Visible to the public File preview

Working together to put CPS Technology on the Sustainability Table

David
 E.
 Culler
 
 
 
 
  University
 of
 California,
 Berkeley
  August
 2,
 2011
 
 

 

What can we put on the table?

CPS-PI-11

8/2/2011

2

Sustainability?
§  “Sustainable
 development
 should
 meet
 the
 

needs
 of
 the
 present
 without
 compromising
  the
 ability
 of
 future
 generations
 to
 meet
 their
  own
 needs”
 
–  Our
 Common
 Future,
 World
 Commission
 on
 Environment
  and
 Development,
 United
 Nations,
 1987
 

CPS-PI-11

8/2/2011

3

Quantifying it - California Law
§  AB
 
 32
 
 
 
ú  Reduce
 
 GHG
 
 emissions
 
 to
 
 
 1990
 
 levels
 
 by
 
 2020
 
 
 
 
 
 

§ 
 Governor’s
 
 executive
 
 order
 
 
 
 S-­‐3-­‐05
 
 (2005)
 
 

§  Renewable
 Portfolio
 Standard
 

ú  80%
 
 reduction
 
 below
 
 1990
 
 
 levels
 
 by
 
 2050
 
 
 
 

ú  33%
 renewables
 by
 2020,
 
 20%
 biopower
  procurement
 

§  480
 
 =>
 
 80
 
 
 mmT
 
 CO2e
 
 in
 
 40
 
 years
 
ú 
 Population
 expected
 to
 grow
 from
 37
 =>
 55
 million
  ú  Economic
 growth
 
4

CPS-PI-11

8/2/2011

CPS-PI-11

8/2/2011

5

What it amounts to …
Historical
 and
 BAU
 Emissions
 
1,000
  900
  800
  700
  600
  500
  400
  300
  200
  100
  0
  1990
  2005
  2020
  2050
  Historical
  BAU
 

GHG
 Emissions
 (MtCO2e/yr)
 

25
 

Per
 Capita
 CO2
 Emissions
 

Metric
 tons
 CO2
 per
 capita
 per
 year
 

Energy
  emissions
  Non-­‐ energy
  emissions
  2020
  Target
 

20
 

15
 

14.62
  12.93
  9.88
  23.4
 

10
 

5
  1.56
  CA
 2005
  CA
 2050
 Target
  CA
 2020
 Target
  CA
 1990
  US
 2003
  0
 

CPS-PI-11

8/2/2011

6

Can we do this?

CPS-PI-11

8/2/2011

7

But

CPS-PI-11

8/2/2011

8

My Take …
§  We
 can’t
 build
 our
 way
 to
 sustainability
  §  The
 discussion
 is
 focused
 on
 the
 Physical
 
ú  New
 Windows,
 Materials,
 Motors,
 Biofuels,
 …
 

§  Need
 to
 make
 the
 best
 of
 what
 we
 have,
 or
 

will
 have,
 
 and
 how
 we
 use
 it…
  §  This
 takes
 observation,
 intelligence
 and
  control
  ⇒ Demonstrable
 CPS
 technology
 on
 the
 table
  ⇒ Efficiency
 &
 Supply-­‐Following
 Loads
  ⇒ 10
 years
 to
 innovate,
 30
 years
 to
 scale
 
