Predictive Analytics for Real-time Energy Management … managed by General Atomics ... Second life...

SAN DIEGO SUPERCOMPUTER CENTER

at the UNIVERSITY OF CALIFORNIA; SAN DIEGO

Presented by

Predictive Analytics for Real-time Energy Management and Dynamic Demand Response

Natasha Balac, Ph.D.Director, Predictive Analytics Center of ExcellenceSan Diego Supercomputer CenterUniversity of California, San Diego



Brief History of SDSC

• 1985-1997: NSF national supercomputer center; managed by General Atomics

• 1997-2007: NSF PACI program leadership center; managed by UCSD• PACI: Partnerships for Advanced Computational

Infrastructure

• 2007-2009: Internal transition to support more diversified research computing• still NSF national “resource provider”

• 2009-future: Multi-constituency cyberinfrastructure (CI) center• provide data-intensive CI resources, services, and

expertise for campus, state, and nation

• Approaching $1B in lifetime contract and grant activity



Today’s Information Technologies Drive 21st Century Solutions

• Labs and Centers• Advanced Query Processing• Spatial Data Systems• Data-Driven High Performance Computing• Scientific Workflow Automation Technologies • Large-scale Data Systems• Predictive Analytics• Data Visualization

• We teach and mentor students

• Graph Data Management• Semantic Web• Machine Learning and Analytics• GIS and Spatial Technologies• Information Integration• Text Analytics• Big Data technologies

Will California levees hold in a large-scale

earthquake?

What therapies can be used to cure or control

cancer?

Can financial data be used to accurately predict market

outcomes?

What plants work best for biofuels?



What is Gordon?

• A “data-intensive” supercomputer based on SSD flash memory and virtual shared memory

• Emphasizes MEM and IO over FLOPS

• Designed specifically for data-intensive HPC applications

• System utilization is cut in half every 18 months while it takes 2x time to read your disk every 18 months



Gordon Speeds and FeedsINTEL SANDY BRIDGE COMPUTE NODE

Sockets & Cores 2 & 16

Clock speed 2.6 GHz

DRAM capacity and speed 64 GB, 1,333 MHz

INTEL710 eMLC FLASH I/O NODE

NAND flash SSD drives 16

SSD capacity per drive & per

node16 * 300 GB = 4.8 TB

SMP SUPER-NODE (VIA VSMP)

Compute nodes / I/O Nodes 32 / 2

Addressable DRAM 2 TB

Addressable memory including flash

11.6 TB

GORDON (AGGREGATE)

Compute Nodes 1,024

Compute cores 16,384

Peak performance 341 TF

DRAM/SSD memory 64 TB DRAM; 300 TB SSD

INFINIBAND INTERCONNECT

Architecture Dual-Rail, 3D torus

Link Bandwidth QDR

Vendor Mellanox

LUSTRE-BASED DISK I/O SUBSYSTEM (SHARED)

Total storage: current/planned 4 PB/6 PB (raw)

Total bandwidth 100 GB/s



PACE: Closing the gap between Government, Industry and Academia

PACE is a non-profit, public

educational organization

oTo promote, educate and innovate

in the area of Predictive Analytics

oTo leverage predictive analytics to

improve the education and well

being of the global population and

economy

oTo develop and promote a new,

multi-level curriculum to broaden

participation in the field of

predictive analytics

Predictive Analysis

Center of Excellence

Inform, Educate and

Train

Develop Standards

and Methodology

High Performance

Scalable

Data Mining

Foster Research

and Collaboration

Data Mining Repository of Very Large Data Sets

Provide Predictive Analytics Services

Bridge the Industry and Academia

Gap



Foster Research and Collaboration

Predictive Analysis

Center of Excellence

Inform, Educate and

Train

Develop Standards and Methodology

High Performance

Scalable

Data Mining

Foster Research and Collaboration

Data Mining Repository of Very Large Data Sets

Provide Predictive Analytics Services

Bridge the Industry and Academia

Gap

• Fraud Detection• Modeling user behaviors• Smart Grid Analytics• Solar powered system

modeling• Microgrid anomaly

detection• Battery Storage Analytics• Sport Analytics• Manufacturing• Genomics



The City in a City

UCSD Microgrid



• 28,000 students and 28,000 faculty and staffRated 7th best public

University in Nation by U.S. News

Jacobs School of Engineering 4th in Bioengineering

• Peak load ~42 MW• Self Generation 30 MW• Solar PV ~1.5 MWSoitec CPV (27.5 kW)

UCSD Microgrid

• Fuel Cell 2.8 MW• Battery systems Second life (60 kWh)Peak shifting (30 kW/30

kWh)3 MW/6 MWH in Q3 2012

Internationally recognized microgrid



Microgrid’s Electrical One Lines



• Provides 85% heating, cooling and electricity for most of the UCSD Campus

• Cogeneration system consists of two 13.5 Megawatt natural gas-fired turbine generators

• These two gas turbines provide 27 Megawatts of electricity for the campus buildings

Central Utility Plant

• 60,000Lb/Hr of Steam per unit, totaling 120,000 Lb/Hr produced as available waste heat, which provides heating and cooling energy for the Chiller Plant and the Heating Plant

• Awarded 1 of 3 “Energy Star Awards” by EPA in 2010 for achieving 66% efficiency (LHV)



Gas Turbines



Central Utility Plant



The Central Utility Plant Saves the UCSD Campus

$850,000 Per Month!



