TY - BOOK

T1 - Optimisation of scheduling and routing for offshore wind farm maintenance

AU - Kingsman, Toby

PY - 2020

Y1 - 2020

N2 - The growing increase in the size and scope of offshore wind farms motivates the need for industry to have access to mathematical tools that reduce costs by efficiently performing daily operations and maintenance activities. Key offshore activities require the transportation of technicians to and within offshore wind farms to complete corrective and preventive maintenance tasks to keep turbines operating efficiently.We provide a new deterministic mixed integer linear programming formulation for deciding the optimal vessel routes for transporting technicians around a wind farm and the scheduling of crew transfers, by minimising downtime, travel and technician costs. The model contains sufficient flexibility to account for multiple vessels, shifts and task profiles, whilst being able to prioritise and omit tasks in environments containing limited resources. Computational experiments are performed which quantify and confirm the impact of key instance characteristics such as technician availability, task profiles and weather conditions. We implement and evaluate the impact of a novel industry safety constraint. The complexity of larger instances motivates a second continuous time formulation, in which preventive maintenance again requires no minimum duration of work before it can provide benefit. We employ a specific decomposition structure to take advantage of variable preventive maintenance and utilise an adaptive large neighbourhood search procedure to solve instances. We evaluate several distinct acceptance criteria in conjunction with random and adaptive operator selection to determine the best option for our model. We produce a statistical model of offshore weather conditions to help quantify the likelihood of limited vessel accessibility to offshore wind farms. We model the joint distribution of key meteorological and oceanographic variables over time whilst accounting for seasonal trends using multivariate kernel density estimation. Our method generates alternative metocean realisations from historical data and reproduces the important long term persistence statistics of good and adverse offshore conditions.

AB - The growing increase in the size and scope of offshore wind farms motivates the need for industry to have access to mathematical tools that reduce costs by efficiently performing daily operations and maintenance activities. Key offshore activities require the transportation of technicians to and within offshore wind farms to complete corrective and preventive maintenance tasks to keep turbines operating efficiently.We provide a new deterministic mixed integer linear programming formulation for deciding the optimal vessel routes for transporting technicians around a wind farm and the scheduling of crew transfers, by minimising downtime, travel and technician costs. The model contains sufficient flexibility to account for multiple vessels, shifts and task profiles, whilst being able to prioritise and omit tasks in environments containing limited resources. Computational experiments are performed which quantify and confirm the impact of key instance characteristics such as technician availability, task profiles and weather conditions. We implement and evaluate the impact of a novel industry safety constraint. The complexity of larger instances motivates a second continuous time formulation, in which preventive maintenance again requires no minimum duration of work before it can provide benefit. We employ a specific decomposition structure to take advantage of variable preventive maintenance and utilise an adaptive large neighbourhood search procedure to solve instances. We evaluate several distinct acceptance criteria in conjunction with random and adaptive operator selection to determine the best option for our model. We produce a statistical model of offshore weather conditions to help quantify the likelihood of limited vessel accessibility to offshore wind farms. We model the joint distribution of key meteorological and oceanographic variables over time whilst accounting for seasonal trends using multivariate kernel density estimation. Our method generates alternative metocean realisations from historical data and reproduces the important long term persistence statistics of good and adverse offshore conditions.

U2 - 10.17635/lancaster/thesis/1139

DO - 10.17635/lancaster/thesis/1139

M3 - Doctoral Thesis

PB - Lancaster University

ER -