Embed Size (px)
Citation preview
LoCal: Rethinking the Energy Infrastructure using Internet Design
Principles
David Culler, Randy Katz, Eric Brewer, Seth Sanders
University of California, Berkeley
LoCal Kickoff5 October 2009
“Energy permits things to exist; information, to behave purposefully.” W. Ware, 1997
Kickoff Agenda
• 1000-1200: Project Overview
• 1200-1300: Experimental Plans (and Working Lunch)
• 1300-1400: Student Poster Session
2
LoCal Beginnings …
• Culler: Need to find information age solutions to THE industrial age problem – Energy
• Katz: The Grid/The Internet: two national scale infrastructures with fundamentally different design principles
• How might we design the Grid in when the Internet exists?
3
The Grid: Marvel of Industrial Age Design
• Deliver high quality low-cost power• To millions of customers over thousands of
miles• Synchronized to 16 ms cycle (60 Hz) • With no orders, no forecasts, no plans• No inventory anywhere in the supply chain
• To enable rapid economic & industrial growth through oblivious consumption
4
A New Reality …
1. Energy becoming increasingly dear– increased cost of acquisition– inclusion of environmental costs
2. Improvements in energy efficiency cause high dynamic variability in the load– high peak-to-ave ratio, bursty
3. Limitations of existing grid present transmission and distribution bottlenecks
4. Incorporation of renewable resources reduces control over supply– most are non-dispatchable (solar, wind)
5
Towards an “Aware” Energy Infrastructure
6
Baseline + Dispatchable Tiers
DistributionTransmissionGeneration Demand
Nearly Oblivious Loads
Non-Dispatchable Sources Interactive
Dispatchable Loads
Communication
Communication
Energy Reduction and Support for Renewables thru Information
DispatchableSupply
Non-Dispatchable
Supply
DispatchableSupply
Non-Dispatchable
Supply
Doing Nothing Well
Scheduling
Storage
Reduce Demand
Increase Effectiveness of Non-Dispatchable
Supply
Energy Network Architecture
• Information exchanged whenever energy is transferred
• Loads are “Aware” and sculptable– Forecast demand, adjust according to
availability / price, self-moderate
• Supplies negotiate with loads
• Storage, local generation, demand response are intrinsic
8
Where to Focus?
• Buildings …• 72% of electrical consumption,
40% of total consumption, 42% of GHG footprint
• 370 B$ in US annual utility bill• 9.5% of GDP in bldg
construction/renovation• Primarily Coal generation• Primary opportunity for
renewable supplies9
Renewable energy consumption
Electricity source
Coal consumption by sector
04/18/23 10
Supply Demand
Figure Courtesy Professor Arun Majumdar, UCB, LBNL
04/18/23 11
Envi
ronm
enta
lO
pera
tiona
l
Start from Scratch?
• No!
12
Grid Exists
13
Conventional Electric Grid
Generation
Transmission
Distribution
Load
Internet Exists
14
Conventional Electric Grid
Generation
Transmission
Distribution
Load
Conventional Internet
Intelligent Energy Network as Overlay on Both
15
Conventional Electric Grid
Generation
Transmission
Distribution
Load
Intelligent Energy Network
Load IPS
Source IPS
energy subnet
Intelligent Power Switch
Conventional Internet
Aware Co-operative Grid
16
• Monitor, Model, Mitigate• Deep instrumentation• Waste elimination• Efficient Operation
• Shifting, Scheduling, Adaptation
• Forecasting• Tracking• Market
• Availability• Pricing• Planning
LoCal Energy Nets in Action
17
IPScomm
power
now
Load profile
w$
now
Price profile
w
now
Actual load
w
Data centerIPS
Bldg Energy
Network
IPS
IPS
IPSInternet
Grid
IPS
IPS
Power proportional kernel
Power proportional service manager
Quality-Adaptive Service
M/R Energy
Net
IPS
IPS
IPS
AHU
Chill
CT
Questions…
• Where does the energy go?– how much is wasted? => do nothing well– how can the rest be optimized?
• How much demand slack is there?– Can it be exercised through shifting?– Energy storage? Electrical Storage?
