists the members of the core team.

Section 1
1-2 Duke Energy Data Modeling and Analytics Final Report
Table 1-1
DMAI Core Team Members
Department
Product and Program Development
Metering Services
Enterprise IT
Grid Modernization
Customer Architecture and Data Management
Emerging Technologies
Market Research/Customer Analytics
Information Management
Environmental Services
IT Architecture and Security
Environmental Services
Emerging Technologies
Emerging Technologies
Revenue Services
Emerging Technologies
Load Forecasting
Emerging Technologies
Emerging Technologies
Grid Modernization
Emerging Technologies
Supply Chain
Grid Modernization
Grid Modernization
Technology Planning, Projects and Reporting
1.2 Dataset Overview
Each participating vendor received a one-week comprehensive dataset of anonymized
Duke Energy customer, weather, and grid-related data to mine for value and to demonstrate
analytical capabilities to turn the data into valuable insights. The goal of the DMAI was to use
data to identify, prioritize, and utilize common use cases that determine the analytics value from
smart grid technology deployment.
Upon completing several information security data requests and agreements, vendors received
access to a secure FTP site to download files for analysis. Below is a summary of what they
received:

Dataset: Information in the Sandbox consists of data from:

AMI systems

Transformers

OVERVIEW OF THE DATA MODELING AND ANALYTICS INITIATIVE
Data Modeling and Analytics Final Report Duke Energy 1-3

Distribution sensors

Distribution grid devices (capacitor banks, reclosers)

Outage Management Systems

Smart grid communications nodes

Weather stations

Billing systems

Socioeconomic databases
For the Data Mining and Analytics Initiative, Duke Energy compiled a dataset of one week’s
worth of data – during the summer coincident peak of 2012. This was comprised of eighteen
(18) different files of data in either the CSV or XLS format, which allowed the vendors to
import information into their own structural format, data warehouse, or platform. We were
interested in learning in general about the vendors’ experiences with data ingestion and model
structure, as Duke Energy is investigating different ways it will manage its big data in the
future.

Data Element Definitions: This file contains a list of all the files that comprise the above
dataset, along with a list of the data elements within each file. Moreover, the data elements
included a short description or definition.

Data Map: This file contains a map of the major data elements as they were set up in the
Duke Energy “Sandbox.” This allowed vendors to understand how each of the files may be
connected with each other.
Vendors were able to access this information via a secure FTP site once they completed and
returned the Data Security Agreement.
1.3 Participation Opportunities during the Initiative
There were several opportunities for participation and feedback during the DMAI. First, vendors
had the ability to submit questions, clarifications, or issues encountered using the Power
Advocate tool. Duke Energy made every effort to address these as quickly as possible. All
questions and answers were made available to the rest of the vendors. This was to help create a
collaborative learning process between Duke Energy and the participants as it learned more
about how to manage and analyze big data sets.
Duke Energy also hosted two conference calls/webinars during the DMAI. These were used to
answer questions and clarify information that may be distributed throughout the Initiative.
Duke Energy also used this time to present information on its analytics activities, and to
introduce new ideas or use cases for consideration and discussion.
Duke Energy learned a significant lesson with the conference calls. The first conference call was
held jointly with all the participating vendors; over 70 people were in attendance. Even though
we stressed that this was to be a collaborative initiative, vendors were reluctant to discuss issues
and questions in front of their potential competitors. Therefore, the second conference calls were
a series of one-on-one calls with each vendor team. This represented a significant investment of
time on Duke Energy’s part. However, it yielded substantial results in terms of feedback and

Section 1
1-4 Duke Energy Data Modeling and Analytics Final Report
participation from the vendors. Coincident with the calls was a request by Duke Energy for
vendors to provide an interim report on the progress being made on data analytics and use case
development. The interim reports gave us insights into what was being considered, allowed Duke
to provide input to each vendor process, and gave us a preview of what we could expect to
receive at the end of the project.
All responses submitted to Duke Energy at the end of the Initiative are considered confidential
information. However, Duke Energy is aggregating ideas and common themes for use in
analyses, presentations, or papers that may be made public. For example, it recently presented a
paper at the 2014 Distributech conference on the DMAI.
1.4 Information Included in Participant Responses
Duke Energy was looking to accelerate its big data analytics activities, and it hoped the DMAI
would enable this process. As a result, we invited a diverse mix of participants to help us
understand the major issues associated with big data analytics. The Duke Energy dataset was the
starting point for participants to learn more about what type of information may be available at
Duke Energy and other utilities. Therefore, we were very interested in not only analytic
solutions, but also insights vendors might have with respect to data mining and analytics.
Examples of information asked for inclusion into the responses were:

New use case ideas

Examples of analytics to investigate further

Insights into what other types of data should be collected

Data warehouse, models and structures

Applications for modeling

Data extraction and ingestion insights

Insights and lessons learned from your data mining and analytics experiences
The vendors provided over 150 unique use case ideas. These are discussed in more detail in
Section 3, and are included in detail in Appendix A.
1.5 Vendors Invited to the DMAI
The DMAI was an invitation-only event. The names of the participants were provided by
members of the core team, and represent a diverse set of experiences and skills. Although
twenty-eight vendors were invited to participate, seventeen completed the DMAI by submitting
final reports.
1.6 Vendor Responses Overview
Figure 1-1 summarizes the number of use cases submitted by the vendors. For each category it
shows the number of use cases (with duplicates omitted), along with the number of vendors that
provided use cases in this category (the total vendors = 17). This figure illustrates how “popular”
the category might be with the vendors. For examples, most of the vendors provided use cases in

OVERVIEW OF THE DATA MODELING AND ANALYTICS INITIATIVE
Data Modeling and Analytics Final Report Duke Energy 1-5
the Meter Analytics and Customer Analysis categories. In short, Duke Energy has many vendor
options to select when conducting further investigations into these categories. Conversely,
although there were many use cases in the Distribution Grid Analysis category, only half the
vendors provided examples of these. Selection of vendors for future discussions may therefore
depend on what use case categories are given the highest priority.

