Sensing & Sensibility of Energy g y gyUse in ‘Mixed-Used’ Buildings

Rajesh K. Gupta• Professor & Chair, Computer Science & Engineering

• Associate Director, California Institute for Telecommunications & Information TechnologyUniversity of California San Diego

Our Team

University of California, San Diego

Yuvraj Agarwal Rajesh Gupta Thomas Bharath Michael JohnSathya JawonSeemanta

Imagine…g

Imagine a small cityImagine a small city 12 000 acres 45 000 occupants 8 000 residents 12,000 acres, 45,000 occupants, 8,000 residents 2 hospitals (with local generation), 15 restaurants 450 buildings 11 million square fee of building space 450 buildings, 11 million square fee of building space Over $250M in capital construction/year Generates 80% of its own electricity usage including Generates 80% of its own electricity usage including

2.8 MW fuel cells, 1.2 MW PV, Wind, 15% of daily energy stored Meters & Monitors everything: Meters & Monitors everything:

50K meters, 4.5K thermostats 16 weather stations, real-time monitoring, , g,

tracks moving clouds across the campus to drive dynamic PV load shifts from 50 kW/sec to 1 kW/sec.

Self-regulating entity, its own police.

Energy Dashboardhttp://energy ucsd eduCurrently working to install the Energy Dashboard asCurrently working to install the Energy Dashboard as http://energy.ucsd.eduCurrently working to install the Energy Dashboard as 

part of the new  “SDGE Innovation Center” in San DiegoCurrently working to install the Energy Dashboard as part of the new  “SDGE Innovation Center” in San Diego

UCSD is Installing Zero Carbon EmissionSolar and Fuel Cell DC Electricity Generators

San Diego’s Point Loma Wastewater Treatment Plant Produces Waste

Methane

UCSD 2.8 Megawatt Fuel Cell Power Plant Uses Methane

2 Megawatts of Solar Power Cells

Being Installed

Localized Co-Generation and storage of energy on the UCSD microgridthe UCSD microgrid

Global ICT Carbon Footprint by Subsector

The Number of PCs (Desktops and Laptops) Globally is Expected to Increase from 592 Million in 2002

to More Than Four Billion in 2020

www.smart2020.org

Applying ICT – The Smart 2020 Opportunityfor Reducing GHG Emissions by 7 8 GtCO efor Reducing GHG Emissions by 7.8 GtCO2e

Smart IntelligentTransportation BuildingsTransportation

S tSmart Electrical

GridTelepresence

Recall total ICT 2020 emissions are 1.43 GtCO2e

Telepresence

Buildings are an important research focusg p All electricity in the US: 3,500 TWh

~500 power plants @7TWh Buildings: 2,500 TWhg , All electronics: 290 TWh

Bruce Nordman, LBNL

ACM B ildS

Buildings consume significant energy >70% of total US electricity consumption

ACM BuildSys

>70% of total US electricity consumption >40% of total carbon emissions

UCSD: A Living Laboratoryg y

Looking across 5 types of buildingsg yp g

more IT

From: Yuvraj Agarwal, et al, BuildSys 2009, Berkeley, CA.

Modern Buildings Are IT Dominated:g50% of peak load, 80% of baseload

Two Steps to Improving Energy p p g gyEfficiency

1. Reduce energy consumption by IT equipment Servers and PCs left on to maintain network presence Servers and PCs left on to maintain network presence Key Idea: “Duty-Cycle” computers aggressively SleepSever: maintains seamless network presence SleepSever: maintains seamless network presence

R d ti b th HVAC2. Reduce energy consumption by the HVAC system Energy use is not proportional to number of occupants Key Idea: Use real-time occupancy to drive HVAC y p y Synergy wireless occupancy node

14

Duty Cycling: Processors, HVACy y g ,

Why not power-down machines that are not working? Or power-down building HVAC systems

Runs into several use model problems Runs into several use model problems “Always ON” abstraction of the internet

Unlike light bulb ‘when not in room turn off the light’ Unlike light-bulb, when not in room, turn off the light Use model for the user/application and the

infrastructure are differentinfrastructure are different Network, enterprise system maintenance: distributed

control of duty-cycling has its own usability problems.control of duty cycling has its own usability problems.

Collaborating Processors

Fundamental Problem: Our Notions of Power StatesH t (PC ) ith A k (A ti ) Sl (I ti ) Hosts (PCs) are either Awake (Active) or Sleep (Inactive)

Power consumed when Awake = 100X power in Sleep! Users want machines with the availability of active machine power of Users want machines with the availability of active machine, power of

a sleeping machine.Somniloquy

Somniloquydaemon

Operating system,

Apps

Host PCSleepServers

Host processor,RAM peripherals etc

Operating system, including networking

stack

wakeupfilters

Appln. stubsRAM, peripherals, etc.

