Time-domain model and optimization for single-axis kinetic energy harvesters driven by arbitrary non-harmonic excitation

Abstract

Abstract Energy harvesting is employed to extend the life of battery-powered devices, however, demanding applications such as wildlife tracking collars, the operating conditions impose size and weight constraints. They also only provide non-harmonic mechanical motion, which renders much of the existing literature inapplicable, which focuses on harvesting energy from harmonic mechanical sources. As a solution, we propose an energy harvesting architecture that consists of variable number of evenly-spaced magnets, forming a fixed assembly that is free to move through a series of evenly-spaced coils, and is supported by a magnetic spring. We present an electromechanical model for this architecture, and evolutionary optimization process that finds the model parameters which describe the time-domain behaviour observed in ground truth measurements. The resulting model can predict the time-domain behaviour of the energy harvester for any configuration of the proposed architecture and for any mechanical excitation. We also propose an optimization process that, using the electromechanical model, optimizes the energy harvester configuration to maximize the power delivered to a resistive load. The resulting optimized harvester design is specific to the particular kind of non-harmonic mechanical excitation to which it will be exposed. To demonstrate the effectiveness of our proposed model and optimization procedure, we constructed four energy harvesters, each with different configurations, and compared their measured behaviour with that predicted by the model, given an excitation that approximates footstep-like motion. We show that the model predictions were consistently within 25% of the RMS load voltage. We then synthesize an optimal energy harvester using the proposed optimization process. The resulting optimal design was constructed and tested using the same footstep-like excitation, and delivered an average power of 1.526 mW to a 30Ωload. This is a 2.8-fold improvement over an unoptimized reference design. We conclude that our proposed behavioural model and optimization process allows the determination of energy harvester designs that are optimized for a non-harmonic and specific input excitation.