摘要
Runoff forecasting which subjects to model pattern and parameter optimization, has an important significance of reservoir scheduling and water resources management decision-makings. This paper proposed a new forecasting model coupled phase space reconstruction technology with relevance vector machine, and its model parameters is optimized by an improved PSO algorithm. The monthly runoff time series from 1953 to 2003 at Manwan station is selected as an example. The results show that the improved PSO has efficient optimization performance and the proposed forecasting model could obtain higher prediction accuracy.