Secondary processorEmbedded

Embedded OS, including

networking stack

Network interface hardware

Embedded CPU, RAM,

flash

Augment the network interface of PCs

Objective: Make PCs responsive even when asleepM i t i il bilit th ti t l t k Maintain availability across the entire protocol stack E.g. ARP(layer 2), ICMP(layer 3), SSH (Application layer)

Without making changes to the infrastructure or user Without making changes to the infrastructure or user behavior

Active State : >140 W Secondary Processor is: Idle State : 100 W Sleep State : 1.2 W

y• Functionally similar – masquerades as the host • Much lower power

Network Interface

Secondary Processor (Power in active state ~1W)

+Low Power

+DRAM

+Flash MemoryLow Power

CPUDRAM Flash Memory

Storage

200n Host Only Somniloquy Stateful applications:

Web download “stub” on the gumstix

100

150

nsum

ption

atts) 

on the gumstix

200MB flash, download when Desktop PC is

l

0

50

ower Con

(Wa

1 600 1200 1800 2400

asleep

Wake up PC to upload data whenever needed

1 601 1201 1801 2401P

Time (seconds)92% less energy than using host PC.data whenever needed

Increase battery life from <6 hrs to >60 hrs

Somniloquy exploits heterogeneity to save power and maintain availability

SleepServers for Enterprises: Architecture

Respond: ARPs ICMP DHCPRespond: ARPs, ICMP, DHCPWake-UP: SSH, RDP, VoIP callProxy: Web/P2P downloads, IM

Average Power 96 Watts

Average Power 26 Watts

DE Total estimated Savings for CSE (>900PCs) : $60K/year

Deployed SleepServers across 50 usersEnergy Savings: 27% - 85% (average 70%)gy g ( g )

Scenario: CSE Energy Use Reductionsgy

• Deploy Somniloquy / Sleepserver80 kBTU/ft2

p y q y / p– Machine room      : 142 kW  71 kWPC Plug loads : 130 kW 70 kW– PC Plug loads        : 130 kW  70 kW

• Ventilation system:– New fans, chillers : 65 kW  52 kW

• Lighting:Lighting:– Fluorescent lighting  LED– Motion‐detector controlled hallway lighting evenings & weekends: 50 kW  11 kW

42 kBTU/ft2

21

Could CSE become a ZNEB?

• Solar energy          : 2700 m2 roof 111 kBTU/ft2

• Solar PhotoVoltaic: 20% efficient 22 kBTU/ft2• Solar PhotoVoltaic: 20% efficient         22   kBTU/ft2 

• How do we achieve 42 kBTU/ft2 ?T ki l PV dd 30% i di 28 kBTU/f 2– Tracking solar PV           : add 30% irradiance           28 kBTU/ft2

– Increase PV efficiency  : 29% efficient   42 kBTU/ft2

22

Dramatic improvements in energy efficiency and solar conversion efficiency needed for ZNEB

Wait for global warming orWait for global warming or better solar cells?

Is that it?Is that it?

Buildings 2.0: Occupancy‐Driven Smart BuildingsBuildings 2.0: Occupancy Driven Smart BuildingsUse occupancy and activity to drive energy efficiency in HVAC system usage.

Red ced cooling hen a

Increased HVAC when a h

Reduced cooling when a room is empty.

room has more occupants.

OccupancyPerformabilityPerformabilityAdaptive Envelope

24

When there are less people in the room, reduce cooling. When there are more, increase cooling as required to maintain comfort.

Reducing HVAC energy consumption g gy p

• Modern buildings have efficient HVAC systemsModern buildings have efficient HVAC systems– Central cooling + chilled water loop is common

• Unfortunately, use of static schedules prevalent– Energy wasted during periods of low occupancy gy g p p y

HVAC   ON

5:15AM 6:30PMHVAC starts at this time Un-Occupied Periods

HVAC stops at this time

Some people actually arrive 2 hours later!

2525

Solution: Fine grained occupancy driven control! 

