A mixed-integer nonlinear program is developed to help optimize post-earthquake electric power restoration, and it is applied to the Los Angeles Department of Water and Power (LADWP) electric power transmission system. The model determines how many inspection, damage assessment, and repair crews to train and where to have them respond so as to minimize the average time each customer is without power. It is solved using a genetic algorithm, and is tightly coupled with a discrete event simulation model of the restoration process that is used to evaluate each candidate solution. The study's key strengths include the detailed representation of the restoration process provided by coupling with the simulation model, simultaneous consideration of all earthquakes that might affect the system, and evaluation of results through comparison with the current LADWP crew allocations using three different measures of restoration effectiveness and including a representation of uncertainty. Copyright © (2006) by Earthquake Engineering Research Institute All rights reserved.