Transcript ICT and Power (Electricity) Prof. Rahul Tongia School of Computer Science CMU 17-899 Fall 2003

ICT and Power (Electricity)
Prof. Rahul Tongia
School of Computer Science
CMU
17-899 Fall 2003
Topics for Discussion



Electricity and Development
Power for ICT
ICT for Power
Rahul Tongia, CMU
2
Fundamentals


Electricity is a form of energy (kWh)
Does not exist in usable forms

Conversion usually requires prime movers (steam turbines,
water turbines, etc.)


Access to fuels (primary energy) is a key issue for developing
countries
Electricity is only about 125 years old

Widespread use is much more recent

US required special programs



Rural Electrification Administration (REA) [now Rural Utilities
Service]
TVA
Electricity from the grid can not be easily stored (AC)

Most electronics use DC
Rahul Tongia, CMU
3
What’s Special about LDCs?

Very low levels of Electrification



2 billion+ lack electricity
Bad quality, intermittent, and often expensive power if available
Lower Level of Economic Development

Large rural agricultural sector




Large quantities of crop residues: primary energy source
Special needs for agricultural services (e.g., pumping water ~ 1/3 of
India’s electricity)
Heavily subsidized in many countries
Industrial-Political Organization

State-centered economies


State-owned enterprises (SOEs) handle not just power but much of
the economy
Weak formal institutions
 E.g.,
Rahul Tongia,
CMU
regulatory institutions, courts, corporate governance
4
Energy-Economy Correlation
10000
GDP (Billion $)
North America
Japan
Germany
Developing
W. Europe
1000
Australia
FSU/E. Europe
OECD Asia/Pacific
100
US
Brazil
China
New Zealand
Russia
Bangladesh
India
South Korea
Pakistan
Mexico
10
Turkmenistan
1996
1
1
Calculated from EIA Data
Rahul Tongia, CMU
10
100
1000
10000
Primary Energy (Trillion BTU)
100000
5
(Lack of) Access to Electricity
People without Electricity Access (millions)
900
South Asia (India)
800
700
600
Sub-Saharan Africa
500
400
300
East Asia (China)
200
100
0
1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030
Source: WEO 2002
Rahul Tongia, CMU
6
Investments in
LDC Power Sector
Source: World Bank (2003)
Rahul Tongia, CMU
7
Where Does Electricity Go?

US


~ 1/3 residential, 1/3 industrial, 1/3 commercial
Developing Countries

Varies significantly by country


Typically higher shares for non-residential (function of large,
centralized design)
Grid penetration to rural areas is very low

Kenya used to have more homes served by Decentralized
Generation (DG) than the grid (mainly solar)
In reality, a fair amount is lost along the way, or stolen!
Rahul Tongia, CMU
8
Electricity in LDCs
Rahul Tongia, CMU
Source: World Bank (2003)
9
How Much Electricity Does ICT
Use?

Numbers as high as 13% of US
electricity were claimed



End users, servers, networking, etc.
Later debunked
ICT – Energy (Power) linkages


Greater Service Economy, even in
developing countries
But, increased globalization
Rahul Tongia, CMU
10
What Consumes Power (ICT
Applications)?

Components of an ICT solution


Computing
Display





CRT
LCD
Storage
Uplinking
80 W normal
15-25 W normal
10 W suspend
5-10 W suspend
variable
12 W Wifi
40 W VSAT
Role of advanced technologies

Chips (processor is largest component)



Pentium 4 uses 50+ watts!
LCD screens, OLEDs, etc.
Wireless


Cognitive Radios – reduce power to lowest required level
But, emitted power is << power drawn from supply