Occupancy Driven HVAC controlp y• Determining occupancy is challenging  

– Cost of deployment, maintenance – Accuracy of detection, privacy and security issuesAccuracy of detection, privacy and security issues

Key Design Requirements:Key Design Requirements: • Inexpensive (less than 10$) • Battery powered – 4‐5 year life • Multiple sensors for accuracy• Multiple sensors for accuracy

Synergy Occupancy Node • CC2530 based design  • 8051 uC + 802.15.4 radio• Zigbee compliant stack 

2626

gbee co p a s ac• PIR + Magnetic reed switch   

Accuracy of Occupancy Detection y p y• Over 96% occupancy accuracy with Synergy node 

2727

Occupant Diversityp y

2828

Occupancy Driven HVAC controlp y

Synergy Occupancy Node Key Design Requirements: • Inexpensive (less than 10$) • Battery powered – 4‐5 year life

Synergy Occupancy Node • CC2530 based design  • 8051 uC + 802.15.4 radio

Zi b li k

2929

Battery powered  4 5 year life • Multiple sensors for accuracy• Zigbee compliant stack 

• PIR + Magnetic reed switch   

30

Deployment across 2nd floor of CSEp y

Floormap: 2nd Floor

‐ 50 Offices, 20 Labs. 8 S B S i

p

‐ 8 Synergy Base Stations

Control individual HVAC zones based on real‐time occupancy information!

3131

32

Implementation: Interfacing with the EMSp g

Occupancy Data Occupancy Data Analysis ServerAnalysis Server (ODAS)

Database

Occupancy Data Analysis Server• Database to store mapping • MetaSys EMS – proprietary protocols

l hWindows Server with OPC Tunneller

Database

Sheeva Plug base stations

Occupancy nodes

D t b

• OPC tunnel to communicate with EMS• Actuation based on modifying status 

for individual thermal zones

BACnet OPC DA ServerHVAC

Control

Database • Use priorities levels  ‐‐ co‐exist with current campus policies. 

• Occupancy data not visible externally NAE NAE

Control

Metasys ADXNAE …p y y

Worked directly with UCSD Physical Plant Services (PPS): gain access to the campus EMS and to actuate 

33

it correctly. 33

HVAC Energy Savingsgy g

HVAC Energy Consumption (Electrical and Thermal) during the baseline day.

HVAC Energy Consumption (Electrical and Thermal) for a test day with a similar weather profile. HVAC energy savings are significant: over 13% (HVAC-Electrical)

d

Estimated 50% savings if deployed across entire CSE!

and 15.6% (HVAC-Thermal) for just the 2nd floor

34

g p yDetailed occupancy can be used to drive other systems.

34

Effect on Comfort  

Evaluating the worst case • Direct sun facing room• Heavy IT load: multiple PCs • Comfort Policy for Cooling y g

• Turning HVAC on at 77F • Turn off HVAC at 71F 

• Our focus is on energy savings• Our HVAC control included safeguard measures

M i d i i f d d it– Maximum and minimum enforced despite occupancy

• Results: Latency to cool room is within minutes

35• Investigating better policies for comfort + energy

35

Recapp

Embedded systems when scaled up naturallyEmbedded systems when scaled up naturally become Cyber-physical systems

o Societal use of energy is a natural culmination of our focus on energy efficiency in electronic systems

When it comes to energy use focus on theWhen it comes to energy use, focus on the long poles in the tent

o Enclosures (Buildings, data center) are a great starting point foro Enclosures (Buildings, data center) are a great starting point for reducing energy use, improving its quality.

IT is a big part of the solution to incorporate optimizations into building energy use.

W i l t f i i th b d idth fo We are in early stages of increasing the bandwidth of power consumption by IT.

Only a beginning: Sustainability 2.0y g g y

37

Some (Recent) Pointers( )• “Evaluating the Effectiveness of Model‐Based Power Characterization”, USENIX 

Advanced Technical Conference (ATC), 2011.• "Duty‐Cycling Buildings Aggressively: The Next Frontier in HVAC Control" , 

ACM/IEEE IPSN/SPOTS, 2011.• "Occ panc Dri en Energ Management for Smart B ilding A tomation"• "Occupancy‐Driven Energy Management for Smart Building Automation" , 

ACM BuildSys 2010.• "SleepServer: A Software‐Only Approach for Reducing the Energy Consumption 

of PCs within Enterprise Environments" , USENIX ATC, 2010.• "Cyber‐Physical Energy Systems: Focus on Smart Buildings" , DAC 2010.• "The Energy Dashboard: Improving the Visibility of Energy Consumption at a• "The Energy Dashboard: Improving the Visibility of Energy Consumption at a 

Campus‐Wide Scale“, ACM BuildSys 2009.• "Somniloquy: Augmenting Network Interfaces to Reduce PC Energy Usage" , 

NSDI 2009.

3838

Thank You

An exciting time to be doing research indoing research in

embedded systems with tremendous potential to solve p

society’s most pressing problems.

Rajesh Gupta

gupta@ucsd.edu