Generator maintenance scheduling (GMS) is important for economic and secure operation of hydrothermal systems. While incorporating unit start-up and shutdown constraints can significantly enhance the practicality of the schedule

it introduces substantial computational challenges

particularly as the number of units or planning periods increases. In this paper

the GMS considering unit ON/OFF states for hydrothermal systems is formulated as a mixed-integer linear programming (MILP) model

in which network constraints and cascade coupling are also taken into account. We employ a mathematical programming-based heuristic named alternating criteria search (ACS) to solve the large MILP model. To improve solution efficiency

we adopt a check-and-add method based on ACS subproblems to deal with a large number of network constraints and extend the basic ACS to aspects of the mathematical forms of subproblems

initial solution generation scheme

and variable fixing scheme according to features of the proposed GMS. Finally

simulations of a modified IEEE 118-bus and 54-unit hydrothermal power system and a real 1348-bus and 155-unit hydrothermal power system are conducted

which show the application and extension of ACS are effective to speed up the process of finding a near-optimal solution for large-scale systems.