Abstract: Unlike traditional energy sources,the large amount of history information of photovoltaic power generation-storageuse,the high complexity of constraints,and the lack of necessary data sharing process often lead to a poor accuracy of power prediction of PV clean energy output. To this end,the information sharing method is introduced into the PV prediction method. The information integration module is used to integrate PV clean energy data from production management system,grid planning system,supervisor information system,dispatching system,historical production data,market transactions,etc. An information sharing method is designed to reduce the degree of differentiation of PV clean energy data. The BP neural network algorithm is optimized by PSOEM to predict the PV clean energy output power,and the particle swarm algorithm with extended memory is introduced to improve the defects of the algorithm falling into local optima,and enhance the accuracy of PV clean energy information prediction and convergence speed,thus build a PV clean energy output power prediction model. The model test results show that the method can accurately realize the real-time sharing of photovoltaic power generation information,power storage information and power consumption information,and the HM accuracy of predicting the output power of PV power system reaches 96.3%,and the RMSE accuracy reaches 93%.