The evaluation of generation and transmission reliability in power system, usually referred to composite reliability, allows specific identification of problems in power supply related to consumer location in transmission network. This work presents an efficient computational algorithm for calculation of composition of composite reliabillity indices using the frequency and duration method, based on Markov techniques. The obtained indices are considered as an adequacy measure of generation and transmission availability in respect to different load levels. The evaluation of these indices is carried out for two distinct operational situations: before and after remedial actions (generation redispatch and load shedding). The main contributions of this dissertaion are concetrated on two fundamental aspects for the composite reliability evaluation: consistent reduction in the number of possible scenarios for generation avaibility, which consequently reduces the state space sizes to be evaluated, and an efficient tool for the performance analysis of selected composite states (generation, transmission and load demand). This analysis is carried out by na algorithm which linearizes load flow equations for defined state space allocations (multilinear model). The proposed algorithm is applied to IEEE reliability Teste System and the results are fully discussed.