Deep Learning Framework with Time Series Analysis Methods for Runoff Prediction

Recent advances in deep learning, especially the long short-term memory (LSTM) networks, provide some useful insights on how to tackle time series prediction problems, not mention development of a model itself for prediction. Runoff forecasting is problem with past runoff data (water level and discharge data) as inputs fixed-length future output. Most previous work paid attention sufficiency input structural complexity while less effort has been put into consideration quantity or processing original data—such decomposition, which can better capture trend runoff—or unleashing effective potential learning. Mutual information seasonal decomposition are two methods handling analysis processing. Based former study, we proposed learning combined daily middle Yangtze River analyzed its feasibility usability frequently used counterpart models. Furthermore, this research also explored quality that affect performance model. With application method, effectively get about amount adopted The comparison experiment resulted different sites, implying improved precision much easier more practical application. In short, exert great may unleash artificial intelligence hydrology research.