In recent years, the global photovoltaic (PV) industry has expended rapidly. The large-scale development of distributed PV systems will inevitably have an impact on traditional distribution network. Based on the collection of multiple data, this paper proposes a data-driven approach to forecasting the distribution of distributed PV systems, which is instructive for distribution network planning and energy policy making. The proposed approach firstly investigates the PV adoption drivers based on the quantitative analysis of historical PV data, then simulates the spatio-temporal diffusion of distributed PV systems using the cellular automation model which can also be used to forecast the development and distribution of installed distributed PV capacity on the basis of multi-source datasets. The proposed forecasting approach is finally applied to analyze the distributed PV systems in Pudong district of Shanghai, China. The forecasting results verify the effectiveness of the approach.