100 mW is legal limit for WiFi
Laptops – much less power but less robust (?)
Rahul Tongia, CMU
11
Details of Desktop Power
AGP video card - 20-30W
PCI video card - 20W
AMD Athlon 900MHz-1.1GHz - 50W
AMD Athlon 1.2MHz-1.4GHz - 55-65W
Intel Pentium III 800MHz-1.26GHz - 30W
Intel Pentium 4 1.4GHz-1.7GHz - 65W
Intel Pentium 4 1.8GHz-2.0GHz - 75W
Intel Celeron 700MHz-900MHz - 25W
Intel Celeron 1.0GHz-1.1GHz - 35W
ATX Motherboard - 30W-40W
128MB RAM - 10W
256MB RAM - 20W
12X or higher IDE CD-RW Drive - 25W
32X or higher IDE CD-ROM Drive - 20W
10x or higher IDE DVD-ROM Drive - 20W
Rahul Tongia, CMU
SCSI CD-RW Drive - 17W
SCSI CD-ROM Drive - 12W
5400RPM IDE Hard Drive - 10W
7200RPM IDE Hard Drive - 13W
7200RPM SCSI Hard Drive - 24W
10000RPM SCSI Hard Drive - 30W
Floppy Drive - 5W
Network Card - 4W
Modem - 5W
Sound Card - 5W
SCSI Controller Card - 20W
Firewire/USB Controller Card - 10W
Case Fan - 3W
Source: FLECOM
CPU Fan - 3W
12
Standalone (DG) Power

What are the options if If AC power is unavailable?


Backup or primary supply?
Non-Conventional Sources of Power

Issues of Scale


For ICT or more (single point or village level)?
Local availability

Solar


Wind


Conversion options limited, typically require tens of kW size
Microhydel


Windspeeds vary by location; highest efficiency for megawatt class turbines
Biomass


Only 3-5 hours equivalent per day (1 kW INPUT/m2 of panel; ~10% efficiency)
Location sensitive, and typically 10s of kW
Diesel

Expensive to run, typically AC output
Rahul Tongia, CMU
13
Designing a DG system

Battery Life examples

Alkaline (from Duracell)
D
C
AA
AAA
NOMINAL VOLTAGE (volts)
1.5
1.5
1.5
1.5
RATED CAPACITY (ampere-hours)
15
7.8
2.85
1.15
Gets very expensive, quickly, even if rechargeable

Lead-acid batteries give much more power and are standardized


Matching supply to demand


AC grid – “infinitely” flexible
Power storage is key



Limits on dischargeability - ~20 kWh total charge
Else peak capacities must be matched
Intermittency issues for many DG systems
Theft is a major concern for DG design (!)
Rahul Tongia, CMU
14
Designing a DG system (cont.)

Solar Systems

Components






PV modules (in series, in panel form)
Power Conditioning Equipment (economies of scale)
Housing (with or without directionalizing)/mounting
Batteries – most expensive operating costs*
Inverter – if AC is required
Costs


Capex at small scale is ~5/peak watt
Gives an operating cost around 20-30 cents/kWh
* cell phone example – Obsolescence of equipment vs.
battery
Rahul Tongia, CMU
15
Designing a DG system (cont.)
Hours per day operational
Days back-up required
Power needs
12
3 (1 current day plus 2 days of no sun)
50
20
15
15
100
Notebook PC
Communication
Lighting
Other*
Watts
AVERAGE
3,600 W-hrs
Equivalent peak sunlight
5 hrs
required to charge up per day
per day
System size calculation
720 peak watts
5 $/peak watt
3,600 $
Capex
Sizing - 1 meter panel
1,000
10%
100
7.2
Thus need
Rahul Tongia, CMU
W
input
efficiency (net)
W
electricity out
sq. m
panel
(peak)
(peak)
16
ICT for Electricity Systems

Two main issues

Supply << Demand


Ability to pay is limited


Requires investments of billions
Often, power companies are loss-making; some of that is
inefficiency
Where can ICT contribute?

