The purpose of load frequency control is to reduce transient oscillation frequencies than its nominal value and achieve zero steady-state error for it. A common technique used in real applications is to use the proportional integral controller (PI). But this controller has a longer settling time and a lot of Extra mutation in output response of system so it required that the parameters be adjusted as appropriate. In this paper, we aim to design a system based on PI controllers using IMPROVED PARTICLE SWARM OPTIMIZATION algorithm for load frequency control. Multi-population approach and local search to improve the OPTIMIZATION algorithms is used and displayed. That this approach will lead to accelerating the achievement of results, preventing entrapment in a local minimum, and get better system output compared with similar methods.