“Smart” Micro-Grids for Developing Regions Achintya Madduri, Javier Rosa, Matt Podolsky, Daniel Soto*, Eric Brewer, Seth Sanders (*) In Collaboration with: Matt Bassinger, John Sarik, and Vijay Modi of the Earth Institute, Columbia University Outline • How we fit into LoCal. • Need for electrifying rural developing regions. • Work done so far by Columbia University in Africa: usage statistics, methodology, etc. • DC microgrid description and features. • Future plans. How Does this fit into LoCal? • Do more with less. • Build a grid from ground up to achieve maximal energy efficiency. • Test-bed for LoCal grid-tied ideas: IPS, thermal slack, demand-response, etc. Need for Rural Electrification • Displace fossil fuels for night-time lighting. Source Cost/Hour Lumens Kerosene $0.04 8-82 10W LED@ $0.67/kWH $0.0067 400 • Low power required per household, but individual generation or fixed monthly cost not practical. Millennium Village Project • The Modi Group at Columbia University has implemented a community generation project: Shared Solar in 20 sites in Africa. • Goal: Was to build a framework to effectively collect tariff. – Existing overhead (administrative) costs are high – Non-payment rates for post-paid plans at 40% • Collected fine-grain data on usage and losses. Shared Solar Overview • Central generation station using a single large source (ex. 2.5 kW Solar Panel) • 220 V-AC transmission & distribution • Centralized metering, switching, and SMS communications. • 100 m radius for transmission. • 20 households with usage of ~200 W-hrs/day Uganda, 7% Electricity Penetration Shared Solar Installation & Gateway Problems Encountered • Off the shelf metering hardware inefficient, expensive, prone-to-failure, and no longer available. – Solution: Custom Hardware that we have developed and tested. • Bigger problem: System losses due to inverters and multiple AC-DC/DC-AC conversions while using mostly DC appliances. DC Micro-Grid • DC generation, distribution, and appliances. • Soft-transmission-voltage control scheme to ensure effective power sharing (ex. 280-320 VDC). • Real-time switching and metering for each household in a tree architecture. • Grid-controlled distributed storage. • Communications over power lines. AC vs. DC Comparison Factors AC micro-grid system with centralized storage DC micro-grid system with distributed storage Sizing Issues Inverter losses can dominate with mismatch to load. Lack of inverter/centralstorage makes it easy to incrementally add more generation. DC-AC/DC-DC conversion losses 20%4 : due to central battery bank + inverter. 15% daytime, 23% nighttime: due to 3 DC-DC conversion steps4,9 Component losses due to internal rectification and conversion > 25%7 0% for DC and >25% for AC loads8 End-to-end Efficiency < 60% 85-77% for DC loads, < 63% for AC loads Overview of Architecture Power Sharing Architecture Power Source ex. Solar Panel (1 KW) MPPT with DC/DC Converter DC line voltage that is variable depending on the load state. Ex. 280-320 VDC Housholds: (~100 W) DC-DC convertor with Storage (100-200 Whrs) … … Power Sharing Controls Distributed Storage • Storage should be sufficient for night-time usage. • Distributed storage minimizes losses of stored electricity. • Allows for incremental growth of high-cost components. Communications Over Power Lines • Low cost communications using a TDMA scheme over power lines. • Allows to close loop for power sharing. • Low bitrate communications: – Charge state of household – Credits – Usage Future Plans • Deploy AC-grid custom metering hardware in Haiti with Columbia University. • Finish framework for DC grid for both power and communications. • Simulate with sample DC-DC convertors and loads with usage patterns obtained from AC grid deployments. Thank You • Acknowledgements: – Modi Group, Columbia University – Alex McEachern, Power Standards Lab – Paul G. Allen Foundation • Questions? Summary • Target a market that is high need but has low usage. • Efficiency is essential to making a solution sustainable. • Re-make the grid to provide the desired utility without the existing constraints.