In view of the current factors affecting the power generation of distributed photovoltaic power generation systems, it is difficult to predict, and the operation optimization strategy between the other power generation systems is not perfect. Based on the typical experience of big data applications in photovoltaic industry at home and abroad, based on photovoltaic power generation data and user load demand data, this paper proposes an RBF neural network based photovoltaic power generation forecasting algorithm and load forecasting model, which is normalized by data. Processing and quantification and similarity processing of weather factors, forecasting PV power consumption and load in a certain period of time; and using the actual data of a PV power plant in Qingdao to learn and predict, and achieve better results, thus verifying the feasibility of the model. In addition, through the prediction of load and the prediction of power generation, the operation optimization strategy is formulated with the goal of economic performance optimization, which realizes the effective utilization of photovoltaic power generation, balances power between power generation side and load side, and greatly reduces network loss and line. Loss increases the reliability and economy of distributed photovoltaic power.