The paper presents 0-1 integer programming models for the control of maintenance expenditure on thermal generating units. Choosing a sound and effective maintenance policy reduces the system down-time and thus increases the revenue to the utility. The objective is aimed at selecting that sot of proposals which will maximize the not present value of its total expected return. The alternative proposals, the forecasts of the net present value of the returns associated with the alternatives, the number of repairmen required for each alternative, and cash outflow required over the next 5 years are available for the analysis. The problem is discussed both under conditions of certainty and uncertainty. A new, simple and efficient tree search method is developed for the solution of the problems. A computer algorithm is developed and results of computation are presented for sample applications.