Under partial shading conditions, the power-voltage characteristic curve of a photovoltaic (PV) array exhibits multiple peaks, making it difficult for traditional maximum power point tracking (MPPT) techniques to accurately track the global maximum power point. To address this issue, an MPPT control method based on an improved arithmetic optimization algorithm (AOA) is proposed. First, the Sobol sequence is used to generate an initial population with a uniform distribution to enhance population diversity. Next, to balance the global search and local exploitation capabilities of AOA, the weight of the acceleration function in the mathematical optimizer of AOA is restructured. Finally, the Lévy flight strategy is introduced into the position update process of AOA, and quasi-oppositional learning is applied to the best solution after each update to enhance the algorithm’s convergence speed and its ability to escape local optima. Simulation and experimental results show that the improved algorithm can accurately and quickly track the global maximum power point under different partial shading and sudden light change conditions, with minimal power oscillation.