Abstract: Aiming at the problem that the fluctuations and randomness of active power for photovoltaic clusters will significantly affect the safe and stable operation of the power system, based on the hierarchical control theory of large systems, the hierarchical prediction control strategy of active power for photovoltaic clusters is proposed. The strategy is divided into three levels in the spatial scale. At the day-ahead cluster tracking layer,with the goal to track the recent scheduling plan, a scheduling plan tracking model of the photovoltaic cluster is established, and the optimized scheduling plan is sent to the sub-cluster layer. At the subcluster layer, coordinated optimization allocation model is established, and at the same time, the control command issued by the cluster is reasonably assigned to each sub-cluster, and the optimized output value of the internal power station of the sub-cluster is given. At the real-time power station control layer, the actual power control model of the power station is established, and based on the operation conditions and received optimized output values, this model adaptively adopts the maximum output mode or following mode for operation. Finally, the experimental results show that the proposed method can improve the photovoltaic accommodation in the target area on the premise of ensuring the security and stability of the system, and make the active power of photovoltaic power stations more stable.