20
1 CenterPoint Energy Time Machines: The Evolution and Application of Predictive Analytics Dr. Steven P. Pratt Chief Technology Officer

Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

Embed Size (px)

Citation preview

Page 1: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

1

CenterPoint Energy

Time Machines: The Evolution and Application of

Predictive Analytics

Dr. Steven P. Pratt

Chief Technology

Officer

Page 2: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential 2

CenterPoint Energy

Page 3: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

The Last Seven Years

3

From 12 to 28 State

Operation

From Analog to Digital Grid

From 80,000 to 221,000,000

meter reads/day

From 700TB to 5.8 PB

From Data Reactive to

Decision Proactive

From Routine Operations to

Disaster Recovery

Page 4: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential 4

Vendor

IT

Provider Hybrid

IT IT/OT

PT Technology and technology related

services are built on a foundation

of global, geographically dispersed

and standardized elements

delivered and supported through

partnerships

FUTURE

Digital services have evolved

from a purely vendor

provisioning model to a

symbiotic and codependent

delivery of business

functionality

PRESENT

The Metamorphosis of Digital Services

Page 5: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Continuing Technology Operation’s Areas of Opportunity

5

• Competition for resources

• Return on technology investment

• Application rationalization

• Cloud deployment

• Operationalization

• Automation

• Resiliency

• Solution Quality

• Incident reduction and response

• Balanced project portfolio

• Operational complexity

• Software rationalization

• Data management

• Technology governance

• Standardization

• Innovation management

• Strategic continuity

• Technology obsolescence

• Mobile maturity

• Metric measurement

Page 6: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Business Drivers

6

Safety Customers Employees Society All

Innovation for Growth Intelligent Meters Smart Grid

Business Transformation Information Technology Operations Technology

Consumer Evolution Access Expectation Preferences

Strategic Consistency Single Reference Architecture Encapsulating Frameworks Execution

Operational Optimization

Service Catalog Automated Operations Consolidated Portfolios

Innovation Agenda

Modernization Integration

Constituency Focus

Customer Vision Employee Satisfaction Societal Benefit Regulatory Compliance

Value Realization

Corporate Data Management & Control Technology Rationalization Big Data

Secured Assets

Technology Challenges

Page 7: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Estimated Five Year Data Growth

7

0 TB

2,000 TB

4,000 TB

6,000 TB

8,000 TB

10,000 TB

12,000 TB

14,000 TB

1 Yr 2 Yrs 3 Yrs 4 Yrs 5 Yrs

Business As Usual EOY Usable Disk Stg

Tier 0/1 Tier 3

Page 8: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Changing Emphasis on Data

8

Virtualization:

Data Structure

How much data you have

Where your data is

What we know

What happened

What’s next

Data persistence

Realization:

Data Interoperability

How much you can do with your data

Where your data is used

What we don’t know

What could happen

What’s now

Data Dynamism

“The difference between Data Virtualization and Intelligence Realization is

analogous to that of saving money or adding value”

Page 9: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Transformation From Data Driven to Intelligence Driven

9

Operation

Independent

Internal Centricity

Customer Driven

Data Driven

Architectures

Divergence

IT/OT

− Dark Data

− Complex

Data

Management

− Information

Centricity

− Data Compression

− Data Derivatives

− Complex Events

− Complex Analytics

Innovation

Co-dependent

External Centricity

Consumer Driven

Intelligence Driven

Architecture

Convergence

CT (CenterPoint

Energy Technology)

Strategic shifts require changes in constructs, methods, and speed in Data

innovation

Page 10: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

“Big Data” as the Basis for Predictability

How an organization chooses to manage “big data” differentiates increased cost from derived value and distinguishes between liability and asset

“Big Data” should be evaluated from three perspectives: Management, Governance, and Insight

“Big Data” requires efficient, effective, and economic management. Advanced compression technologies, automation, archiving, and storage tiering reduce costs and lessen dependence on specialized skill sets

Data growth without bound creates a costly and unnecessary burden. Organizational governance structures ensure data created and maintained is relevant, meets business needs, and follows processes for creation, use, and retirement of data resources

Organizations must recognize data as an asset to be mined for its residual value. “Big Data” provides a broad sample space for knowledge and value creation beyond that for which the data was originally created. Analytics and Advanced Analytics provide for practically unlimited data analysis yielding insights such as situational awareness, operational decision making, customer knowledge, and potential new business opportunities

10

Page 11: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Hypothesis

11

Time travel does not require us to be present in the past or the future but

simply understand the context of either.

