In order to solve the problem that the multi-peak power voltage (P-U) characteristic curve generated by a photovoltaic power generation system under unbalanced illumination and the traditional maximum power point tracking (MPPT) algorithm easily falls into the local optimal solution, a photovoltaic multi-peak MPPT method based on an improved manta ray foraging algorithm is proposed.The algorithm combines several mathematical optimization strategies with the original manta ray foraging optimization algorithm. Firstly, chaos mapping is used to initialize the search population, increasing population diversity. At the same time, reverse learning is integrated after summarizing the characteristics of photovoltaic P-V characteristic curves to improve the distribution efficiency of the population; Secondly, a nonlinear change function is introduced to provide a better balance between the development and exploration of the algorithm; Then the linear change curve is introduced to improve the ability of the algorithm to jump out of the local extreme value; Finally, levy flight strategy is introduced to improve the search ability of the algorithm. It is proved by the simulation results of MATLAB/Simulink, In the process of tracking, the algorithm has a small oscillation amplitude, good tracking accuracy, and good convergence time.