Figure 1-1. Use Cases by Category and Vendor Coverage
1.7 Data Quality Issues
The DMAI represented an opportunity to receive external feedback with respect to the new
dataset developed from the integration of data from various systems. Duke Energy wanted to use
the Initiative to understand what issues may have arisen from developing models and analytics
with the data. Most vendors provided insights throughout their final reports. Table 1-3 below
contains examples of the major data issues identified by the vendors.
Table 1-3
Data Quality Issues
Data Quality Issue Description
Lack of data history More historical data would allow for better model results. At least two years of
data is required.
No meter event information was provided. This
is critical for event processing analytics.
In fact, all asset event information is important, especially if there are ways to
link the events across the different assets using key variables like timestamps,
GPS location, etc.
Network model and line GPS coordinates Provide connectivity information between devices.
0
5
10
15
20
25
30
Use Cases
Vendor Coverage

Section 1
1-6 Duke Energy Data Modeling and Analytics Final Report
Data Quality Issue Description
Line sensor device/line information missing Line sensor dataset did not have Device IDs or associated Circuit IDs.
Therefore, you cannot relate these to other devices or events (meters,
reclosers, outage events, etc.).
No Substation or Circuit SCADA Data This information could be compared with sum of meter readings for a way to
calculate losses.
No customer billing information Need rate and billing information to understand how analytics may segment or
predict customer usage based on price.
No Capacitor Banks and Recloser location
information
This is needed to be able to relate events stored in Pi Historian with
communication node, line sensor and meter events.
Missing/Null GPS coordinates for transformers This is needed to help identify asset locations.
Asset attribute data was not provided In order to do asset failure analysis, asset information (type, age, etc.) is
required.
Social media datasets were missing These could provide significant customer sentiment information that could be
used for segmentation and targeting opportunities. Twitter feeds are the most
common example cited.
No firmographic information for non-residential
customers
For non-residential customers, information on square footage, number of
employees, and NAICS (North American Industrial Classification System) or
SIC (Standard Industrial Classification) code would also be useful to aid
development of load forecasting models as well as segments of customers.
One-minute granularity data for line sensor
events not enough for fault detection
Seconds or milliseconds are required; or set up Sandbox 2.0 to export data
directly from the line sensor system.
Line sensor dataset did not include circuit ID
and threshold value of the current data
collected
As a result, they could not be used to verify/correlate outage information.
Transformer ratings (kVa) were not included This is a key piece of data - representing the size of the transformer – is
required for monitoring, overload and failure analysis.
Socioeconomic dataset not complete Additional demographic data would make analytics more insightful. Substantial
number of records did not have this.
Customer/Transformer relationship No customers mapped to several transformers, yet these transformers have
load on them
Limited outage information available for review The Outage_Info file contained a limited set of events that were confirmed in
the AMI data. However, that AMI data revealed much more extensive and
complete outage information when analyzed on its own

Data Modeling and Analytics Final Report
Section 2
USE CASES OVERVIEW AND VALUE ESTIMATION
2.1 Vendor Responses Overview
Duke Energy received over 150 use cases as part of the DMAI. These were grouped into major
categories for additional analysis and prioritization. As part of the DMAI, we asked the vendors
to provide business case value estimates for the use cases they provided. Whereas the vendors
spent a significant amount of effort developing and analyzing use cases, limited quantifiable
financial information was provided on the benefits. As the core team reviewed the vendor final
reports, this issue was raised again. Given the significant number of use cases that were
provided, Duke Energy wanted some initial indication of value it could use to help it develop
short-term prioritization of activities.
Therefore, Duke Energy asked the participants for insights as to where it should refine its focus
to extract value from the use cases through advanced analytics. Duke Energy provided a table
with the different use case categories, along with some examples of the use cases that were
provided in the final report. Duke Energy then asked the participants to provide an estimate of
what the value may be from pursuing use cases within the category. Value could come in the
form of reduced operational expenses; increased revenues from existing or new product or
service ideas; or protection of existing revenues from increased reliability, customer satisfaction,
theft reduction, etc.
Understanding that providing a financial metric would be difficult, we asked instead for them to
use one of the codes below:

Value Code Value Estimate
$ Thousands in value
$$ Millions in value
$$$ Tens of millions in value
$$$$ Hundreds of millions in value
$$$$$ Billions in value

To discourage guessing, we stressed that we were looking for the vendor experiences and
insights with providing analytics under each category; if they did not have experience in a
particular category, they were to just put an N/A there.
The value estimates were based on a wide range of assumptions. Some vendors developed “value
potential” estimates. These represented the lifetime value of the use case. Lifetimes ranged from
5 to 20 years. Others developed a value per year estimate. We will be working to convert the
estimates to a common denominator for use in planning and prioritization of 2014 activities.
Table 2-1 contains the value estimates by Vendor. Table 2-2 consolidates all the comments
received by the vendors regarding their value estimates.

Section 2
2-2 Duke Energy Data Modeling and Analytics Final Report
2.2 Observations and Insights
The following insights came from review of the vendor use cases and value estimate responses:

Over 150 unique use cases were provided by the participating vendors; these were grouped
into 12 macro categories for value analysis and prioritization.

The average score for each category ranged from 2.6 to 3.6. While this was expected, what is
more interesting is that for most categories, value scores varied widely. All categories had the
highest score (5 - $$$$$) by at least one vendor.

Time should be spent reviewing individual vendor scores and comments, to gain a better
understanding of what motivated their responses.

In every category, at least one financial metric or “rule of thumb” was given for the basis for
value estimation. These need to be reviewed and standardized before a consolidated business
case is developed.

Many examples of value cited come from applications that cross more than one use case
category. Therefore, as estimates are refined, care should be taken to ensure we are not
double-counting the value of a particular use case. Some examples of cross value include:

Smart Grid Monitoring and Analysis Systems: these incorporate meter and
communications infrastructure data and analytics

Program Development: incorporates customer analytics, energy efficiency and demand
response information, and potential benefits

Load Forecasting: could be considered a building block for other categories, such as
demand response, energy efficiency, or distribution grid analysis

Significant potential value may come from the incorporation of social media, socioeconomic,
and interval data to segment new customers and identify new revenue opportunities.
Duke Energy has significant new revenue goals over the next five years; knowing its
customers on a more granular and segmented level will be required to meet this goal.

