Posters

Abstract

Offshore wind developments continue to grow globally at a rapid pace. To facilitate this rapid growth costs must continue to be reduced. Reducing Operations and Maintenance (O&M) costs offers opportunities. With Offshore Wind Farms (OWFs) increasing in size and moving further offshore, it is becoming increasingly important to develop and deploy O&M strategies that can reduce O&M costs. Producing maintenance schedules and routes that can be used by the operators to minimise the costs and lost revenue of performing maintenance tasks could facilitate these cost reductions. Planning maintenance at OWFs can be very challenging. Teams need to consider the impacts of weather on accessibility and power generation, the electricity market in some cases, spare part and technician availability and vessel routes. Ideally a predictive maintenance approach should be implemented which entail accounting for expected failure events as well. For large windfarms it can be almost impossible for a human to find the optimal maintenance.   Therefore, this research presents a novel framework and solution methods for producing maintenance schedules and routes across a combined yearly and daily time horizon that can be used by maintenance teams at offshore wind farms. The framework facilitates the planning of maintenance in a way that minimises both the cost to perform maintenance and revenue losses from Wind Turbine (WT) downtime at an OWF. The problem of simultaneously solving both the yearly and daily timescales is divided into a long-term job allocation problem and a shorter-term Maintenance Routing and Scheduling Problem (MRSP). Starting with a set of known maintenance jobs that need to be performed for the year and forecasts of weather and electricity prices for each week, each job is allocated to a week within the year to minimise the expected revenue losses and costs for performing these maintenance actions. The problem is formulated in a Mixed Integer Linear Program (MILP) solved using commercial optimisation software. Once an initial LT plan has been produced an updated list of maintenance jobs, such as corrective maintenance, and new forecasts are acquired by the maintenance team. The long-term job allocation problem is then resolved based on the updated information as a multi-objective optimisation problem to minimise the disruption to the original long-term plan and the total cost of maintenance. The maintenance team then select their preferred schedule. Then daily schedules and maintenance vessel routes are provided by solving the MRSP. The MRSP is also formulated as a MILP and solved with commercial optimisation software. This process continues daily. Evaluated against a year of historic maintenance, the framework was able to achieve a cost reduction of 49% compared to the historic baseline. The savings were driven by significant reductions in the revenue losses from WT downtime thanks to strategic scheduling of maintenance. Further work will look at implementing the effects of uncertainty in weather forecasting, maintenance task duration and the electricity markets for planning maintenance. Implementing computational methods for planning maintenance will help to reduce wind farm revenue losses and aid the work of maintenance teams.