A photovoltaic power interval prediction is proposed based on an optimized Back Propagation (BP) neural network by Simulated Annealing Modified Particle Swarm Optimization (SAPSO-BP) and quantile regression. First, a dynamic SAPSO algorithm is used to predict the Photovoltaic (PV) output, taking the local minimization of BP and precocious convergence of the particle swarm into account. Then it optimizes the parameter setting of the BP neural network and improves the accuracy of prediction. From a deterministic prediction of PV power, the prediction power error of the SAPSO-BP model under different weather types is analyzed. Finally, a quantile regression model is established based on uncertain weather factors to realize the interval analysis of future power fluctuation. The model does not have to predict the distribution of the PV predicted power error, and the calculation method is simple. The fluctuation range of the predicted power of the PV can be provided at any confidence level. The model provides more abundant information for power system scheduling decision and operational risk assessment. This work is supported by the National Natural Science Foundation of China (No. 51577086).