The performance of various Contingency Selection methods has been investigated within the framework of accuracy for application to steady-state power system security assessment and suitability for execution in a real-time environment. In the study the following requirements have been considered: (a) Effectiveness: in identifying contingencies which may cause limit violations and discarding all others; (b) Adaptability: to model both permanent and temporary changes in the system; (c) Flexibility: to model any number and type of contingencies; (d) Computational efficiency: in terms of speed in selecting the sub-set of contingencies as well as in terms of storage requirements; (e) Ability: to update and augment on-line the list of contingencies given the actual system operating data.