In a small hydropower (SHP) enriched area

accurate forecast of SHP generation is very important for ensuring power grid safety and coordinating SHP generation with other power sources. Different from the generation forecast of large hydropower

most SHP stations are located in remote mountainous areas where information gathering is hard and management weak

making it nearly impossible to make use of the locally available forecast methods. By considering the actual situation of SHP stations

and with the SHP stations of an area as the object

a SHP generation forecast method is proposed. This method uses the improved BP neural network as the forecast model and partial mutual information to select remarkable input variables. In the power forecast stage

climate forecast system (CFS) data is used to get the precipitation of the target area and as the input of the model to forecast SHP generation. The case study of Dehong and Dali in Yunnan shows the effectiveness of the proposed method. This work is supported by National High Technology Research and Development Program of China (863 Program) (No. 2012AA050205) and National Natural Science Foundation of China (No. 51209029).