Energy Harvesting Circuits for Battery-Free IoT Powered by Plant Microbial Fuel Cells

Battery-free Internet of Things (IoT) systems can reduce maintenance costs and environmental impact by operating directly from harvested energy. However, replacing batteries with weak and variable sources introduces a fundamental energy-management challenge: harvested power is often in the microwatt range, whereas sensing, computation, communication, and monitoring tasks require short bursts of higher energy. This mismatch is especially critical for Plant Microbial Fuel Cells (PMFCs), which are sustainable bioelectrochemical energy sources but provide low voltage, low power, and slow recovery dynamics. This dissertation investigates energy-harvesting circuits and storage architectures for battery-free embedded systems powered by PMFCs. First, it presents a controlled intermittent power-extraction interface that imposes explicit extraction and recovery timing while remaining compatible with energy-harvesting converters. Experiments on real PMFCs show that some intermittent extraction periods are associated with higher harvested power: in preliminary tests, a 10 s period reached 83 μW, 30% higher than 100 s and 63% higher than 1000 s. In a 9-day experiment, two intermittently loaded cells increased from 47.2 to 122.4 μW and from 19.4 to 67.6 μW, compared with 32.8 to 43.0 μW for the continuously loaded reference. Electrochemical impedance spectroscopy further shows that loading history affects the anode and cathode responses. Second, the dissertation presents CapDYN, an autonomous dynamic-capacitance architecture for battery-free intermittent systems. CapDYN adapts the equivalent storage capacitance during charge and discharge, exposes intermediate energy states to the runtime, and matches them to task requirements. The measured hardware reduces leftover energy from 23.67% in a fixed-capacitor baseline to 0.16%–5.57%, while keeping the control infrastructure below 10 μW. In the LTspice-based application-level evaluation, the energy-aware runtime avoids failed executions, improves effective efficiency by up to 34.10%, and advances aggregate first task completions by up to 45.00%. Overall, this thesis shows that efficient battery-free operation requires co-design across the source interface, storage architecture, and runtime, increasing the fraction of scarce sustainable energy that becomes completed embedded tasks.