Table 2-1
Value Estimates for Use Case Categories
Use Case Category

V
e
n
d
o
r

1

V
e
n
d
o
r

2

V
e
n
d
o
r

3

V
e
n
d
o
r

4

V
e
n
d
o
r

5

V
e
n
d
o
r

6

V
e
n
d
o
r

7

V
e
n
d
o
r

8

V
e
n
d
o
r

9

V
e
n
d
o
r

1
0

V
e
n
d
o
r

1
1

V
e
n
d
o
r

1
2

V
e
n
d
o
r

1
3

V
e
n
d
o
r

1
4

V
e
n
d
o
r

1
5

V
e
n
d
o
r

1
6

V
e
n
d
o
r

1
7

A
v
e
r
a
g
e

S
t
a
n
d
a
r
d

D
e
v
i
a
t
i
o
n

#

V
e
n
d
o
r

R
e
s
p
o
n
s
e
s

Meter Analytics N/A 3 4 1 3 2 4 2 3 2 3 2 3 5 3 3 3 2.9 0.927 16
Customer Analysis N/A 4 4 3 3 2 3 5 4 3 4 2 4 2 2 2 1 3.0 1.061 16
Distribution Grid Analytics N/A 3 4 N/A N/A 3 4 5 4 3 N/A 3 2 2 4 3 N/A 3.3 0.850 12
Outage Analysis N/A 3 3 1 4 3 3 5 3 N/A N/A 2 3 2 3 2 N/A 2.8 0.948 13
Revenue Protection N/A 5 3 2 3 3 3 5 4 N/A N/A 2 3 5 4 1 4 3.4 1.172 14
Demand Response N/A 3 4 2 N/A 2 3 5 3 3 3 3 3 4 2 2 3 3.0 0.816 15
Asset Management Analysis N/A 3 3 2 N/A 4 5 5 3 N/A N/A 3 2 2 5 1 3 3.2 1.231 13
Load Forecasting N/A 4 3 1 3 3 5 N/A 4 N/A N/A 3 4 5 2 2 4 3.3 1.136 13
Distributed Generation N/A 2 3 2 N/A 2 3 5 3 N/A N/A 2 N/A N/A 3 N/A N/A 2.8 0.916 9
Energy Efficiency N/A 3 4 2 3 2 5 2 3 3 3 2 3 N/A N/A 2 3 2.9 0.833 14
Communications N/A 3 3 N/A N/A 3 2 N/A N/A N/A N/A 2 N/A 2 N/A 3 N/A 2.6 0.495 7
Security N/A 3 N/A N/A N/A 4 4 N/A N/A N/A N/A 3 N/A 4 N/A N/A N/A 3.6 0.490 5

Value Estimate Value Code
Thousands in value $ - 1
Millions in value $$ - 2
Tens of millions in value $$$ - 3
Hundreds of millions in value $$$$ - 4
Billions in value $$$$$ - 5

S
e
c
t
i
o
n

2

2
-
4

D
u
k
e

E
n
e
r
g
y

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

Table 2-2
Use Case Value-Consolidated Responses
Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Meter Analytics
Meter event analysis
Meter interval usage outlier analysis
Meter operations monitoring
Efficiency benefits though optimizing work order deployment and identifying and analyzing problems that may
suggest infrastructure operations issues
Analyzing operations data, identifying outliers, and responding to operations events will increase operational
efficiency
Better operational response time to identify and repair the estimated 1-2% of meter failure events that occur each
year
1-2% savings ($20/customer) from billing system improvements that utilize meter data
Significant value overlap with this category and Communications when utilizing data via a smart grid monitoring and
analysis system (SGMAS) to operation the smart grid more efficiently
Customer Analysis
Segment customers for utility program
participation using socioeconomic data
Customer load pattern analysis
Develop rate class profiles for demand and
voltage
Profiles from AMI data will support more exact rate design calculations and will also support more detailed rates. This
has significant revenue enhancement and cost reduction potential.
AMI data will allow more careful targeting of program marketing, which will lead to higher program participation or
lower program costs or both.
Customer targeting and segmentation analytics can lead to major improvements in customer satisfaction.
$2/meter/year opportunity. Huge opportunity to look at customers the way on-line advertisers do.
Social media data, when combined with socioeconomic and interval usage data, will allow Duke new visibility into
customer actions and sentiments.
Value may vary across service territories depending on the value Duke Energy places on increasing customer
satisfaction scores.
Customer and segmentation analysis can lead to new program and product ideas, which are new sources of value.

U
S
E

C
A
S
E
S

O
V
E
R
V
I
E
W

A
N
D

V
A
L
U
E

E
S
T
I
M
A
T
I
O
N

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

D
u
k
e

E
n
e
r
g
y

2
-
5

Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Distribution Grid Analysis
Feeder/phase load prediction, monitoring and
management
Identify and manage voltage variations by
device along the feeder
Implement condition-based vegetation
management
Customer Volt/Var applications could save up to 3% of energy savings with optimization analytics.
Typical Volt/Var applications save 0.5-1.5% of energy savings, or around $100M a year for the complete Duke
Energy service territory.
Reduction of vegetation management costs through optimization and forecasting of operations. Annual vegetation
management spending is 1% of revenues; savings could be up to 20% of this.
Another vendor estimates this at $0.5/meter/year, although the value may be farther out.
Understanding device activities, operations, and events along the entire feeder can provide optimization value of that
feeder, from a voltage and loading perspective.
Load balancing will help enhance asset lives.
Value can come from using asset analysis to deploy and/or enhance condition-based maintenance (CBM) programs.
Visualization applications will aid operators to make quicker and better decisions to reduce technical losses and
increase operations efficiency.
Reduction in capital budgeting costs is achievable through better asset management.
Development of locational marginal pricing (LMP) forecasts will provide significant value if Duke Energy joins an RTO
in the South.
Outage Analysis
Reduce SAIFI and CAIDI through faster
restoration management
Automate outage notification and restoration
documentation process
Identify and resolve momentary outage events
Customer attrition may be affected by outages. This could be in retail competition states like Ohio, or by the
increased installation of DG and microgrids ($$$$).
Using analytics to respond faster to outages will lead to reliability index (SAIDI, SAIFI, CAIDI) improvements. This is
an excellent example of direct benefits from complex event processing analytics in grid operations.
Visualization and event processing analytics can also be used to identify, locate, and resolve momentary outages,
improving the MAIFI score.
Better analytics may help utilities manage “mutual assistance” crews (from other utilities), and predict how many
resources they may need for impending storms.
Using analytics just for increasing the efficiency of automated outage notifications is not that valuable. Value is higher
in states where regulators penalize utilities for excess outages.

