This work presents a practical and general method base on discrete convolution techniques for evaluating the expectation and variance of the frequency and duration (FeD) reliability indices, for a single area system. The proposed method models simultaneously and without any restriction, derated generating capacity generation units with all possible load levels. The expectation and variance of Frequency and duration reability indices are evaluated based on discrete convolution, treating the the generating unit transition rates and forecast peak loads as random variables. Clustering techniques are used to model the incertainties in the output capacities of hydraulic units. Under this interpretation, the reliability indices are also random variables. The generation capacity expansion problem under uncertainties, for a single-area system, is addressed from a new point of view. This method determines confidence limits (or percenties) not only for the loss of load indices but also for the frequency and duration failure indices. The generating capacity expansion is then calculated through risk indices percenties instead of the usual expected values. The proposed FeD method is then used for evaluating the impact of load and generating parameter uncertainties on typical systems including the IEEE Reliability Test System, and the Brazlian South/Southeastern Generating System.