Components of power sector vertical




Rahul Tongia, CMU
Generation
Transmission
Distribution
Consumption
17
Conventional Wisdom

One can not do real-time power flow
management (transactions and billing) for
transmission level flows


Today, pools operate based on historical or
aggregated information
One can not measure demand (usage) from
all consumers in real-time with high
granularity
What has changed to make these outdated –
the growth of IT technology
Rahul Tongia, CMU
18
Focus here on
Distribution/Consumption


IT is already extensively used in
generation/transmission in developed
countries
Other Synergies


Stringing Optical Fibers along power lines
Smart Cards (pre-payment)


Found extensive use in S. Africa in Black Townships (12
years experience)
Can link to other utilities or consumer services (pre-paid
cell-phone cards are very popular)
Rahul Tongia, CMU
19
Using IT to Enable
Sustainability

Sustainability has many components

Resource utilization


Economic viability



Efficiency and loss reduction are sine-qui-non
Theft reduction
Management
IT can improve power sector distribution,
consumption (utilization), and quality of service

Requires a change in mindset, and the willingness of
utilities to innovate
Rahul Tongia, CMU
20
Case study on IT for power sector
improvement in India

India today has the world’s largest number of
persons lacking electricity


 400 million (equivalent to Africa’s unserved!)
Reforms began in 1991



Vertically integrated government department
monopolies are being broken
Initial focus was on generation
New realization that distribution is the key to
India’s power sector viability

Newer entities should be run as businesses
Many parallels to other developing countries
Rahul Tongia, CMU
21
India’s Power Sector Overview

5th largest in the world – 107,000+ MW of capacity

But, per capita consumption is very low



350 kWh, vs. world average over 2,000 kWh
40% of households (60% of rural HH) lack electricity
In very dire straits

Supply << Demand



Blackouts are common, with shortfall estimated between 10-15%
Most utilities are heavily loss-making, with an average rate of
return of negative 30% or worse (on asset base)
High levels of losses = 25+%


Technical losses – poor design and operation
Commercial losses (aka theft) often over 10%
Rahul Tongia, CMU
22
Reasons for the problems

Agricultural sector


Consumes 1/3 of the power, provides <5% of revenues
Pumpsets are overwhelmingly unmetered – just pay flat rate
based on pump size


Adds to uncertainty in technical losses vs. commercial losses and
usage
Utilities lack load duration curves to optimize
generation and utilize Demand Side Management

All generation is assumed to be baseload, and priced
accordingly



Leads to poor energy supply portfolio
Doesn’t send correct signals to consumers, either
Utilities end up using just average costing numbers, not
recognizing the marginal costs
Rahul Tongia, CMU
23
Idea – use IT for power sector
management

Posit – If new meters are to be installed, why
not “smart” digital meters, which are also
controllable, and communications-enabled?



Incremental costs would be low
Instead of just quantity of power, can also
improve quality of power
Analysis presented is based on collaborative
work with a major utility in India (name
withheld for confidentiality reasons)
Rahul Tongia, CMU
24
Quality of Power

India is focusing on quantity of power only

Current “shortfall” numbers are contrived




Quality norms are often missed




Based only on loadshedding with minor correction for frequency
Do no factor in peak clipping fully
Do not account for lack of access (e.g., over 60% of rural homes lack
connections)
Voltage – often deviates by 25+%
Frequency – often deviates by 5% (!)
Even farmers pay a lot for their bad quality power (around 1
cent/kWh implicit, even higher in some regions)
Use of voltage stabilizing equipment


Additional capital costs (in the multiple percent range)
Efficiency losses (2-30% lost!)
Rahul Tongia, CMU
25
Power Quality:
ITI CBEMA
Curve
Rahul Tongia, CMU
26
Why the Focus on
Distribution?