S
e
c
t
i
o
n

2

2
-
6

D
u
k
e

E
n
e
r
g
y

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Revenue Protection
Implement revenue protection alerts using
anomaly/outlier analysis
Develop billing and rate analysis to estimate,
track, and recover lost revenues due to theft
Predict delinquencies using billing data and
socioeconomic information
Accepted theft rates for North American utilities is 2-4% of annual revenues. Advanced algorithms that detect
anomalies can identify where this may occur, allowing for quicker resolution.
Energy theft is the third largest form of theft, behind credit card and autos, at around $6 billion a year; other estimates
cited are $5/meter/year.
There are some reality checks associated with identifying theft that will temper the results:
You may never get all the money stolen back, since there was no formal way of estimating the theft in the past
You also may not have protected any future revenues from those caught stealing, as they probably reduce their
overall use (one way or the other…)
Revenue protection can also come from prediction and proactive resolution of impending customer delinquencies.
Revenue protection also should emphasize protecting base revenues from future competitive and disruptive
influences (like what is happening with SDG&E and RWE concerning renewables). These influences include:
Retail competitors
Distributed Generation/Renewables
Customer Flight due to cost or reliability issues
“Unwanted reduction” of use by customers fuel switching or implementing energy efficiency projects
Remote disconnect switches will allow for faster responses to event alerts. This has shown significant reductions in
slow/non-pay customers, increasing cash flow.
Demand Response (DR)
Develop demand response impact evaluation
metrics
Identify potential DR participants
Predict system load curtailment from DR
programs using real time meter data
Social media could result in increased customer response in DR programs, resulting in a 3-4% increase in peak
reduction.
Analytics can help optimally forecast and dispatch DR, maximizing the ROI of the system.
Identify potential DR participants – Growing a DR program with targeted participants can mean the difference
between a successful program and a complete failure. Actual value will depend upon the program delivered.
Analysis of interval data can help determine peak period usage by customer, allowing Duke to prioritize where it
would be able to reduce load most effectively. This would also help identify new DR participants.
The ability to accurately document savings from DR programs will allow for more value to be recovered in ISO DR
offerings.
Real future value of DR will come from decentralized, localized, and optimized activities. Analytics will help achieve
this.
Development of metrics does not inherently provide direct value, but can attempt to measure the value realized.

U
S
E

C
A
S
E
S

O
V
E
R
V
I
E
W

A
N
D

V
A
L
U
E

E
S
T
I
M
A
T
I
O
N

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

D
u
k
e

E
n
e
r
g
y

2
-
7

Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Asset Management Analysis
Implement condition-based asset
management for distribution devices
Identify and replace overloaded transformers
Identify and monitor abnormal device activity
or loading
Value will come from using analytics to operate conditioned and predictive based maintenance programs.
Reduction of catastrophic asset failures and loss of energy sales from effective asset management.
Reducing costs of maintaining assets.
Analytics can lead to better management of assets to increase reliability and reduce outages.
More precise operation of assets can extend their lives and allow for capital expenditure deferral.
Load Forecasting
Predict customer energy usage using local
weather station information
Develop load forecasts at the meter,
transformer, phase, circuit, and substation
level
Develop optimization strategies to maximize
operational value of distribution assets
More granular information will help make forecasts more accurate, and will allow for the use of different forecast
horizons.
Robust forecast models can improve accuracies, leading to more cost effective decisions.
Accurate load modeling/forecasting – short and long term – allows for better resource planning.
New models can be applied to the big data to help forecasting accuracy. Examples of these models include non-
linear regression, neural networks, and random forecast models.
Forecasts can be key inputs into DMS and DR systems. These systems continue to evolve; benefits will depend on
the capabilities of these systems to translate forecasts into actionable and automated strategies in response to
changing conditions.
There is value from using forecasts to predict equipment overloading or system congestion issues.
Load forecasting benefits are often interrelated with other use cases, such as asset management, energy efficiency,
demand response, customer analytics.
These forecasts are key inputs to distribution automation and distribution management systems as well as demand
response management systems. These systems are evolving and the benefits will depend on the capabilities of
these systems to translate forecasts into actionable and automated strategies in response to evolving conditions.

S
e
c
t
i
o
n

2

2
-
8

D
u
k
e

E
n
e
r
g
y

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Distributed Generation (DG)
Forecast/estimate distribution grid impacts
from customer DG operations
Develop PV adoption and usage models
Identify customers with DG who consistently
produce more electricity than they consume
DG has the potential of being a huge negative value if not managed. Therefore, the more knowledge Duke Energy
has as to where these are located and how they are operated, the better. Duke also has to learn how to profit from
them, since they are already here. Monitoring weather variables against various renewable generation (wind, solar)
could help manage and/or predict grid impacts from DG assets.
Developing various DG scenarios will help Duke Energy understand what the different operational impacts they will
have on the grid.
Advanced analytics may help identify where newly installed DG is located, and how it may be affecting the grid.
Understanding how DG is being used, and its impacts on the grid will help Duke Energy refine its tariffs accordingly.
Modeling and analyzing DG will help Duke understand its variability, and how much spinning reserves it may need to
have ready.
Identification of variability in operations: need for additional spinning reserves (assume 20% of photovoltaic (PV)
capacity must be met by GT), then assume value is 2% of this number.
Cost of developing grid distribution models to help identify and integrate intermittent PV.
Customer Plug-in Electric Vehicle (PEV) usage analytics will help Duke understand how vehicle-to-grid (V2G)
programs may affect the grid operations at the edge.
Energy Efficiency (EE)
Develop/implement an energy efficiency
impact evaluation system
Identify potential energy efficiency program
participants using interval data and
socioeconomic information
Create new programs targeted at specific
customer segments
1.5% -2.5% of additional EE savings can come from the strategic use of social media.
Using Signal Analysis to deconstruct interval data into specific end uses could be used for better energy
recommendations to customers. Combining this with socioeconomic information may allow for better segmentation
and target marketing.
Analytics will help better run and segment customers for EE programs.
The issue is how to calculate value for programs that essentially pay customers to use less energy.
Analytics brings limited additional value to EE efforts if it is only used for impact evaluation activities.
Communications
Implement a smart grid communications
monitoring and analysis system
Implement mobile communications and fleet
telematics system
More effective management of communications infrastructure can occur from the use of complex event processing.
Using analytics to optimize bandwidth and costs of using public communications networks (cellular).
Effective use of analytics with network management systems could help resolve issues and outages more quickly,
reducing costs.
Analytics will help manage service level agreements, with better crew response times.

