With the growth of demand for electrical energy, it is a common practice in the power industry to shorten the maintenance duration and often to postpone maintenance tasks. Maintenance or overhaul of generating units is one of the crucial factors in delivering reliable electrical energy. A large part of the overhaul work is to examine the condition of the mechanical and the electrical elements. This examination can highlight faults in some components requiring management of new activities. This paper presents a method using a dependency structure matrix for managing schedules within uncertain conditions of information dependency. The matrix is optimized using an evolutionary algorithm for scheduling overhaul and planned outages at hydro power plants. The dependency between tasks or the information flow between activities is formulated in a structural matrix with the objective to minimize the total completion time. The method is used to examine the staff scheduling of a twelve years overhaul at one of Landsvirkjun's hydro power units. Staff scheduling or rostering is a large scale constrained optimization problem. Using the Dependency Structure Matrix (DSM) and evolutionary methods (GA) to optimize the order of activities shows improvements in the overhaul operational plan.