Energy-Aware Sensing in Data-Intensive Field Systems Using Supercapacitor Energy Buffer

摘要

Energy sustainability is important for field data sensing and processing in intelligent transportation, environmental monitoring, and context awareness. Rechargeable batteries in self-sustainable systems suffer from adverse environmental impact, low thermal stability, and fast aging. Advancements in supercapacitor energy density and low-power processors have reached an inflection point, where a data-intensive (e.g., operating on high-frame-rate visual data) field-deployed system can rely solely on supercapacitors for energy buffering. This paper demonstrates the first working prototype of such a system, consisting of eight 3000-Farad supercapacitors, a 70-mW controller/harvester board, and a Nexus tablet. We address the challenges of maintaining quality-of-service (QoS) on a limited energy buffer. We leverage the voltage-to-stored-energy relationship in capacitors to enable precise energy buffer modeling (no more than 3% error in time-to-depletion prediction). To achieve high precision, we find that it is necessary to account for the variation of effective capacitance, particularly lower capacitance at lower voltages nearing energy depletion. Modern mobile processors operate most efficiently at very high load (when most cycles are effectively utilized) or very low load (when fewer cores are active at a lower frequency). We propose delayed bursts-continuous low-power data capture and bursts of data processing at a higher CPU configuration-to improve the power proportionality and realize high QoS at varying energy budget. Our working prototype has been successfully deployed at a campus building rooftop where it analyzes nearby traffic patterns continuously.