CPS-PI-11 8/2/2011

9

What’s on the Table? – buildings

CPS-PI-11

8/2/2011

10

4 Part GHG Reduction Plan
§  Efficiency
 

CPS-PI-11

8/2/2011

11

4 Part GHG Reduction Plan
§  Efficiency
  §  Electrify
 

CPS-PI-11

8/2/2011

12

4 Part GHG Reduction Plan
§  Efficiency
  §  Electrify
  §  Decarbonize
 

the
 electricity
  §  Decarbonize
  the
 fuel
 

CPS-PI-11

8/2/2011

13

4 Part GHG Reduction Plan
§  Efficiency
  §  Electrify
  §  Decarbonize
 

the
 electricity
  §  Decarbonize
  the
 fuel
 

CPS-PI-11

8/2/2011

14

All required for even 60% reduction

but still fall short of 80%
CPS-PI-11 8/2/2011

15

Efficiency

CPS-PI-11

8/2/2011

16

Electrification

CPS-PI-11

8/2/2011

17

Low Carbon Electricity Options

CPS-PI-11

8/2/2011

18

Build Rate

CPS-PI-11

8/2/2011

19

The Renewables are there

CPS-PI-11

43% agriculture, 3.4% urbanized

8/2/2011

20

The Problem: Supply-Demand Match
Baseline + Dispatchable Tiers Oblivious Loads

Generation

Transmission

Distribution

Demand

CPS-PI-11

8/2/2011

21

Load-following Supply (?)
Growing proportion of renewables leads to higher price volatility. October 2008 to March 2010: >90 hours with negative prices; highest price reached: +€500/MWh, lowest -€500/MWh
494.26
330 300 270 240 210 180 150 120 90 60 30 0

01.10.2008

01.11.2008

01.12.2008

01.01.2009

01.02.2009

01.03.2009

01.04.2009

01.05.2009

01.06.2009

01.07.2009

01.08.2009

01.09.2009

01.10.2009

01.11.2009

01.12.2009

01.01.2010

01.02.2010

-30

€/MWh

-60 -90

-120 -150

-500.02 Daily minimum price (indicated in red when negative)

Daily maximum price
Source: EEX spot prices.

CPS-PI-11

8/2/2011

22

01.03.2010

… and in CA

CPS-PI-11

8/2/2011

23

Zero Emissions Load Balancing (ZELB)
§  Just
 the
 emissions
 from
 the
 natural
 gas
 to
 

firm
 the
 33%
 renewables
 exceeds
 GHG
 target
 

§  Even
 with
 50%
 with
 natural
 gas
 &
 50%
 with
 

some
 yet-­‐to-­‐exist
 storage
 

CPS-PI-11

8/2/2011

24

We Understand Supply

CPS-PI-11

8/2/2011

25

Not demand

CPS-PI-11

8/2/2011

26

Limits to Renewable Penetration
§  Variability,
 Intermittency
 of
 Supply
  §  Visibility
 into
 Availability
 of
 Supply
  §  Ability
 of
 Loads
 to
 Adapt
  §  Algorithms
 and
 Techniques
 for
 Reactive
 Load
 

Adaptation
 

§  Capability
 of
 the
 Infrastructure
 to
 maintain
 

the
 match
 

CPS-PI-11

8/2/2011

27

CPS and the 4 Part GHG Reduction Plan
§  Efficiency
  §  Electrify
  §  Decarbonize
  §  Decarbonize
 
Monitoring,
 Analysis,
 Modeling,
  Waste
 Elimination,
 Power
  Proportionality,
 Optimal
 Control
  Intelligence,
 Communication,
  adaptation
 in
 Everything
  ZELB.
 Supply-­‐Following
 Loads,
  Energy
 SLA,
 Cooperative
 Grid
 
8/2/2011

the
 electricity
  the
 fuel
 

CPS-PI-11

28

CPS Technology …
§  National-­‐scale
 Physical
 Information
 Service
 
ú  Data
 collection,
 streaming,
 archiving,
 querying
  ú  Modeling,
 Analysis,
 Control
  ú  Representation,
 metadata,
 life-­‐cycle
  ú  Use
 others’,
 contribute
 yours
 

CPS-PI-11

8/2/2011

29

Some starts

CPS-PI-11

8/2/2011

30

sMAP: Uniform Access to Diverse Physical Information
Applications
Modeling
  Control
  Storage
  Visualization
  Location
  Debugging
  Personal
  Feedback
  Continuous
  Commissioning
 

Actuation
  Authentication
 

JSON
 Objects
  REST
 API
 

sMAP
 
Physical Information
Electrical
  Geographical
  Occupancy
 

Water
 

HTTP/TCP
  …
 
 

