A Fused Gaussian Process Modeling and Model Predictive Control Framework for Real-Time Path Adaptation of an Airborne Wind Energy System

This article presents a computationally tractable adaptive control strategy suitable for mobile systems operating in a stochastically and spatiotemporally varying environment by fusing Gaussian process modeling and receding horizon control. This strategy ideally manages the tradeoff between exploration (maintaining an accurate estimate of the stochastic resource) and exploitation (maximizing a performance index, which generally consists of harvesting the resource) subject to partial observability (stochastic resource only measurable at the system’s location) and mobility constraints, which are characteristic of dynamic systems. The case study in this article focuses on a crosswind airborne wind energy (AWE) system where the wind turbine tower is replaced by tethers and a lifting body, allowing the system to adjust its altitude, with the goal of operating at the altitude that maximizes net energy production in a wind environment that is changing in altitude and time. Real wind speed versus altitude data has been used to validate the strategy and results are presented for a variety of control strategies applied to a rigid wing-based AWE system.