• What limits renewable penetration?– vs storage, scheduling, cooperation
• What are the protocols involved?• System and network design• …
18
19
Intelligent Power Switch
(IPS)
Energy Network
PowerComm Interface
EnergyStorage
PowerGeneration
Host Load
Intelligent Power Switch
(IPS)EnergyStorage
Intelligent Power Switch
(IPS)EnergyStorageEnergyStorage
Intelligent Power Switch
(IPS)EnergyStorage
Intelligent Power Switch
(IPS)EnergyStorageEnergyStorage
Intelligent Power Switch
(IPS)EnergyStorage
Intelligent Power Switch
(IPS)EnergyStorageEnergyStorage
Intelligent Power Switch
(IPS)EnergyStorage
Intelligent Power Switch
(IPS)EnergyStorageEnergyStorage
Host LoadHost Load
energy flows
information flows
Intelligent Power Switch
• PowerComm Interface: Network + Power connector• Scale Down, Scale Out
Understanding Diverse Load
20
Energy Consumption Breakdown
21
Re-aggregation to Purpose
22
OS for Building, Datacenter, Grid, …
23
24
Datacenters
3-19-2004 25
Server Power Consumption
230
15
248
87
190
13
190
13
200
14
161
19
287
48
0
50
100
150
200
250
300
350
Wat
ts
Pow
erE
dge
1850
Del
l Pow
erE
dge
1950
Sun
Fire
V60
x
Sun
Fire
x21
00 -
Cyb
er S
witc
hing
Sun
Fire
X22
00
Com
paq
DL3
60
HP
Int
egrit
y rx
2600
Server Power Consumption
Active
Idle Soda Machine Room Power Consumption
26.5 30.6 31
18.118.9 19
44.5
50.9 50
9.5
10.117
10
31
0
20
40
60
80
100
120
140
160
180
est kW min est kW max kW meas
KW
290 Soda
288 Soda
530 Soda
420A Soda
340 Soda
287 Soda
• x 1/PDU efficiency + ACC
• If Pidle = 0 we’d save ~125 kw x 24 hours x 365 …
• … Do Nothing Well
26
“Doing Nothing Well”
• Existing systems sized for peak and designed for continuous activity– Reclaim the idle waste– Exploit huge gap in peak-to-average power consumption
• Continuous demand response– Challenge “always on” assumption– Realize potential of energy-proportionality
• From IT Equipment …– Better fine-grained idling, faster power
shutdown/restoration– Pervasive support in operating systems and applications
• … to the OS for the Building• … to the Grid
27
EnergyInterconnect
LocalGeneration
Local Load
IPS
LocalStorage
IPS
IPS
IPS
IPS
IPS
Scaling Energy Cooperation
• Hierarchical aggregates of loads and IPSs• Overlay on existing Energy Grid
Energy InterconnectCommunications Interconnect
Tools and Techniques
• Doing Nothing Well
• Scheduling
• Storage
Scheduling
Forecasting Supply
Shifting
Prioritizing
Storage
Monitoring
Modeling
Manage/Reduction
Consumption
Constructive PlanTools andTechniques
Supplies Transport Loads
Storage
Scheduling
Doing NothingWell
Generation Consumptioncooperation
Constructive PlanTools andTechniques
Supplies Transport Loads
Storage
Scheduling
Doing NothingWell
Generation Consumption
StaticPlannedProactiveDispatch
DynamicUnplannedReactiveNon-Dispatch
cooperation
Constructive PlanTools andTechniques
Supplies Transport Loads
Storage
Scheduling
Doing NothingWell
Generation Consumption
CentralizedAggregatedGlobal Control
DecentralizedDisaggregatedLocal Control
cooperation
Constructive PlanTools andTechniques
Supplies Transport Loads
Storage
Scheduling
Doing NothingWell
Generation Consumption
ComputeNodes
Machine RoomsDatacenters
Building
HVACLighting
Plug Loads
Buildings
AC/DCDistribution
LoCal-izedLocal Grid
LoCal-izedGeneration
cooperation
Google Earth Models
Google Earth Models
Berkeley
Available Testbeds
Richmond FieldStation
La Jolla
Wide-Area Network
Planned Testbeds
• Loads (with storage/supply/transport)– LoCalized Rack– LoCalized Machine Room/Datacenter– LoCalized Distributed Datacenters (with UCSD)– LoCalized Building– LoCalized Buildings/Campus/Local Grid
• Supplies– LoCalized Renewable Energy Source
• Beyond– Standalone Testbed (aka “Burning Man”)
36
38
Summary and Conclusions
• Monitor, Model, Manage: scalable infrastructure for integrated energy generation and storage
• IPS: points where information and energy flows come together
• Information overlay to the Grid, visualize usage patterns by facilities and individuals, do nothing well, enable markets
• Initial focus on buildings aware of energy usage and integration of renewable sources
