Maximizing Energy Capture in Hydrostatic Wind Energy Conversion Systems via Variance-Driven Adaptive Horizon Soft-Constrained Stochastic MPC

The hydrostatic transmission system provides structural decoupling in wind energy conversion systems; however, controlling it under turbulent conditions remains challenging. Utilization of conventional model predictive controllers is limited in two key aspects: hard pressure constraints lead to early torque cut-offs and thus energy loss; and fixed prediction horizons cause computational burdens as the variance of the wind forecast increases. We hypothesize that the combination of a soft-constrained stochastic tube formulation and a variance-driven adaptive horizon will yield the best energy extraction, while guaranteeing computational tractability. In this paper, an economic soft-constrained stochastic tube model predictive control with a sigmoid-based adaptive mechanism is proposed, which adjusts the prediction horizon dynamically according to the wind forecast variance in real time. When simulated using standard turbulent wind profiles, the proposed method obtains a mean total efficiency of 44.11% and a power gain of 8.21% against the rule-based baseline. Importantly, it also maintains a 0% constraint violation rate for the 190 bar pressure limit under high turbulence, while maintaining maximum quadratic programming solve time below 200 ms. This overcomes the trade-off between maximizing energy extraction and ensuring hydraulic safety, and demonstrates superiority over the conventional rigid architectures.