A major problem related to the growing popularity of grid connected photovoltaic power generation is the operational challenges (such as overload and overvoltage) due to the variability of photovoltaic power generation. Flexible power point tracking (FPPT) can limit the photovoltaic output power to a specific value to solve some integration problems. The traditional FPPT algorithm based on disturbance observation has the problem of slow dynamics. Therefore, an adaptive FPPT algorithm is proposed in this paper. The proposed algorithm has fast dynamic characteristics under the quick changing environment conditions (such as cloud layer passing), while maintaining low power oscillation in the steady state. The proposed algorithm uses additional measurement samples under each disturbance to observe changes in operating conditions (such as solar irradiance). Then, the voltage step is calculated adaptively according to the observation conditions (such as transient or steady state) to improve the tracking performance. Finally, the simulation experiment on a 3 kVA single-phase photovoltaic grid connected system verifies the effectiveness of the algorithm in terms of fast dynamics and high accuracy under different operating conditions.