Cost optimization of maintenance scheduling for wind turbines with aging components

Abstract: A major part of the wind turbine operation cost is resulted from the maintenance of its components. This thesis deals with the theory, algorithms, and applications concerning minimization of the maintenance cost of wind power turbines, using mathematical modelling to find the optimal schedules of preventive maintenance activities for multi-component systems.     The main contributions of this thesis are covered by the four papers appended. The unifying goal of these papers is to produce new optimization models resulting in effective and fast algorithms for preventive maintenance time schedules. The features of the multi-component systems addressed in our project are: aging components, long-term, and short-term planning, planning for a wind power farm, end of the lifetime of the wind farm, maintenance contracts, and condition monitoring data.     For the long-term maintenance planning problem, this thesis contains an optimization framework that recognizes different phases of the wind turbine lifetime. For short-term planning problem, this thesis contains two modeling frameworks, which both focus on the planning of the next preventive maintenance activities. Our virtual experiments show that the developed optimization models adopt realistic assumptions and can be accurately solved in seconds. One of these two frameworks is further extended so that available condition monitoring data can be incorporated for regular updates of the components' hazard functions. In collaboration with the Swedish Wind Power Technology Center at Chalmers and its member companies, we test this method with real-world wind farm data. Our case studies demonstrate that this framework may result in remarkable savings due to the smart scheduling of preventive maintenance activities by monitoring the ages of the components as well as operation data of the wind turbines.     We believe that in the future, the proposed optimization model for short-term planning based on the component age and condition monitoring data can be used as a key module in a maintenance scheduling app.