SENSORLESS ENERGY HARVESTING STRATEGIES FOR DIELECTRIC ELASTOMER GENERATORS

This thesis investigates self-sensing and control strategies for Dielectric Elastomer Generators (DEGs) applied to energy harvesting from low-frequency mechanical sources. In these systems, the effectiveness of the harvesting cycle depends on the synchronization between the mechanical deformation of the elastomer and the electrical activation of the device. The work focuses on the estimation of the DEG capacitance without using external mechanical sensors. To this aim, the device is described through an equivalent electrical model with time-varying parameters, and the capacitance, together with the equivalent resistance, is estimated online from voltage and current measurements by means of a Recursive Least Squares (RLS) algorithm. The estimated capacitance is then used as the main control variable for the real-time detection of the extrema of the deformation cycle. Several detection strategies are analyzed, including direct sample-comparison methods, an RLS-based Ramp Detector, an approach inspired by Very Fast Simulated Annealing, and derivative-based methods. Finally, the thesis addresses the energy cost associated with the self-sensing process. For this purpose, energy-saving strategies based on time-windowed activation and adaptive adjustment of the measurement-signal amplitude are proposed. The developed methods are evaluated in simulation under sinusoidal and polychromatic mechanical profiles, showing the potential of self-sensing for sensorless synchronization and energetic optimization of DEG-based harvesting systems.