Structural
 

Weather
 

Environmental
 

Actuator
 

CPS-PI-11

8/2/2011

31

sMAP ecosystem
sMAP
 Resources
 
California
 ISO
  sMAP
 Gateway
 

Applica?ons
 

sMAP
  sMAP
 
AC
 plug
 meter
  Vibra?on
 /
 Humidity
  Proxy
 Server
 

Database
 

sMAP
 
Weather
 

Google
 PowerMeter
 

EBHTTP
 /
 IPv6
 /
 6LowPAN
  Wireless
 Mesh
 Network
  sMAP
  sMAP
 

EBHTTP
  Transla?on
 

Edge
 Router
 

Internet
 

Every
 Building
  Light
 switch
  Temperature/PAR/TSR
  Dent
 circuit
 meter
 

sMAP
 

sMAP
 

Modbus
 
sMAP
 Gateway
 

RS-­‐485
 
sMAP
 Gateway
  Cell
 phone
 

CPS-PI-11

8/2/2011

32

Cory CEC B2G Testbed
DOP HVAC

MCL equip

Whole Bldg

MCL infra Central vent

MCL vac

office HVAC

servers Plug loads
CPS-PI-11

Lighting Parking Lot
8/2/2011

inst Lab 199 HVAC

33

DOE MELS => Appliance Energy

LBNL Bldg 90 611 of 1200 loads

CPS-PI-11

8/2/2011

34

DOE/UCB/Siemens Auto Demand Response Sutardja Dai Hall
400
  350
  300
  Power
 (kW)
  250
  200
  150
  100
  50
  0
  Lights
 &
 Plugs
  Servers
 &
  Cooling
  Nano
 Fab
  HVAC
  Other
 

§  www.openbms.org
 
CPS-PI-11 8/2/2011

35

BACNet => sMAP
Sensors, Actuators Controllers

Sutardja Dai Hall
Technical Specification Sheet Rev. 9, August 2003

Internet
Modular Building Controller Siemens
Features
Modular hardware components to match equipment to initial control requirements while providing for future expansion Modular, snap-in design simplifies installation and servicing Transparent viewing panels on the enclosure door to view the status indicator LEDs and override switch positions Integration platform for communications and interoperability with other systems and devices Proven program sequences to match equipment control applications Advanced Proportional Integral Derivative (PID) loop tuning algorithm for HVAC control to minimize oscillations and guarantee precise control Built-in energy management applications and DDC programs for complete facility management

P2 over RS-485

safety, security, and lighting). Up to 100 modular field panels communicate on a peer-to-peer network.

Remote Login

sMAP

Figure 1. Modular Building Controller.

Description
The Modular Building Controller (MBC) is an integral part of the APOGEE® Building Automation System. It is a high performance, modular Direct Digital Control (DDC) supervisory field panel. The field panel operates stand-alone or networked to perform complex control, monitoring and energy management functions without relying on a higher level processor. The MBC provides central monitoring and control for distributed Floor Level Network (FLN) devices and other building systems (e.g., chiller, boiler, fire/life

Apogee BACnet/IP Server

Gateway

Comprehensive alarm management, historical data trend collection, operator control and monitoring functions Support for peer-to-peer communications over Industry standard 10/100 Base-T TCP/IP networks.

Modem
8/2/2011

CPS-PI-11

36

Data from 1 Modern Building
§  1358
 control
 settings
  §  2291
 meters/sensors
 
ú  Set
 points,
 Relays
 (lights,
 pumps,
 etc),
 Schedules
  ú  Power
 (building,
 floor,
 lights,
 chiller,
 pumps,
 etc)
     Current,
 voltage,
 apparent,
 real,
 reactive,
 peak
  ú  Temp
 (rooms,
 chilled
 water,
 hot
 water)
  ú  Air
 volume
  4+ million Commercial, ú  Alarms,
 Errors
  ú  Dampers,
 valves,
 min/max
 flow,
 fan
 speed,
 PID
 

§  2165
 control
 outputs
 
parameters
 

110+ million Residential

§  72
 other
 

CPS-PI-11

8/2/2011

37

Current sMAP demography
Name
  Cory
 Hall
 Submetering
  Cory
 Hall
 Metering
  Cory
 Lab
 Temperature
  Cory
 Lab
 Machines
  Cory
 Chilled
 Water
  Cory
 Weather
  Soda
 Hall
 Sun
 Blackbox
  Soda
 Lab
 Machines
  Soda
 Lab
 Panel
  Soda
 SCADA
 Data
  LBNL
 Building
 90
  Campus
 Power
 Data
  Citris
 SDH
 BACnet
  Brower
 Buildinbg
  Sensor
 Type
  Dent
 Three-­‐Phase
  ION
 and
 pQube
 Meters
  TelosB
  ACme
  HeatX
 
