Revenue-based maintenance scheduling for wind turbines

Abstract

The ability to effectively schedule maintenance so that lost revenue is minimized is an important aspect of wind turbine management, and a recent interest at the company Rabbalshede Kraft AB. The aim of this thesis was to formulate short– term maintenance scheduling models as integer linear optimization problems and to implement these using open–source software. Two models were formulated for the cases of in-house maintenance teams and hired external workers, respectively. These models take a wind forecast and spot prices as input to form the revenue forecast used in the optimization. Both models were implemented and the former was thoroughly tested in hypothetical maintenance situations. The results showed that a significant reduction in lost revenue is possible by solving for the optimal schedule. Further, the wind speed forecast and turbine performance were analysed with respect to the expected error in lost revenue. Several methods of reducing run time of the optimization were studied. One of these, the so-called Dantzig–Wolfe decomposition, was formulated explicitly for both models and implemented for the in–house model. The result from a run time analysis showed that this method effectively reduces run time for a number of maintenance scheduling cases and allowed for practical solution of complex problems with free open–source software.