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

Rahul Tongia, CMU2 Topics for Discussion Electricity and Development Power for ICT ICT for Power

Rahul Tongia, CMU3 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, CMU4 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., regulatory institutions, courts, corporate governance

Rahul Tongia, CMU5 Energy-Economy Correlation 1996 Calculated from EIA Data

Rahul Tongia, CMU6 Source: WEO 2002 (Lack of) Access to Electricity East Asia (China) Sub-Saharan Africa South Asia (India)

Rahul Tongia, CMU7 Investments in LDC Power Sector Source: World Bank (2003)

Rahul Tongia, CMU8 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, CMU9 Electricity in LDCs Source: World Bank (2003)

Rahul Tongia, CMU10 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, CMU11 What Consumes Power (ICT Applications)? Components of an ICT solution Computing Display CRT 80 W normal10 W suspend LCD15-25 W normal5-10 W suspend Storagevariable Uplinking12 W Wifi40 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, CMU12 Details of Desktop Power AGP video card W PCI video card - 20W AMD Athlon 900MHz-1.1GHz - 50W AMD Athlon 1.2MHz-1.4GHz W 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 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 CPU Fan - 3W Source: FLECOM

Rahul Tongia, CMU13 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 Only 3-5 hours equivalent per day (1 kW INPUT/m 2 of panel; ~10% efficiency) Wind Windspeeds vary by location; highest efficiency for megawatt class turbines Biomass Conversion options limited, typically require tens of kW size Microhydel Location sensitive, and typically 10s of kW Diesel Expensive to run, typically AC output

Rahul Tongia, CMU14 Designing a DG system Battery Life examples Alkaline (from Duracell) NOMINAL VOLTAGE (volts)RATED CAPACITY (ampere-hours) D1.515 C AA AAA Gets very expensive, quickly, even if rechargeable Lead-acid batteries give much more power and are standardized Limits on dischargeability - ~20 kWh total charge Matching supply to demand AC grid – “ infinitely ” flexible Power storage is key Else peak capacities must be matched Intermittency issues for many DG systems Theft is a major concern for DG design (!)

Rahul Tongia, CMU15 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 cents/kWh * cell phone example – Obsolescence of equipment vs. battery

Rahul Tongia, CMU16 Designing a DG system (cont.)

Rahul Tongia, CMU17 ICT for Electricity Systems Two main issues Supply << Demand Requires investments of billions Ability to pay is limited Often, power companies are loss-making; some of that is inefficiency Where can ICT contribute? Components of power sector vertical Generation Transmission Distribution Consumption

Rahul Tongia, CMU18 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, CMU19 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, CMU20 Using IT to Enable Sustainability Sustainability has many components Resource utilization Efficiency and loss reduction are sine-qui-non Economic viability 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, CMU21 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, CMU22 India ’ s Power Sector Overview 5 th 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, CMU23 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, CMU24 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, CMU25 Quality of Power India is focusing on quantity of power only Current “ shortfall ” numbers are contrived 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) Quality norms are often missed 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, CMU26 Power Quality: ITI CBEMA Curve

Rahul Tongia, CMU27 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 Pumpsets Refrigerators Synchronous motors Heating (?)

Rahul Tongia, CMU28 US Refrigerator Efficiency Standards Similar standards can be established for “smart appliances” Source:

Rahul Tongia, CMU29 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 E.g., refrigerators – don ’ t operate or defrost during peaks (5% of Indian electricity usage) 5% peak load management could lead to a 20% cost reduction Feasible, as most peak loads are consumer-interfaced Bimodal peaks in India, residential driven Italy is already implementing such a system (ENEL)

Rahul Tongia, CMU30 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, CMU31 Additional Benefits A system which will offer Outage detection and isolation Remote customer connect & disconnect Theft and tamper detection Real time flows To allow real time pricing Suitability for prepayment schemes Load profiling and forecasting Possible advanced communications and services 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

Access (440, 220, or 110 V) Low Voltage Smart Meter (Can be off-site outside user Control; Is partly a modem) Secondary Distribution Voltage House Users Distribution (~11 kV) Medium Voltage Couple r Coupler ~ 20 km Last Few Hundred Meters Substation Data Center Distribution Transformer (pole or ground) Coupler Sub-Transmission and Transmission (> 11 kV) LV Concentrator Network Schematic Uplink

Rahul Tongia, CMU33 Components of the solution One segmentation – locational At consumer Meter/Gateway Meter could be pole-side if required In home network 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, CMU34 Solution Components (Cont.) Concentrator upwards Concentrator – Each Distribution Transformer (aka Low Voltage Transformer) feeds on the order of 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, CMU35 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, CMU36 Technologies for various segments (cont.) Access Low Voltage PLC is available today Being explored for Internet access, in fact (Megabits per second) MV 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, CMU37 Technologies vs. Capabilities AccuracyTheft Detection CommunicationsControlCapabilities Electro- mechanical Meter low (has threshold issues for low usage) poorexpensive add-onnil Digital (solid state) highNode onlyexternalLimitedHistorical usage reads only Next Gen. Meter (proposed) Arbitrarily high High (network level) Built-in (on-chip)* *Can do much more than Automated Meter Reading (AMR) Full (connect/dis- connect); Extending signaling to appliances Real-Time control; DSM

Rahul Tongia, CMU38 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, CMU39 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 Possibly new services

Rahul Tongia, CMU40 Economics of case system Estimated System (Rural- centric) 62 Consumers (all classes) per Distr. Transformer 98 Distribution Transformers per Sub- Station

Rahul Tongia, CMU41 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, CMU42 Challenges Protocols Use of thin protocols to reduce capex for embedded systems Security – PLC can be a shared medium 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 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, CMU43 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, CMU44 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