Abstract: Aiming at the problem of power multi-peaking in locally shaded photovoltaic power generation, this paper proposes an improved squirrel search algorithm. In order to improve the global search ability of this algorithm, Tent mapping and dynamic refraction reverse learning strategy are introduced to jump out of the local search and enter the global search scope. A 30×50 PV array is built by the simulation platform, and the maximum power point tracking simulation is carried out. And it is compared with the traditional bionic algorithm: particle swarm optimization algorithm(PSO). The results show that the improved SSA algorithm has small fluctuations in the modulation process while ensuring short response time, good GMPP tracking effect, and effectively reduces the energy loss in the power generation process of PV array.