Long-Term Streamflow Forecasting Based on Relevance Vector Machine Model

摘要: Long-term streamflow forecasting is crucial to reservoir scheduling and water resources management. However, due the complexity of internally physical mechanisms in process influence many random factors, long-term a difficult issue. In article, we mainly investigated ability Relevance Vector Machine (RVM) model its applicability for forecasting. We chose Dahuofang (DHF) Reservoir Northern China Danjiangkou (DJK) Central as study sites, selected 500 hpa geopotential height northern hemisphere sea surface temperatures North Pacific predictor factors RVM Support (SVM) model, then conducted annual Results indicate that results DHF much better than DJK when using SVM, because latter basin has magnitude bigger 1000 m3/s. Comparatively, accurate both two basins can be gotten with Nash Sutcliffe efficiency coefficient 0.7, they are those from SVM model. As result, an effective approach forecasting, it also wide discharge dozen thousand cubic meter per second.