U
S
E

C
A
S
E
S

O
V
E
R
V
I
E
W

A
N
D

V
A
L
U
E

E
S
T
I
M
A
T
I
O
N

D
a
t
a

M
o
d
e
l
i
n
g

a
n
d

A
n
a
l
y
t
i
c
s

F
i
n
a
l

R
e
p
o
r
t

D
u
k
e

E
n
e
r
g
y

2
-
9

Use Case Category
(with select use case examples)
Participant Responses – Value Opportunities
Security
Implement a distribution cybersecurity
monitoring system
Implement remote visualization and monitoring
security system around critical assets
Value of analytics to increase asset and operations security extends well beyond the grid.
There are examples of value here in other industries that should be researched for parallel benefits to utilities.
$$$-$$$$ in the banking/fraud areas.
What happened with Target is an excellent example of the need for strong cybersecurity measures.
Other
List any other categories or use cases that
may produce value for Duke Energy
Examples were also provided for gas operations and transmissions operations (see Space-Time Insight’s final report
for more information on these).

Value Code Value Estimate
$ Thousands in value
$$ Millions in value
$$$ Tens of millions in value
$$$$ Hundreds of millions in value
$$$$$ Billions in value

Data Modeling and Analytics Final Report
Section 3
TOOLS, SYSTEMS AND ANALYTICS USED
3.1 Overview
The Duke Energy Team evaluated and summarized the IT tools and analytics models found in
the vendor responses. Vendor responses included interim reports, final reports, and notes taken
during phone call meetings. The summary of this content is documented in the following
sections.
3.2 Data Analytics Tools used by Vendors
There was significant breadth and diversity of software tools used by the vendors in DMAI.
Tools used are organized into the categories of Ingestion, Storage, Visualization, Analytics, and
Business Intelligence. Also shown, summarized by vendor, is whether their software platform is
offered through Cloud computing Big Data as a Service (BDaaS). Vendor Industry focus is listed
here as well. Table 3-1 summarizes the tools by each big data category.
Each vendor has a different history that shows their approach to analytics and also reveals their
strengths. Each vendor in the DMAI has strong interest in competing into the energy sector
analytics space. Vendor expertise can be analyzed through the following characterizations:

Energy Sector expertise. These vendors have worked in the Energy Sector for many years.
They have hardware products, software systems, and services that address the needs of this
sector. They have moved into analytics through organic development, or by purchasing or
partnering with analytics and IT companies.

Analytics expertise. These vendors have deep analytics skills and approach the market by
providing solutions across multiple business sectors.

Information Technology (IT) expertise. These vendors have focused on the IT solutions that
support the Big Data analytics ecosystems. Their products can be considered the base
infrastructure for analytics and the enabler for handling Big Data. Many are working to grow
their businesses into vertical offerings by business sector. Some have their beginnings in data
warehouse products, while others began with visualization tools.
No one vendor can provide the best-in-breed solution for all three skill categories.
The IT expertise space is where you find open-source software (OSS) solutions. Standards do not
exist in the visualization, analytics, and business intelligence categories; thus, it is not typical to
find OSS in these categories.
The vendors that bring expertise in the Energy Sector and Analytics typically utilize the IT tools
from the IT experts.

Section 3
3-2 Duke Energy Data Modeling and Analytics Final Report
Table 3-1
Summary List of Tools by Big Data Component
Ingest Storage Visualize Analytics BI
Collect, Ingest,
ETL
Store, Warehouse,
Data Science
Discover,
Visualize
Model, Statistics
Apps, Business Intelligence,
Framework
Sqoop Hadoop Pig MapReduce EDP
SSIS HDFS Hive Python DROMS
Informatica HBase Tableau CustomerIP CustomerIP
Custom ETL Redis SQL SAS Cognos
First Logic Cloudera Hadoop Microstrategy Matlab SAP BO
VEE MS-SQL Visiokio R Decision
MySQL Hawq Model Builder Falcon
MongoDB Autonomy IDOL RStudio Score
Oracle RDB Vertica Mahout Xpress Optimization
Couchbase SAP BO Spotfire Predix
Pivotal GPDB BAS SAS DS2 Integrated DSM GridIQ
SAS Data SAS FedSQL LoadSeer LoadSeer
Enterprise DW SAS/Graph DSMore QDMP
OSI Pi Visual Analytics SmartSpotter SAS BI
CEP OLAP IDROP SEF
Netezza Presto Enterprise Miner GeoAnalytics
SAP HANA STI Dashboard High-Performance Analytics
Apama CEP Spotfire RabbitMQ

Table 3-1 above shows that there is a wide range of tools used by the 17 Vendors in DMAI. The
Duke Energy Team took some time to take a cursory look at these tools and by doing so, the
following insights arose:
Ingestion: The data ingestion process involves processing data from various sources and
cleaning the data as it comes in. This process is called ETL (extract, transform, and load). It
involves understanding how to communicate with the data sources, rules to transform the data
into formats that are recognizable by downstream systems, and data-loading into new data stores.
This process is costly and time-consuming and warrants the question: “Why is the data quality so
poor and the data formats so diverse that it needs to be cleansed and transformed by Big Data
processes in the data center?” Future systems, based on a distributed architecture enabling the
autonomous Grid, should address this time-consuming and expensive process by slicing up the
process and moving its functionality to the collecting device.
Storage: In the last 30 years, the primary data store has been the relational database management
systems (RDBMS) with its standardized structured query language (SQL). The relational
database space has been ruled by the enterprise giants Oracle, IBM, and Microsoft. The SQL
base is a strong standard and the space has become more commoditized, allowing OSS providers
Open-source software

