Under the conditions of partial shading , many peaks will exhibited in the power characteristic curve of the PV array . The traditional MPPT methods are ineffective and are easy to fall into the local optimum. Particle swarm optimization can effectively solve the multi-peak problem by its strong global search ability, but the conventional PSO algorithm has a slow convergence rate and is easy to be precocious. An improved particle swarm genetic algorithm is proposed .The algorithm can balance the ability between local search and global optimization by changing the inertia weight and learning factors constantly. And use the crossover and mutation operation of the genetic algorithm to increase the diversity of the population. The simulation results show that improved algorithm has good tracking speed and precious in the process of multi-peak maximum power tracking.