A Data‐driven Optimization Model for Generating Robust Templates of Scheduling Offshore Wind Farm Maintenance Tasks

Abstract

This paper studies the offshore wind farm maintenance scheduling from a new and more practical angle, generating templates of scheduling maintenance tasks robust to uncertain maintenance demand in scenarios with limited data. To realize such data-driven optimization of wind farm maintenance task scheduling, a distributionally robust optimization model with a ϕ-divergence-based ambiguity set is formulated to maximally meet the maintenance requirement while minimizing resource consumption under the worst-case distribution of maintenance demand. The proposed model can be easily transformed into its equivalent computationally tractable formulation and efficiently solved by off-the-shelf solvers. The ambiguity set is appropriately chosen by cross-validation techniques based on simulated data of maintenance demand. Computational results demonstrate that the proposed method can generate maintenance task schedules with a better out-of-sample performance compared to deterministic, stochastic programming, and robust optimization approaches given that only a small size of related data is available. Generated templates can support decision-making in scheduling maintenance tasks and can be particularly valuable to such needs for offshore wind farms having limited data collected.