From Prediction to Prescription: Integrating Risk Estimation with Optimization for Offshore Wind Farm Maintenance

Abstract

Offshore wind farms are crucial in the transition to renewable energy. But their operation and maintenance (O&M) tasks remain costly and difficult. Maintenance planning is challenging due to harsh weather, limited accessibility, vessel requirements, and crew shortages. Predictive maintenance methods can identify possible equipment failures, but many of the studies that are now available focus mainly on prediction and offer little assistance for maintenance decision-making.

This thesis presents a predictive-prescriptive maintenance framework for offshore wind farms that integrates predictive maintenance alerts with maintenance scheduling. The system prioritizes maintenance tasks and evaluates the urgency of maintenance using Remaining Useful Life (RUL) and failure probability data. The tasks are then assigned with respect to real-life constraints like weather accessibility, crew availability, vessel capacity, and maintenance time.

Two scheduling strategies, a Mixed-Integer Linear Programming (MILP) optimizer and a Greedy heuristic scheduler, were tested. Predictive maintenance data was used to test the framework under various workforce availability conditions. The results show that the vessel capacity becomes increasingly important as workforce resources increase, but manpower availability is the key constraint in maintenance operations. By achieving higher maintenance completion rates, lower operational costs, and better resource use, the MILP optimizer consistently outperformed the Greedy scheduler.

The result shows that predictive maintenance is not enough to improve offshore maintenance performance. Effective scheduling is also necessary to convert predictive data into useful maintenance operations. Therefore, this thesis concludes that smart scheduling is as important as accurate prediction for improving the efficiency and cost-effectiveness of offshore wind farm maintenance.