TOOLS, SYSTEMS AND ANALYTICS USED
Data Modeling and Analytics Final Report Duke Energy 3-3
to enter. In order to answer known questions more quickly across large data typically in time-
series, OLAP was introduced and cubes were pre-formed to support fast visualization.
In the last 10 years, the relational database model has been placed under great strain with the
growth of the Internet, unstructured data sources, and the “Internet of Things” devices coming
on-line. The inability of the RDBMS to handle vast quantities of unstructured data and the
advent of high-speed commodity hardware, opened the world of parallel processing to all, which
at one time, was only accessible to those who could afford super-computers. In 2003, Nutch (to
become Hadoop) was created and in 2004 Google created MapReduce. Parallel processing
moved to the forefront to solve the problem of processing billions of web pages.
In-memory databases and complex event processing (CEP) has become popular and necessary
where high bandwidth and millions of transactions are present, like in the banking industry
where fraud detection in milliseconds is a requirement.
Since then, several types of NoSQL database structures have been created to better manage
unstructured data:

Key-value store: example Redis

Tabular: example HBase

Document oriented: example MongoDB
NoSQL does not support joins, has no complex transactions, and offers no constraints. NoSQL
enables the ability to store and retrieve large quantities of data. It supports dynamic growth of
data and easily takes in new types of data added to the system. The data is not highly structured
and the structure that is present is allowed to change over time. Relationship understanding and
constraint management is moved to the programs and scripts that process the data. For example,
joins can be pre-formed in the data prior to storage.
It is important for Duke Energy architects to consider the types of data that support the primary
use cases of the utility of the future so that the determination of the right-fit data structures and
stores can be made.
Visualization, Analytics, and Business Intelligence: The DMAI was not focused on deep
analysis of Vendor visualization, analytics, and BI tools. It is clear that these tools are an
important aspect of a data analytics strategy. Each Vendor final report included graphs, charts,
geo-views, and dashboards to show the results of their analysis and in some cases to differentiate
their offering. A comprehensive dashboard, given the data and use cases discovered in the
initiative, would include: Transformer data, Geo-location, Socioeconomic information, Outage
data, Meter readings data, Time-series events, User interaction through queries, filters, tables,
and graphs.
A quick search of the Internet reveals hundreds of visualization tools and libraries. Many OSS
solutions are available. An example of a strong OSS software solution is the R language for
statistics and modeling; it includes a rich visualization library.
The analysis of visualization, analytics, and BI tools is an important task in the building a data
analytics strategy.

Section 3
3-4 Duke Energy Data Modeling and Analytics Final Report
3.3 Analytics Used by Vendors
The Duke Energy Team reviewed the Use Cases and Models presented by the Vendors in the
final reports. The following is a summary of the analysis tools and models mentioned:

Fundamental Statistics:

Descriptive Statistics (Max, Min, Average, Median)

Graphs and Charts: Box Charts, Whisker Charts

Inferential Statistics

Outlier Analysis and Correlation:
– Statistical Process Control, R-Chart
– Linear Regression
– Multivariate Regression with Correlation Coefficients
– Time Series Analysis

More Complex Analysis:
– Self-learning Analytics
– Predictive Analysis
– Profiling
– Scorecards
– Clustering
– Decision Trees
– Visual Analytics
– Random Forest
– Support Vector Machines
– Gradient Boosting
– Logistic Regression
– Neural Networks
– Tree Ensemble Modeling
– Econometric Models
3.4 Insights and Observations

No one vendor can do it all: IT, data science, and Energy sector.

Each vendor had a starting point. Some started with:

Data warehousing products

Relational database tools

TOOLS, SYSTEMS AND ANALYTICS USED
Data Modeling and Analytics Final Report Duke Energy 3-5

Visualization software tools

Data analytics and statistics

In-memory quasi database solution

Energy Sector knowledge and problem solving

Financial industry knowledge and problem solving

Health care sector knowledge and problem solving

Digital sensors and appliances, low-cost communications and storage, demand for
information, and the Internet of Things (IOT), have created an IT environment where Big
Data is a problem.

Duke Energy needs to have a Big Data problem; current state is that we have thousands of
small data problems.

Each use case category evaluated contains silos of systems, and the categories themselves are
silos; the greatest value will come when all data is reachable by all for analysis.

Vendors that focus on analytics only, and on the energy sector, are the visionaries we should
collaborate with to find new value streams.

The majority of Big Data analytics vendors do not know our business, but they are learning.

The strength of the Relational Database (RDBMS) is its schema, but this is also its greatest
weakness since the schema is inflexible and difficult to change.

When your questions about the data become more complex and cross silos, then the complex
RDBMS schemas break down.

If you know the questions that you want to ask and the data structures are well-known and not
dynamic, then the RDBMS is the best solution.

If you are not sure of the questions and do not know the relationships across the data, then
you may plan to disaggregate your relational data into an unstructured format and query the
data with parallel processing tools to seek out new relationships.

The strength of Hadoop, NoSQL, and MapReduce is the ability to store data in its original
form in simple structures and process it with low-cost parallel computers.

Hadoop usage can reduce IT costs and can scale; runs on parallelized commodity hardware
and OSS has no license costs.

IT Tools for Ingestion and Storage are becoming more commoditized and open; the
Visualization, Analytics, and BI applications are more specialized by vendor and are not yet
standardized.

Technology is evolving so fast that Duke cannot keep up.

A small group of vendors are focused on complex event processing and in-memory databases.

Recommend to work with energy-focused analytics vendors to drive vision and use case
focus.

Work with the best data science vendors to drive IT systems.

