An immune algorithm, which includes a fuzzy system and an annealing immune operator is presented. The proposed immune algorithm (PIA) is applied to the short-term unit commitment (UC) problem in power system operation. The PIA differs from its counterparts in three main aspects: (i) the crossover and mutation ratios are changed from having fixed values and become a variable that is determined by the fuzzy system; (ii) it uses a memory cell; and (iii) it uses an annealing immune operator. These modifications result in three major advantages for the PIA: (i) it does not fall into locally optimum solutions; (ii) it can quickly and correctly find the full set of globally optimum solutions; and (iii) it can easily obtain the most economic solution to the UC problem. In particular it can determine the start-up and shutdown schedules for the generation units so that they are able to meet forecasted demands at the minimum cost while satisfying adverse range of constraints. The PIA is used to generate schedules for cases containing 10, 20, 50, 70 and 90 generators. The schedules generated by the PIA are compared to those generated using the dynamic programming, Lagrangian relaxation, genetic algorithm, simulated annealing and tabu search methods. It is shown that the proposed method is valid and that it is able to produce excellent solutions.