  Vaisala
 WXT520
  Fan
 speed;
 environmental
  ACme
  Veris
 E30
  Barrington
 controls
  Acme
  Obvius
 Aquisuite;
 various
  Siemens
 Apogee
  Johnsons
 Control
  Physical
 Layer
  Modbus
 +
 Ethernet
  HTTP/Ethernet
  802.15.4
 +
 Ethernet
  804.15.4
 +
 Ethernet
  Modbus
 +
 Ethernet
  SDI-­‐12
 +
 Ethernet
  Http/Ethernet
  802.15.4
 +
 Ethernet
  Modbus
 +
 Ethernet
  RS-­‐485/various
  802.15.4
 +
 Ethernet
  XML/HTTP/Ethernet
  BACnet/IP
 +
 RS-­‐485
  BACnet
  Sense
 Points
  79
  3
  4
  8
  1
  1
  10
  40
  1
  “1”
  550
  4
  “1”
  “1”
  Channels
  3318
  150
  8
  16
  11
  11
  84
  80
  42
  1670
  1650
  100
  600+
  1500
 

CPS-PI-11

8/2/2011

38

A Stream Engine
2000 1800 1600 1400

Number of streams

1200 1000 800 600 400 200

insert

sMAP
 

0 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Compression achieved

~300k pts/sec

resample aggregate

query

Time-­‐series
 Interface
 
  Bucketing
  RPC
  Key-­‐Value
 Store
  Page
 Cache
  Lock
 Manager
  Storage
 Alloc.
  Compression
  Storage
  mapper
  SQL
  MySQL
 

CPS-PI-11

8/2/2011

39

readingdb

streaming pipeline

sMAP
 

Scaling Out in the Cloud
Columnar Data Storage Query Translator SQL è MR

Map-Reduce!

Stream Management System!

Hadoop!
HBase HDFS Hive

Data Source

Data Processing

Data Sink

Analytics Workloads!

Real Time Monitoring!

Interactive Workloads!

CPS-PI-11

8/2/2011

40

15,000 pts and growing…

CPS-PI-11

8/2/2011

41

CPS Technology
§  National-­‐Scale
 Physical
 Info
 Service
  §  Software
 foundations,
 platforms
 and
 

solutions
 for
 Energy
 Efficiency
 and
 Agility
 

CPS-PI-11

8/2/2011

42

Stages of Energy Effectiveness
§  Waste
 Not
 
ú  Do
 Nothing
 Well
 !!!
 

§  Power
 Proportionality
 
 
ú  Peak
 Performance
 :
 Power
 =>
 Safety
  ú  Optimize
 Partial
 Load
 -­‐
 from
 nothing
 to
 peakl
 

§  Sculpting
 
ú  Identify
 the
 energy
 slack
 and
 utilize
 it
 

§  Negotiated
 Grid
 /
 Load
 /
 Human
 Interaction
 
ú  Plan,
 Forecast,
 Negotiate,
 Manage
 

CPS-PI-11

8/2/2011

43

Our Buildings
Lighting HVAC IT and Plug Load PDUs, CRACs Servers

8/2/2011

CPS-PI-11

44

PowerProportional Buildings ?
Cory Hall: Office + Semiconductor + IT

950 KW

1150 KW

Min = 82% of Max

CPS-PI-11

8/2/2011

45

PowerProportional Buildings ?
1.45 MW Stanley Hall:
Office + BioScience - 13 NMRs

2.02 MW

Min = 72% of Max

CPS-PI-11

8/2/2011

46

PowerProportional Buildings ?
LeConte Hall: Office

202 KW 62 KW

Min = 31% of Max

CPS-PI-11

8/2/2011

47

Re-Flash the HVAC …
Whole Bldg

LoCal + ActionWebs

inst Lab 199 HVAC

CPS-PI-11

8/2/2011

48

Learning-based Model Predictive Control
§  Mathematical
 model
 from
 Newton’s
 law
 of
 cooling
 
time constant of room change in temperature over time AC cooling weather heating from occupants and equipment

dT/dt = -krT - kcu(t) + kww(t)) + q(t)