Data Modeling and Analytics Final Report
Section 4
RECOMMENDATIONS AND NEXT STEPS
4.1 Next Steps
The following are recommendations and next steps for Duke Energy to implement in 2014:

Expand the list of collaborative vendors beyond the initial participants, as they continue to
provide valuable insights and feedback with respect to their offerings and experiences. Meet
with them individually to discuss the technical and analytic results of their research.

Continue work on refining and forecasting the potential value of having a big data and
analytics platform at Duke Energy. Prioritize the use case categories to pursue, identify the
top use cases, and develop a base case forecast for realizing the value.

Utilize the results of the IT systems review to start development of technical and functional
specifications that will become part of Sandbox 2.0 and the big data platform.

Present final results to Analytics Leadership Team and make the resources, research, and
documentation available to the team for input into their 2014 activities.

Prepare an organizational structure, resource, and skills gap analysis to identify areas that
Duke Energy may want to supplement with future employees or external resources. Develop
job descriptions for the key positions.

Meet with Duke IT to complete technical evaluation of new tools.

Consolidate vendor data issues, comments, and modeling methodologies.

Continue to define Duke Energy Big Data Architecture.

Data Modeling and Analytics Final Report
Appendix A
LIST OF USE CASES

Use Case
#
Use Case Name Category
1 Develop Asset Lifecycle Model Asset Management
2 Cluster or identify outliers in communication events Communications
3 Smart grid communications network health monitoring and reporting Communications
4 Customer Peak Analysis (Duke DMAI) Customer Analysis
5 Aggregate Customer Usage Profile (Duke DMAI) Customer Analysis
6 Identify customers with distributed generation Customer Analysis
7 Correlate socioeconomics data with energy data (kWh) Customer Analysis
8 Correlate customer demand (kW) and socioeconomic variables Customer Analysis
9 Identify and analyze abnormal customer use patterns Customer Analysis
10 Segment Customers for Power Manager using socioeconomic information Customer Analysis
11 Predict potential AC failure through analyzing load patterns on in-home
circuits
Customer Analysis
12 Develop household energy cluster profiles using billing and socioeconomic
data
Customer Analysis
13 Develop residential end use energy estimates using interval meter data Customer Analysis
14 Customer Load Pattern Analysis Customer Analysis
15 Identify PEV Charging Location Customer Analysis
16 Identify Vampire loads Customer Analysis
17 Determine optimal customer thermostat settings Customer Analysis
18 Calculate HVAC System Sizing from meter data Customer Analysis
19 Target Marketing and Consumer Engagement Customer Analysis
20 Predict customers at risk of attrition using socioeconomic and social media
information
Customer Analysis
21 Customer Sentiment Analysis Customer Analysis
22 Identify overloaded customer service points Customer Analysis
23 Develop rate class profiles for demand and voltage Customer Analysis
24 Compare monthly consumption for potential high bill complaint cases Customer Analysis
25 Calculate usage statistics by neighborhood for customer comparative
analysis
Customer Analysis
26 Develop energy usage/sq. ft. metrics for gas and electric by rate class and
customer segment
Customer Analysis
27 Seasonal energy consumption analysis Customer Analysis

Appendix A
A-2 Duke Energy Data Modeling and Analytics Final Report 6/16/14
Use Case
#
Use Case Name Category
28 Consolidate and analyze customer interactions (payment history, call
center interaction, outages, etc.) to evaluate and predict customer
sentiments to the utility
Customer Analysis
29 Develop annual voltage profiles by customer Customer Analysis
30 Proactive Identification of PEVs Customer Analysis
31 Demand Response Impact Evaluation Demand Response
32 Identify Potential DR Participants Demand Response
33 Demand Response Equipment Monitoring Demand Response
34 Demand Response Forecasts Demand Response
35 Identification of non-operational demand response devices Demand Response
36 Identify and monitor customers with central air conditioning as potential
demand response participants
Demand Response
37 Identify high energy customers during summer and winter peaks for
demand response programs
Demand Response
38 Identify customers with electric water heating for demand response
programs
Demand Response
39 Predict system load curtailment from demand response programs using
real time meter data
Demand Response
40 Monitor home area network (HAN) connectivity Device Analytics
41 Calculate feeder Losses Distribution Grid Analysis
42 Customer DG Power Production Analysis Distributed Generation
43 PV Generation Modeling Distributed Generation
44 PV Adoption Forecasting Distributed Generation
45 Identify DG customers who are producing more than they are consuming Distributed Generation
46 Calculate loads at meter, transformer, circuit, and substation level by time
interval
Distribution Grid Analysis
47 Identify network connectivity from GPS Distribution Grid Analysis
48 Update Duke Energy Network Model Distribution Grid Analysis
49 Validate Peak Load Shaving Impacts Distribution Grid Analysis
50 Optimize Distribution Grid Sensor Locations Distribution Grid Analysis
51 Feeder Loss Analysis Distribution Grid Analysis
52 Phase Load Balance Analysis Distribution Grid Analysis
53 System Load Evaluation Distribution Grid Analysis
54 Distribution Facility Forecasting Distribution Grid analysis
55 Locational Distribution and Transmission Benefits Distribution Grid Analysis
56 Locational Uncertainty and Option Value Distribution Grid Analysis
57 Distributed Optimization to Maximize Benefits Distribution Grid Analysis
58 Aggregate customer loads on a feeder to compare with SCADA data to
estimate feeder losses
Distribution Grid Analysis
59 Calculate peak feeder or feeder segments Distribution Grid Analysis

