Maintenance of power generators is essential for reliable and efficient electricity production. Because generators under maintenance are typically inactive, optimal planning of maintenance activities should consider the impact of maintenance outages on the system operation. However, finding a minimum cost maintenance schedule in hydropower systems is a challenging optimization problem due to the nonlinearity of the electricity production, the uncertainty of the water inflows and the intrinsic complexity of scheduling problems. We propose the first two-stage stochastic programming formulation for the hydropower generator maintenance scheduling problem, and we implement a parallelized Benders decomposition method with several acceleration techniques for its solution, considering a large number of scenarios. We apply statistical methods for selecting the best combination of acceleration techniques for the decomposition algorithm, and we compare the computational time of the parallelized decomposition against a mixed-integer linear programming solution approach using a testbed adapted from a real hydropower system in Canada.
Paru en mai 2018 , 29 pages