It’s where the consumer (and hence, revenue) is
High losses today



Technical losses, 10+ % in rural areas
DSM and efficiency measures possible
Use of standards required



Use a combination of technology, industrial partnership, and
regulations
Learn from experiences elsewhere
Bulk of India's consumption is for just several classes of devices




Rahul Tongia, CMU
Pumpsets
Refrigerators
Synchronous motors
Heating (?)
27
US Refrigerator Efficiency
Standards
Similar standards can be established for
“smart appliances”
Source: www.standardsasap.org
Rahul Tongia, CMU
28
Future of Appliances and Home
Energy Automation Networks

Incremental cost of putting networking and
processors into appliances approaching a few dollars


Could allow time of use and full control (utility
benefit/public good/user convenience)
Link to a smart distribution system

Micro-monitor and Micro-manage every kWh over the network


5% peak load management could lead to a 20% cost reduction


E.g., refrigerators – don’t operate or defrost during peaks (5% of
Indian electricity usage)
Feasible, as most peak loads are consumer-interfaced
 Bimodal peaks in India, residential driven
Italy is already implementing such a system (ENEL)
Rahul Tongia, CMU
29
Objectives and design goals
for a new IT-enabled

Implement a basic infrastructure to…




Micro-measure every unit of power across the network
Allow real-time information and operating control
Devise mechanisms to control the misuse and theft of power
through soft control
Which would…





Reduce losses
Improve power quality
Allow load management
Allow system-level optimization for reduced costs
Increase consumer utility, satisfaction, and willingness to pay
Rahul Tongia, CMU
30
Additional Benefits

A system which will offer




Outage detection and isolation
Remote customer connect & disconnect
Theft and tamper detection
Real time flows




Suitability for prepayment schemes
Load profiling and forecasting
Possible advanced communications and services



To allow real time pricing
Information and Internet access
Appliance monitoring and control
Managing such “extra” power (from theft) is enough to give
subsistence connectivity to the poor

Requires ICT to determine and manage the margin effectively

Telecom is special – very short-run low marginal cost; in electricity it is
much more difficult
Rahul Tongia, CMU
31
Data Center
Network Schematic
~ 20 km
Couple
Coupler
r
Uplink
Sub-Transmission
and Transmission
(> 11 kV)
Last Few Hundred Meters
Coupler
House
LV Concentrator
Coupler
Substation
Distribution
Transformer
(pole or ground)
Secondary
Distribution
Voltage
House
Users
Smart Meter
(Can be off-site outside user
Control;
Is partly a modem)
Distribution
(~11 kV)
Medium Voltage
Access
(440, 220, or 110 V)
Low Voltage
Components of the solution

One segmentation – locational

At consumer

Meter/Gateway


In home network



Meter could be pole-side if required
Needed connect to enabled devices (appliances)
Eventually, homes would also have Decentralized Generation
available (?fuel cells, flywheel storage, etc.)
Access (low voltage distribution)

From gateway to a concentrator, on user side of
distribution transformers – Using PowerLine Carrier (PLC)
Rahul Tongia, CMU
33
Solution Components (Cont.)

Concentrator upwards


Concentrator – Each Distribution Transformer (aka Low
Voltage Transformer) feeds on the order of 100-200
homes in India (as in Europe). In contrast, US
Distribution Transformers feed 5-10 users.
Communications medium

Over Medium Voltage PLC to the Sub-station
or


Wireless
 Limited Coverage in Developing Countries
Substation upwards (uplinking)

Usually based on leased lines or optical fiber
Rahul Tongia, CMU
34
Technologies for various
segments

In-Home Network

Appliances



Emerging Standards are talked about by appliance
companies (Maytag, Samsung, GE, Ariston etc.)
Using Simple Control Protocol (or other appropriate “thin”
protocols)
Meters



Solid-State meters exist, but not yet the norm in
developing countries
Most have communications capabilities for
external ports
Lowest cost solution (if feasible) – PLC – target 5$
incremental cost
Rahul Tongia, CMU
35
Technologies for various
segments (cont.)