LIST OF USE CASES
Data Modeling and Analytics Final Report Duke Energy A-3
Use Case
#
Use Case Name Category
60 Determine where segments of grid have reached __% of capacity Distribution Grid Analysis
61 Provide power factor for all feeders and feeder segments Distribution Grid Analysis
62 Augment SCADA and Pi system data with voltage and power
measurements from the meter
Distribution Grid Analysis
63 Identify anomalies in meter-transformer mapping Distribution Grid Analysis
64 Identify 3-phase meters assigned to a single phase transformer Distribution Grid Analysis
65 Analyze distribution device operations by feeder Distribution Grid Analysis
66 Identify voltage variations by device Distribution Grid Analysis
67 Plot voltage profile by meter, transformer then by feeder over time Distribution Grid Analysis
68 Cluster or identify outliers in distribution device data (voltage, PF) Distribution Grid Analysis
69 Project value of vegetation management activities Distribution Grid Analysis
70 Develop Grid Parity Model Distribution Grid Analysis
71 Energy Efficiency Impact Evaluation Energy Efficiency
72 Develop energy efficiency program impact evaluation analytics Energy Efficiency
73 Financial Closing Analysis Financial Analysis
74 Revenue Forecasting Financial Analysis
75 Predicting Customer Energy Usage Using Weather Variables (Duke
DMAI)
Load Forecasting
76 Develop load forecasts at meter, transformer, circuit, and substation level Load Forecasting
77 Identify customer movement with socioeconomic information Load Forecasting
78 Develop regional zone load forecasts Load Forecasting
79 Create load profiles for each rate class using interval meter data Load Forecasting
80 Forecast storm path and impact on distribution grid Load Forecasting
81 Detect "Hot Meter Sockets" Meter Analytics
82 Reconcile all installed meters with their provisioning status Meter Analytics
83 Reconcile all installed meters with those that have successfully billed Meter Analytics
84 Meter failure analysis and notification Meter Analytics
85 Identify/Reconcile correct asset location on GIS Meter Analytics
86 Meter malfunction trend analysis and reports Meter Analytics
87 Detect/report physical meter breaches Meter Analytics
88 Identify meters without GPS coordinates Meter Analytics
89 Identify bottleneck meters within the mesh network connectivity Meter Analytics
90 Meter temperature monitoring and analysis Meter Analytics
91 Identify orphan meters Meter Analytics
92 Monitor and analyze meter remote disconnect functionality Meter Analytics
93 Collect, analyze and report DOE metrics Meter Analytics
94 Interactive map to display meters and network equipment in various
operational states
Meter Analytics
95 Identify and track customer power factors Meter Analytics
96 Identify meters with incorrect multipliers Meter Analytics

Appendix A
A-4 Duke Energy Data Modeling and Analytics Final Report 6/16/14
Use Case
#
Use Case Name Category
97 Detect abnormal gas spikes Meter Analytics
98 Identify gas meters with low variability summer load to identify potential
gas leaks
Meter Analytics
99 Identify gas register reverses Meter Analytics
100 Usage Outlier Analysis – Meter Interval Data (Duke DMAI) Meter Analytics
101 Collect and analyze meter event data Meter Analytics
102 Identify potential change in energy use after a meter change out Meter Analytics
103 Validate Meter constant Meter Analytics
104 Determine phase/circuit for single-phase meters based on AMI events Meter Analytics
105 AMI Load Profile Development Meter Analytics
106 Calculate Power Factor using interval meter data Meter Analytics
107 Use interval meter data to fix meter-to-transformer topology Meter Analytics
108 Predict meter device failures Meter Analytics
109 Duke Energy Outage Analysis (Duke DMAI) Outage Analysis
110 Identify individual/localized outages Outage Analysis
111 Identify large outages Outage Analysis
112 Compute and visualize time/speed of power restoration Outage Analysis
113 Correlate outage and storm data Outage Analysis
114 Correlate outage with distribution equipment Outage Analysis
115 Predict outage duration Outage Analysis
116 Compute outage statistics by distribution device Outage Analysis
117 Identify "false positive" meter outages Outage Analysis
118 Identify outage locations in real-time Outage Analysis
119 Identify disturbances/momentary outages Outage Analysis
120 Outage Impacts on Feeder Reliability Outage Analysis
121 Automated Outage Reports Outage Analysis
122 Develop CAIDI, SAIFI and SAIDI statistics using interval meter data Outage Analysis
123 Analyze feeder segments for frequent momentary outages Outage Analysis
124 Track geographic areas of multiple momentary outages Outage Analysis
125 Identify meters with repeated sustained outages Outage Analysis
126 Identify meters with repeated momentary outages Outage Analysis
127 Improve CAIDI using smart meter restoration timestamp Outage Analysis
128 Detect usage on inactive or disconnected accounts Revenue Protection
129 Detect silent meters Revenue Protection
130 Detect failed meters Revenue Protection
131 Detect anomalies using interval meter data and meter events data Revenue Protection
132 Predict delinquencies using billing data and socioeconomic information Revenue Protection
133 Identify Energy Bypassing meter using kW and Voltage data Revenue Protection

LIST OF USE CASES
Data Modeling and Analytics Final Report Duke Energy A-5
Use Case
#
Use Case Name Category
134 Optimize delinquent account collections using socioeconomic and social
media information
Revenue Protection
135 Identify service points where electric meter has been removed but gas
meter is still active
Revenue Protection
136 Analyze meter consumption for those homes with services disconnected
(as identified by billing)
Revenue Protection
137 Develop Consumption Pattern Analytics - monitor for changes over time Revenue Protection
138 Switched/Swapped Meter Analysis Revenue Protection
139 Identify meter potential fuse failure Revenue Protection
140 Inactive meter usage analysis Revenue Protection
141 Identify stopped electric meters Revenue Protection
142 Identify service points where electric usage is low by gas usage is normal Revenue Protection
143 Monthly report on inactive meters Revenue Protection
144 Revenue Protection Alerts - Based on Threshold Criteria Revenue Protection
145 Track and report customer billing to recover lost sales because of theft Revenue Protection
146 Identify active gas meters showing no consumption over a specified period
of time
Revenue Protection
147 Identify logical security breaches Security
148 Incorporate fleet telematics Service
149 Technology Analytics Technology Analysis
150 Transformer Coincident Peak Analysis (Duke DMAI) Transformer Analysis
151 Calculate transformer loading using customer interval meter data Transformer Analysis
152 Calculate Transformer Load Diversity Transformer Analysis
153 Investigate impact of additional load on existing transformer Transformer Analysis
154 Evaluation of transformer usage Transformer Analysis
155 Transformer loading and voltage profile Transformer Analysis
156 Endpoint loading and voltage profile contributing to transformer loading Transformer Analysis
157 Capture kVA data from meter to and use (with others) to calculate kVA
loading of a transformer
Transformer Analysis

doc_974419104.pdf