A unit commitment（UC） algorithm based on the improved immune algorithm is proposed according to the characteristics of UC， especially those of the security constraints in the power system. The advantage of a typical immune algorithm lies in handling 0-1 variables. This algorithm has improved the immune algorithm in two aspects. First， an antibody segment is used to represent a unit state during a dispatching period. Furthermore， information is stored in an antibody memory library by decomposing the good antibodies. Second， the information on the active security constraints is stored as an expanded part of antibody. A security constraints memory library is formed to simulate the artificial active constraint selection process during an economical dispatching. Furthermore， an optimization mutation based on group search is adopted to reduce the blindness of random mutation. The IEEE 118-bus system is used to verify the validity and rationality of the proposed method.