When a photovoltaic array panel is exposed to uneven light, the power-voltage (P-V) characteristic curve becomes multi-peaked, and conventional maximum power point tracking (MPPT) algorithms will not be able to track the correct global maximum power point (GMPP), and artificial intelligence algorithms with global search capabilities are usually highly parameterized and complex. To address the above problems this paper proposes a composite tracking algorithm combining the adaptive cuckoo algorithm and perturbation observation method (ACS-P&O). This improved method takes the switching probabilities and Lévy flight step coefficients from the cuckoo search (CS) algorithm and adaptively adjusts them to extend the search range of the algorithm at an early stage. The introduction of a processing strategy for bounding individuals further reduces the number of algorithm iterations. The improved algorithm makes it easier for the system to jump out of the local maximum power point (LMPP), while at a later stage the algorithm operates precisely in a small area, improving the local exploitation capability. The addition of the perturbation and observation (P&O) method mitigates power oscillations when the system is located near the GMPP and stabilizes the output. Simulation results show that the ACS-P&O composite algorithm can adapt to the effects of environmental changes and track the GMPP quickly and accurately. This work is supported by the Natural Science Foundation of Shaanxi Province (No. 2021JM-393).