§  Model
 identified
 using
 semi-­‐parametric
 regression
 
23

Temperature: Experimental (blue) Simulated (red)

(°C)

22 21 11AM 12PM 1PM 2PM 3PM Time 4PM 5PM 6PM 7PM 8PM

Heating from occupants and equipment

(°C/s)

0.029 0.0285 11AM 12PM 1PM 2PM 3PM Time 4PM 5PM 6PM 7PM 8PM

(Aswani, Master, Taneja, Culler, Tomlin, Proc. of IEEE, 2011)
CPS-PI-11

8/2/2011

49

Living Lab…
§  LBMPC
 saved
 
ú  30%
 energy
 when
 hotter
 outside
  ú  70%
 energy
 when
 cooler
 outside
 

§  Standard
 MPC
 was
 inconsistent
  §  Two-­‐position
 control
 was
 inefficient
 
Method
 
Measured
  Energy
  Estimated
  Energy
  Tracking
  Error
  Temperature
  Variation
  Average
  External
  Load
 

Two-­‐Position
  LBMPC
  Control
  Linear
 MPC
  Experiment
  Two-­‐Position
  LBMPC
  Experiment
  LBMPC
  Linear
 MPC
  Two-­‐Position
 
CPS-PI-11

23.6kWh
  30.5kWh
  32.6kWh
  11.8kWh
  35.1kWh
  13.3kWh
  8.6kWh
  34.5kWh
 

0.75°C
  0.62°C
  0.61°C
  0.86°C
  0.93°C
  0.55°C
 

0.13°C
  0.30°C
  0.20°C
  0.17°C
  0.21°C
  0.19°C
 
50

11.0°C
  11.0°C
  11.0
 °C
  7.2°C
  7.2°C
  7.2°C
 

8/2/2011

LBMPC to Minimize Energy…
§  Estimates
 heating
 load
 using
 model
 

§  Best
 control
 that
 considers
 estimated
 occupancy
 
Cost
Temperature Dynamics

Temperature Constraints AC Constraints

CPS-PI-11 8/2/2011 (Aswani, Master, Taneja, Culler, Tomlin, 2011, submitted)

51

… or optimize to follow the supply

Generation Transmission

VPG

Distribution Load

CPS-PI-11

8/2/2011

52

at Building-Scale, and beyond

Temperature Blue – Measured; Red - Simulated
1 24 24 2 24 3

Heating from occupants, equipment, etc.
1 1.9 0.95 0.9 1.85 1.4 1.8
°C °C

2 1.45

3

°C

°C

°C

°C

22

22

22

0.85 0.8 0.75 0.7

1.75 1.7 1.65 1.6

1.35 1.3 1.25 1.2 Tu W Th F Sa Su M Tu W Th F Sa Su Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

Tu W Th F Sa Su M Tu W Th F Sa Su

4 24 24

5 24

6

4 1.65 1.8 1.75 1.6 1.55
°C

5 1.6 1.55 1.5
°C

6

°C

°C

°C

°C

22

22

22

1.7 1.65 1.6

1.5 1.45 1.4

1.45 1.4 1.35 Tu W Th F Sa Su M Tu W Th F Sa Su Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

Tu W Th F Sa Su M Tu W Th F Sa Su

7 24 24

8 24

9

7 1.65 1.6 1.75 1.7 1.65
°C

8 1.55 1.5
°C

9

°C

°C

°C

°C

22

22

22

1.55 1.5 1.45

1.45 1.4 1.35 1.3

1.6 1.55 1.5

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

20

Tu W Th F Sa Su M Tu W Th F Sa Su

1.4

Tu W Th F Sa Su M Tu W Th F Sa Su

Tu W Th F Sa Su M Tu W Th F Sa Su

Tu W Th F Sa Su M Tu W Th F Sa Su

CPS-PI-11

8/2/2011

53

Making Sense out of Sensors
° ° ° ° ° °

° ° °
CPS-PI-11 8/2/2011

54

CPS-PI-11
55
  75
  70
  60
  90
  80
  1
  64
  127
  190
  253
  316
  379
  442
  505
  568
  631
  694
  757
  820
  883
  946
  1009
  1072
  1135
  1198
  1261
  1324
  1387
  1450
  1513
  1576
  1639
  1702
  1765
  1828
  1891
  1954
  50
  65
  85
 

