Download - CIGRE Presentation
LISBON 2013
1.4 Information Technologies for the Smart Grid
Advanced Utility Analytics with Object-Oriented Database Technology
Paul Myrda, John Simmins, Bert Taube
USA
Paper Reference 144
LISBON 2013 Overview
• Uniqueness of the utility data context
• Data provides a “menu” of opportunities for improvement
• Focus of EPRI research in the area of “Big Data”
• Lack of situational awareness and analytics in power transmission demonstrated by August 14, 2003, blackout
• Demo showing an effective analytic approach to create situational awareness and decision making in transmission
• Data challenge and technology solution to create real-time situational awareness and analytics based on PMU data
LISBON 2013 Big Utility Data Management
and Analytics – Data Diversity Challenge
Big Energy Data Features Energy Data Types Energy Data Sample Rates
µs – Range
(e.g. HF Switching Devices)
ms – Range
(e.g. PMU Devices)
sec – Range
(e.g. DER Output Variations)
min – Range
(e.g. Service Restoration)
hour – Range
(e.g. Demand Response)
Day – Range
(e.g. Day-ahead Scheduling)
Year – Range
(e.g. Life of IT Asset)
Decade – Range
(e.g. Life of OT Asset)
Data Volume
(e.g. TBytes per Day)
Data Velocity
(e.g. 300,000 Data Objects/sec)
Data Variety
(e.g. Large Variety of Data
Object Types/Classes)
Data Validity
(e.g. Large Variety in Data
Object Shelf Life)
Data Veracity
(e.g. Large Variety of Data
Objects with different Data
Quality)
Telemetric Data
(e.g. in SCADA Systems;
normally in Historian)
Oscillographic Data
(e.g. in Power Quality Monitor;
normally in Historian)
Usage Data
(e.g. in Meter Data Management
System; normally in RDBMS)
Asynchronous Event Messages
(e.g. in Distribution Managem.
System; normally in RDBMS)
Meta Data
(e.g. in Geospatial Info System;
normally in RDBMS)
LISBON 2013 Building Business Value through Increased Situational Awareness
• Telemetry
• Oscillography
• Usage Data
• Asynchronous Event Messaging
• Meta-data
• Leveraging Asset
• Increased Reliability
• Increased Efficiency (Workforce)
• Increased Efficiency (Grid)
• Silo-busting
LISBON 2013 Situational Awareness & Analytics in Power Transmission – August 2003
Largest ever North American Power Grid Blackout on August 14, 2003 (after 4PM)
Affected 50 million people and more than 70GW of electrical load in OH, MI, NY, PA, NJ,
CT, MA, VT, and the Canadian provinces Ontario and Quebec. Some areas were left
without power for several days with rotating blackouts for up to 2 weeks.
Final NERC Report (7/13/2004)
Lack of situational Awareness of
Line Outages and degraded
Conditions (Group 1 Cause)
No effective Management of
Vegetation in Transmission
Rights-of-Way (Group 2 Cause)
No Provision of effective diagnostic
Support (Group 3 Cause)
DOE/FERC Report (2/2006)
Recommendation to establish a
Real-Time Transmission
Monitoring System (based on
Synchrophasor Measurement
Network)
Recommendation to develop
Computer-based operational and
Management Tools
LISBON 2013 Prior to the August 2013 Blackout
Generation, Demand, Interregional Power Flows on
August 14 before start of more severe part of Blackout
Scenario at 15:05PM
LISBON 2013 Evolution of August 2003 Blackout
Power Flows at 16:05:57, Prior
to the Sammis-Star Trip (25deg)
Power Flows at 16:05:58, After
the Sammis-Star Trip (12deg)Power Flows at 16:09:25
(45deg)
Power Flows at 16:10:44
(115deg)
Power Flows at 16:10:45
(115deg)
Changes in Direction of Power Flow during the Course of Blackout Events correlated
with severe Changes in Phase Angular Differences between Locations involved!
LISBON 2013 Divergent Phase Angle
LISBON 2013 Phase Angle Divergence – Time Series
LISBON 2013 Synchrophasor Measurement Networks
– The Real Big Data Challenge
TimePast (August 14, 2003) Present Future
15 PMUs 150 PMUs 1,000 PMUs
Generated Data
Data Volume
470 – 620 Million
Data Points per Day
Data Size
2.3 – 6.0
GByte per Day
Generated Data
Data Volume
4.7 – 6.2 Billion
Data Points per Day
Data Size
22.6 – 60.3
GByte per Day
Generated Data
Data Volume
31.1 – 41.5 Billion
Data Points per Day
Data Size
150.6 – 401.7
GByte per Day
LISBON 2013 Synchrophasor Measurement
Networks – The Real Big Data Solution
JPA
(IEC CIM Model) Integration Layer
Server Network I/O Layer
V/OD1
(IEC CIM)
Osiris Server
Network/Cloud
Data
Ingestion API
Data Streams
In-server
Analytics API
Sensor/Machines
(e.g. Synchrophasor)
Networks
(e.g. Transmission)
Hadoop/MapReduce
(e.g. AMI/GIS)
Server Network I/O Layer
R/Revolution Analytics
(e.g. Time Series)
Graph Analysis
(e.g. Suppl. Cost Opti.)
Complex Event Proc
(e.g. Real-Time Oper.)
Data Analysis
V/OD1
(IEC CIM)…
Transactional Apps
(e.g. Blackout Demo)
Analytical Apps
(e.g. R Desktop)
Client Network I/O Layer
Network/Cloud
LISBON 2013 JPA Performance Benchmarking Demo:
Linked NoSQL vs. Hibernate Postgres
Database Performance Benchmarking Test Results: Cost
Data Ingestion Test for IEC CIM
End Device Event Profile Objects
3 Years 5 Years T
Cost
$777k
$4,504k
Linked NoSQL
Hibernate Postgres
F = 5.8
10 Years
In-Database Mixed Workload Test for IEC CIM
End Device Event Profile Objects
3 Years 5 Years T
Cost Linked NoSQL
Hibernate Postgres
10 Years
$948k
$7,148k
F = 7.5
$13,758k
$1,378k
F = 10
$2,002k
$4,023k
F = 2 $2,943k
$7,276k
F = 2.5 $5,295k
$15,409k
F = 2.9
Assumptions: (i) 1 DB Server requires 2 App Servers (Linked NoSQL) and
(ii) 1 DB Server requires 5 App Servers (Hibernate Postgres)
LISBON 2013 Building Business Value through Increased Situational Awareness