Co-optimization of generator maintenance scheduling and operation using SDDP and partial Benders decomposition

Generators in a power system must undergo regular maintenance to ensure their longevity
and reliability. Maintenance requirements are often based on number of hours operated, number of
generator startups or simply elapsed time. With ever-increasing intermittent generation and price
volatility, it is imperative that maintenance scheduling is integrated into system planning to ensure
a reliable and cost-effective service. To this end, we present a multi-stage stochastic program that
co-optimizes maintenance planning and generator dispatching, including reservoir operation. We
use the SDDP algorithm and incorporate maintenance requirements as state variables, while the first
stage is dedicated to computing maintenance schedules using partial Benders decomposition.