Aiming at the problems of large scale, multi-constraint and non-linear optimal substation locating, an algorithm is proposed which adapts to the actual geography state, has optimal systems and ensures the efficient speed. Based on Genetic Algorithm and Particle Swarm optimization (GA-PSO), this algorithm adopts real-coded strategy and elite-preservation strategy, then uses a minimum annual cost of substation planning to be fitness. This algorithm realizes the self-adjusted search of solution in spatial range, and avoids the problem of prematurity and the situation of trapping in local best optimization. The speed of convergence of this algorithm is faster than that of GA and its precision is rather higher than those of PSO and GA. Besides, the results are evaluated and locally modified by the evaluation function based on Analytic Hierarchy Process (AHP), which is closer to the actual situation. The results demonstrate that GA-PSO algorithm has good converging speed and find-best ability, doesn’t need the process of coding and crossover, and the speed of convergence is fast and implementation is easy. The method proposed has a promising application in large-scale practical problems.