Page 12: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Historical Context

12

Everything that possibly could

happen has likely already

happened.

There is a taxonomic classification

of historical events.

Every historical event can be

defined in terms of the four

dimensions.

Historical events are necessarily

directly or indirectly related.

Future state is a composite of

historical elements, taxonomic class,

dimensional context, and

association.

Mathematics determines the

accuracy of the future state

representation

The amount, order, structure,

connectivity, and type of data is the

basis for predicting future state

H.G. Well’s Depiction of a Time Machine

Page 13: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Value of Data-in-Memory Computing

13

Memory Data

Data-in-Memory

Transform

Load/Unload

Merge

Data is stored in a column store directly in

memory, naturally compressed and optimized

for high speed analytic processing HANA

Application logic is pushed into HANA’s in-

memory predictive and calculation engine as a

strategic platform for all SAP future applications

Eliminates latency and significantly simplifies the

landscape including infrastructure, resource

development and maintenance of applications

Virtually transform or model data for direct

consumption by in-memory, embedded

predictive functions

Enables agility in development and support of

applications and Reduces resources required to

develop, maintain and support applications

Enables insight to ALL data with exponentially

greater performance and time to results

Page 14: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

The Predictability Horizon

14

“If we have sufficient data from the past we can sufficiently predict the future”

HANA

BIG DATA

Load Study

Transformer Load Study

Demand Forecasting

Diversion Detection Asset Management

Usage Insight Regulatory Market Study

Gas Forecasting Event Aggregation

Financial Modeling Customer Segmentation

and Sentiment

Page 15: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

HANA as an Analytics Solution

15

HANA established

as an interim

platform for data

aggregation,

applications, and

advanced reporting

HANA evolves to a

strategic solution

for data integration

and business

functionality and

analytics

Page 16: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

A CRM Example

16

Ba

ck O

ffic

e

Customers HANA

Analytics

Marketing

Call

Center

Channel

One

Channel

Two

Channel

Three

Scenario: Customer contacts the call

center with the potential of one or

more of 40 potential reasons for

calling.

Utilizing a HANA based Predictive

Analytics Engine (PAE), the most

likely reason for the customer call is

predicted, either deflecting the call

entirely or reducing the call agent

handling time.

The same prediction can be used to

proactively communicate with the

customer.

This degree of predictability not only

increase customer satisfaction but

also the productivity of the call center

agents.

Solution: The PAE combines historical data

from multiple sources to create

approximately 14,291,200 data points

resulting in the most likely reason for a

customer call.

The calculation is completed in 1 second and

represents a 9000% reduction in the time

required over prior methods of prediction.

Page 17: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

A Prediction Example

17

Page 18: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

HANA Application View

18 SELECT DRILL-DOWN PREDICT

Page 19: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

CenterPoint Energy Proprietary and Confidential

Predictive Analytics serve as a Roadmap for Deriving Value from Data

19

Deliver maximum value from our combined information assets through 2020 and beyond:

Optimization of the data resources we create and maintain

Development of an optimal cost and support model to balance exponential growth with resultant data value

Return on our investment in data and information assets through application of advanced analytics and automated operations decision-making

Institutionalizing long-term data management through a strong and sustainable governance model

The Application:

Building an “active” Smart Analytics System that captures, virtualizes, interrogates, and realizes data outcomes (Intelligence Realization)

Assembling a framework of interoperability between technology including system management, complex processing, real-time analytics, and Service Orientation

Creating a Decision Services team to develop and support both business and operational analytical context

Transforming to an Industry specific data model for consistency of data constructs across the Enterprise.

Evolving from data virtualization to Intelligence Realization

Page 20: Time Machines: The Evolution and Application of Predictive Analytics-Dr Steven P. Pratt, PhD., CenterPoint Energy, Inc

20

Thank You on Behalf of CenterPoint

Energy [email protected]