Supply Air Fan VFD

OAT
 

Input
 

Output
 

8/2/2011

55

What’s really going on?
§  Bring
 comfortable
 air
 from
 outside
  §  Cool
 it
 down
 till
 its
 really
 cold
  §  Push
 it
 out
 everywhere
 thru
 VAVs
 that
 are
 at
 

minimum
 opening
  §  Reheat
 it
 to
 set
 point
  §  So
 the
 empty
 rooms
 will
 be
 comfortable
 
§  This
 is
 going
 on
 everywhere!
 

§  And
 we
 supply
 perfect,
 precious
 energy
 to
 do
 it!
 
CPS-PI-11 8/2/2011

56

And IT is no better …

Atom 333
350

Westmere
Server Power Consumption
48 300 250 200 150 230 100 50 0
PowerEdge 1850 Dell PowerEdge 1950 SunFire x2100 Cyber Switching SunFire V60x

87

Active Idle

15 13 13 14 19 287 248 190 190 200 161

Watts

SunFire X2200

CPS 2011

Core i7

4/12/11

57

HP Integrity rx2600

Compaq DL360

Supply-Following Computational Loads Background Processing (shiftable)

Requests Availability Forecasts

IPS

QoS (fidelity & latency)
Power

Controllable Storage
CPS-PI-11 8/2/2011

58

Energy-Availability Driven Scheduling
Non-dispatchable, variable supply
14 12 10 x 10
4

Power (W)

8 6 4 2 0

0

200

400 Time (hrs)

600

800

Pacheco wind farm
x 10 5
4

x 10 5 4

4

greedy pacheco 2.0x

Power (W)

Power (W)
0 200 400 Time (hrs) 600 800

4 3 2 1 0

3 2 1 0

Power proportional, grid-aware loads

200

250 Time (hrs)

300

Scientific computing cluster
CPS-PI-11
NREL Western Wind and Solar Integration Study Dataset 59 8/2/2011 http://wind.nrel.gov/Web_nrel/

Scheduling & Energy Storage

CPS-PI-11

8/2/2011

60

CPS Technology
§  National
 Physical
 Info
 Service
  §  Software
 foundations
 for
 Energy
 Efficiency
 

and
 Agility
  §  Infrastructure
 for
 Energy
 Innovation
 

CPS-PI-11

8/2/2011

61

Personalized Automated Lighting Control
HTTP

Python Control Process MySQL Python Django

sMAP

BACnet

§  Three
 controllable
 

Gateway
CPS-PI-11

ballasts
 per
 fixture
  §  ~5
 zones
 per
 floor
 
8/2/2011

62

Real Energy Savings

SDH
 4th
 Floor
 Lighting
 Energy
 Usage
 
4
  3.5
  3
  2.5
  KW
  2
  1.5
  Colab
 Savings
  1
  Floor
 kW
  0.5
  Collab
 kW
  0
  0
  0
  58%
  53%
  68%
  69%
  63%
  65%
  68%
  75%
  80%
  71%
  70%
  61%
  60%
  50%
  40%
  30%
  20%
  10%
  0%
 

CPS-PI-11

8/2/2011

63

From the agony of load-following …

CPS-PI-11

8/2/2011

64

to Co-operative Energy Mgmt in a
Cyber-Physical Grid
•  Availability
•  Pricing •  Planning
Source IPS energy subnet Intelligent Power Switch Load IPS

•  Forecasting
•  Tracking •  Market

•  Monitor, Model, Mitigate
•  Deep instrumentation •  Power proportional Design •  Energy-Agile Control •  Shifting, Scheduling, Adaptation
CPS-PI-11 8/2/2011

65

Thanks

CPS-PI-11

8/2/2011

66