Optimization of routing and scheduling for performing maintenance at offshore wind farms
Abstract
As the demand for energy in general and renewable energy in particular grows, so does the wind energy industry. As the more easily accessible sites available for wind turbines are exhausted, new sites must be found elsewhere. Therefore, offshore wind farms have to be built further from shore, leaving them more exposed to rough weather and less accessible for maintenance. Traveling times and the cost of maintenance are both increased. This increases the potential for savings and the importance of proper routing and scheduling of maintenance vessels.In this thesis a model for such routing and scheduling is developed. It allows for leaving maintenance technicians at wind turbines to pick them up later, which gives it a structure similar to a pickup-and-delivery problem with time windows. Maintenance operations are divided into two classes, corrective and preventive maintenance. The model includes time-dependent downtime costs, and the preventive operations create some extra challenges when it comes to modeling since costs are dependent on both start and end time of the operation.Two alternative approaches for solving the wind turbine maintenance routing problem are explored, an arc-flow and a path-flow formulation. Both models are tested on instances of varying number of vessels and operations. The arc-flow model is solved with commercial software using branch-and-bound. The path-flow model is solved using a labeling algorithm to enumerate feasible paths and using dominance to eliminate less promising paths, before selecting the optimal paths by solving a set partitioning problem. The path-flow model is not exact for instances involving preventive operations, but produces objective function values at or very close to optimum.Both methods are able to solve small and medium-sized instances in satisfactory time, and the path-flow heuristic is substantially quicker and can handle a few more maintenance operations than the arc-flow model.