As a prevailing solar energy utilization equipment, the three-phase grid-connected photovoltaic (PV) inverter is widely operated in partially shaded conditions and thus tends to generate multiple local maximum power points on its power-to-voltage and current-to-voltage characteristic curves. In order to identify the global maximum power point (GMPP) quickly and precisely, this paper proposes a ripple-based maximum power point tracking method. It aims to perform the optimization of tracking using the segmented scanning of DC-side voltage. An improved adaptive perturb and observe (AP&O) method is introduced to maximize the solar conversion and to increase working stability. This method applies a hybrid model of fixed and variable step-size perturbation to restrain the fluctuation of PV-side voltage. It belongs to a two-stage GMPP tracking method. That is, when environmental factors such as irradiance and temperature change quickly PV power fluctuates sharply. Correspondingly, the AP&O method tracks the latest maximum power point (MPP) with a large fixed-step voltage reference command. When the PV power fluctuates smoothly under a slow environmental change rate, the algorithm applies multiple small and variable step-size voltage perturbations to vibrate round the location of GMPP. Simulation and experimental results show that this method improves the efficiency of the PV inverter tracking performance. In addition, the stability of DC bus voltage is guaranteed, and the operational stability of the photovoltaic power generation system is improved.