This paper presents a novel optimization technique using genetic algorithms to develop an optimized emergency defence plan for power systems. The technique determines the optimal combination of generator tripping, load shedding, and islanding to regain system stability following severe contingencies. It was applied to the Libyan power system using time-domain simulations to evaluate solutions. Results showed the optimized defence plan required less load shedding than the existing Libyan plan and improved system response during a 2003 blackout event.