Access



MV



Low Voltage PLC is available today
Being explored for Internet access, in fact
(Megabits per second)
Crossing through transformers remains a technical
challenge
Going long distances an issue
Uplinking

Availability of optical fiber or leased lines can be
met through planning
Rahul Tongia, CMU
36
Technologies vs. Capabilities
Capabilities
Accuracy
Theft
Detection
Communications
Control
Electromechanical
Meter
low (has
threshold
issues for
low usage)
poor
expensive add-on
nil
Digital (solid
state)
high
Node only
external
Limited
Historical
usage reads
only
Arbitrarily
high
High
(network
level)
Built-in (on-chip)*
Full
(connect/disconnect);
Extending
signaling to
appliances
Real-Time
control; DSM
Next Gen.
Meter
(proposed)
Rahul Tongia, CMU
*Can do much more than
Automated Meter
Reading (AMR)
37
Design Model and Business
Case

Only target specific users



All agricultural (almost one-third of the load)
All Industrial and larger commercial users
Only the larger-size domestic users


Estimated 2/3 of homes only use <50 kWh per month
Include every network node that needs monitoring
and/or control




Substations
Transformers
Capacitor banks
Relays
etc.
Rahul Tongia, CMU
38
Design Model and Business
Case (cont.)

Investment in long run only a few thousand rupees
per targeted user (Target <75$ capex)


When amortized, implies requirement of improvements in
system of only a few percent!
Savings will come from






Lower losses/theft
Increased sales possible
Lower operational costs
Load management
Better consumer experience (and hence, possibility for higher
tariffs)
Future interaction with smart appliance and smart home
networks

Rahul Tongia, CMU
Possibly new services
39
Economics of case system

Estimated
System (Ruralcentric)


62 Consumers
(all classes)
per Distr.
Transformer
98
Distribution
Transformers
per SubStation
Domestic (applicable)
Commercial
Agricultural
High-Tension
Distribution Transformers
Substations
Needed Savings
$
Number of Nodes Equipment cost ($)
200,000
75
383,000
75
673,000
75
70,306
714
500
5,000
35,153,000
3,570,000
132,923,000
Other IT and infrastructure (capitalized)
10,000,000
142,923,000
15% <-annualized rate incl. Amortization
$ 21,438,450 annually
11,625,000,000 kWh sold annually
0.06 Electricity Rate ($/kWh)
697,500,000 Annual Costs
3.1% <- Need improvements worth
Rahul Tongia, CMU
Cost ($)
15,000,000
28,725,000
50,475,000
<- Average only;
Excludes peak
savings potential
40
Economics (cont.)


6-7 year payback on investment
(conservative) possible with just 3%
improvement in system
Savings will come from




Theft Reduction
Time-of-Day and DSM measures (peak reduction)
System Quality, reliability, and uptime
Higher Collection
Rahul Tongia, CMU
41
Challenges

Protocols



PLC






How to couple around transformers or other obstacles
How to go long runs with low errors (and high enough bandwidth)
– Shannon’s theorem provides a limit
Noisy line conditions in many developing countries
Appliances


Use of thin protocols to reduce capex for embedded systems
Security – PLC can be a shared medium
Need for standards to bring down costs and ensure inter-operability
Design – Should the PLC signals pass through the
meter/gateway directly to appliances?
How active or passive should consumer behavior modification
be?
Costs (as always)
Rahul Tongia, CMU
42
Challenges – Implementation
and Management


Utilities are typically risk-averse
They face increased regulatory uncertainty



Without some portions of a market, how do they benefit?
Will they (should they) pass all pricing information on to the
consumer?
Developing country management issues


Utilities were typically State Owned Enterprises (SOEs)
Utilities were run with social engineering goals

Increased automation, control, and sophistication (and theft
detection) poses risks to the large cadre of current employees
Rahul Tongia, CMU
43
A New World for Power
Systems




Includes “smarts” for significant
improvements in efficiency
New services can be enabled once the
appropriate infrastructure is in place
Segmentation of development allows
independent, modular innovation, e.g., home
automation and appliances
Developing countries (esp. Asia) can lead the
way through leap-frogging
Rahul Tongia, CMU
44