Today, the large-scale Photovoltaic (PV) power system, connected to grid, is in their advanced development stage and is extremely interest in whole world. The real large-scale PV array can be partially shaded by the shadow of building, cloud, bird and dirt. The output characteristic of PV materials in partially shaded conditions is strong nonlinear, and there are multi local peaks in output power voltage curve, and the only one real peak exists in these local peaks. Certainly, the maximum power point tracking (MPPT) method is very important to extract the as much as possible energy from the costly PV materials. The variant weight Particle Swarm Optimization (PSO) method is proposed to track the real peak by using the excellent multi-peak value optimization characteristic of PSO algorithm. The simulation results shows that the proposed PSO method can improve the response speed and output efficiency of PV materials in partial shading as compared to the conventional Incremental conductance (IC) method.