• 1.2 MWs under PPA• 58 kW Demo sites• 63 kW Owned by campus• 190 kW Sustainable Comm. Program

Campus Solar Power

Currently have 1.5 MWs on campus



Price Center And Student Center



30 kW/30 KWH PV Integrated Storage System



30 kWh Energy Storage



UCSD’s Electrification of the Transportation Sector



• Molten Carbonate fuel cell

technology from FuelCell

Energy of Danbury CT

• CA’s first and only

“directed biogas” project

• Largest commercially

available design

• Commenced commercial

operations in Dec ‘11

2.8 MW Fuel Cell Demonstration

• Installed and operated under a PPA with BioFuels Energy LLC at a cost that is parity with the grid

• 10% of the campus’ baseload energy supply has a footprint the size of a tennis court



2.8 MW Fuel Cell Demonstration



August 14 Demand Response

• “SDG&E has notified the campus that energy demand reduction is needed between 1 and 6 p.m. today due to continued high temperatures taxing regional power reserves. Without demand reduction, SDG&E may be required to take actions to stabilize the utility grid

• Beginning at 1 p.m. today, Facilities Management will automatically reduce power demand by adjusting campus comfort cooling settings

• Heating, ventilation and air conditioning in office areas will go into "unoccupied" mode and spaces will be warmer or cooler than normal, depending on the space. Temperature set points in lab areas will range between 68 and 76 degrees. Airflow will not be affected”



Comparison between August 13 and August 14

Pow

er (k

W)

Time (24 hours)

August 13

August 14 Demand Response

Imports = Black; Total Consumption = Blue



DR: Stacked Area Chart Comparison

0

5000

10000

15000

20000

25000

30000

35000

40000

45000

Po

wer

Dem

an

d (

kW

)

Time (hour of day)

UCSD Power Demand (Imported and Generated)on August 13th and 14th

SDGE Imports

Estimated SolarProduction

Steam Turbine

Generator B

Generator A

Fuel Cell

August 13th 2012 August 14th 2012

Po

wer

Dem

an

d (

kW

)



Comparison with September 14

Po

wer

(k

W)

Time (hours)



Significance

• UCSD can reliably reduce load by 6 MW in 54 minutes

• Sufficient chilled water must be available to allow converting the steam chiller to a steam generator

• Can building energy response be improved by using existing HVAC data for efficiency?

• Building KPI Intern program • Summer of 2012



Why work on Building control systems

• As of 2010, the DoE reported that building energy consumption represents 41.1% of the U.S. Primary Energy Consumption which was greater than the industrial and transportation sectors

• A small improvement in building HVAC efficiency can result in large reduction in carbon emissions and lower costs of operations

• Building might be useful “control” devices in local area power systems, especially to help compensate for solar PV and EV charging system intermittency



•Measures performance of a building

•Annual energy per square foot

•Can include other variables:

Occupancy

Building type

Direction the windows face

Energy source

•Comfort KPI can include:

Amount of air flow

Zone temperature

Quality of air (CO2)

Key Performance Indicator



“You Can’t Manage What You Don’t Measure”

2

9



• Problem

Legacy naming conventions not documented

Multiple names for same object begin measured

• Approach

Use object oriented design of naming structure

Use IEC 61850 recommended names for leaf objects

Use ISO 50001 Building Automation Standards

• Goals

Set foundation for auto discovery and naming of BACnetsystems

Use PI AF data references to auto create PI tags with rigid structure

Create PI AF names and tagnames with fixed width components

Tag naming issues and approach



California UCSD WarrenComputer &

Science Engineering

Building

Floor 1Room 1106

HVAC Room Control Actual Cooling Setpoint

Actual Heating Setpoint

Actual Supply Air

Flow Setpoint

MeasurementActual

Supply Air Flow

Damper Position

Zone Temperature

Before

After

Naming Structure



Results of the Building Model

https://sc-osi-wp-2.ad.ucsd.edu:1325/Coresight/

https://sc-osi-wp-2.ad.ucsd.edu:1325/Coresight/



Calculated HVAC Power



Supply Air Flow & Damper Position



Computer Science & Engineering Building



Weekly Pattern Large Building Energy Usage



UCSD Smart Grid

• UCSD Smart Grid sensor network data set• 45MW peak micro grid; daily population of over 54,000 people

• Self-generate 92% of its own annual electricity load

• Smart Grid data – over 100,000 measurements/sec

• Sensor and environmental/weather data• Large amount of multivariate and heterogeneous data streaming from

complex sensor networks

• Predictive Analytics throughout the Microgrid



Big Data from Microgrids

• Complexities introduced by the large amount of multivariate and heterogeneous data streaming from complex sensor networks

