Abstract: Combining the theory of gray system with self memory principle, a new model for predicting time series of reservoir inflow is proposed. The model has been used to predict the annual and flood period inflow to reservoir of Caohe Hydropower Station on the Jin River. The case study indicates that the model can deal with non stationary time series and provide the inflow trend and extreme values with improved accuracy.

Key words: inflow, self memory, gray model, prediction