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
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
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