The perturbation observation method is one of the most commonly used methods to track the maximum power of photovoltaic (PV) system, but there are few reviews on perturb and observe (P&O) algorithm. In this paper, the maximum power point tracking (MPPT) based on P&O algorithm is discussed. The classical fixed-step P&O algorithm, improved classical P&O algorithm and hybrid P&O combine with intelligent algorithms are analyzed in terms of accuracy and fast tracking. By comparing the study of MPPT method in P&O algorithm, it is found that in recent years, researchers have improved the classical P&O algorithm through the improved classical P&O algorithm and the hybrid algorithm of P&O combined with machine learning/meta-heuristic. These two methods can reduce the oscillation of MPP, accelerate the convergence speed, improve the tracking accuracy, reduce the power loss and improve the conversion efficiency of PV system.