As one of the important components of heterogeneous energy, the distributed photovoltaic power will affect the safety and stability of the traditional distribution network, due to the randomness and the fluctuation of its output. With the increase of power generation capacity, the ability of distribution network to incorporate the distributed photovoltaic power becomes the main constraint. Therefore, under the premise of considering multiple operation targets, a multi?model learning method is constructed in this paper, which consists of a PV force change prediction model, a load distribution prediction model, and a distributed PV maximum capacity evaluation model. The impacts of photovoltaic access on the distribution networks during the whole process can be analyzed by the proposed model. The Elman neural network model is proposed to ensure the prediction accuracy of the photovoltaic power generation variations. The BP neural network model is established to consider the prediction accuracy as well as the prediction efficiency. The PSO model for the maximum capacity of distributed photovoltaic power generation is built, so as to realize the accurate evaluation of the limit of distribution network acceptance of distributed photovoltaic power. The empirical results show that the proposed method can ensure the safe and stable operation of the distribution network, and is also beneficial for the planning of the maximal capacity of distributed photovoltaic power generation in the distribution networks.