• Extremely large, complex sensor networks, enabling a novel feature reduction method that scales well



4 V’s of Big Data

IBM, 2012



Predictive Analytics for Discovering Energy Consumption Patterns

• The utility and the consumer both benefit from consumption analytics

• Forecasting the energy consumption patterns in the UCSD campus microgrid

• Different spatial and temporal granularities

• Novel Feature Engineering

• Machine learning for demand response optimization



Predicting Energy Consumption

• Modeling the energy usage behavior of campus buildings is an important step to improving the efficiency of the system

• An accurate model of future energy usage enables:• anomalies discovery and point to wasteful usage of energy

• combat high future demands

• optimize pre-heating or pre-cooling scheduling

• maximize the energy efficiency

• help guide decisions regarding demand response, excess generation, renewable supply and load balancing and manage power outages



Multivariate Time Series Classification Approach

Metafeature Extraction

Global Feature Extraction

Feature Space Segmentation

Feature Space Abstraction

Intermediate Dataset Creation

Time Series Input X

Final Model Y

Step 2: Global Feature

Extraction

Step 3: Feature Space

Segmentation

Step 1: Metafeature

Extraction

Step 6: Standard

MineTool

Step 5: Intermediate

Dataset Creation

Step 4: Feature Space

Abstraction

MineTool-TimeSeries [Karimabadi, Sipes, et al



• Observed 3 MW

demand response from

buildings

• Observed 30 minute

time constant from

buildings

• Asked the question

- Can buildings be used

for microgrid control?

How can buildings be used as control variables?

• What effect will solar

PV have on grid

stability?

• What effect will fast

EV charging have on

grid stability?

• Can the Microgrid

deliver ancillary

services to the area

power system?



Solar Power Integration Through Predictive Analytics

Predictive

analytics

Solar + EV

integration

Business &

Societal change

Battery Storage

Analytics



Distributed Energy Generation

• Rapid growth of renewable generation capacity

• Photovoltaic (PV) generation on residential and commercial rooftops

• Interconnected in the electric distribution system, rather than at the high-voltage transmission level

• Population of plug-in hybrids and electric vehicles (EVs) is expected to grow substantially



Analytics/Visualization Examples



Power Generation Dashboard with PowerPivot



Power Generation Dashboard with Power View



Power Generation with Maps



Building Energy Consumption Analytics with PowerPivot



Building Energy Consumption Analytics with Maps



Building Carbon Emission Analytics with Power View



Building Carbon Emission Analytics with Power Pivot



Building Carbon Emission Analytics with Maps



Building Energy Consumption with GeoFlow Heat Map



Windows 8



Smart City San Diego



• Clean Tech San Diego, OSIsoft, SDG&E and UC San

Diego Common data infrastructure connects physical

assets: electrical, gas, water, waste, buildings,

transportation &traffic

• Platform to securely transfer high volumes of Big Data

from multiple, distributed measurement units

• Crowd-sourced Big Data in a cyber-secure, private cloud

• Predictive analytics on real-time time-series data

Sustainable San Diego Partnership

Predictive Analytics Center of Excellence

White House Big Data Event: “Data to Knowledge to

Action” – Launch Partners Award



Smart City (San Diego)

PI 2012 + SQL 2012 EE + SharePoint 2013

PI Visualization (DL, WP, CS,

PB)

AF, Asset Based Analytics

PowerView, PowerPivot,

GeoFlow

San Diego Gas &

Electric

(SDG&E)

SDSU UCSD

PI Cloud Connect

PI

PI Cloud Connect

Pipeline

Interface

Pipeline

Interface

Pipeline

Interface

Downtown Buildings

Azure IaaS (VM)Windows 8“Prius” dashboard

Power ViewBusiness Insight “drill down”



A common data infrastructure that connects all of the various physical assets/infrastructures

–Electrical

–Gas

–Water

–Transportation & Traffic

–Critical Facilities

–Communications

–Buildings

Why does a City need a Data Infrastructure?

Smart Infrastructure

Gas & Electric

Water & Wastewater

Waste Management

Critical Facilities & Buildings

Transportation & Traffic

Communications & Data Centers



PACE EducationEducating the new generation of Data Scientist

• Data mining Boot Camps

• Boot Camp 1

• April, 2014

• Boot Camp 2

• May 2014

• On-site training

• Tech Talks - 3rd Wednesday

• Workshops – Hadoop, PMML

• Summer Institute

• Institute for Data Science and Engineering



Mike Gualtieri's blog

http://blogs.forrester.com/mike_gualtieri

© Copyr i gh t 2014 OSIso f t , LLC.

Thank You!

• www.pace.sdsc.edu

• For further information, contact Natasha Balac [email protected]

63

http://www.sdsc.edu/

Predictive Analytics for Real-time Energy Management … managed by General Atomics ... Second life...

Documents

Transcript of Predictive Analytics for Real-time Energy Management … managed by General Atomics ... Second life...