Short-term Runoff Prediction Optimization Method Based on BGRU-BP and BLSTM-BP Neural Networks

Abstract Runoff forecasting is one of the important non-engineering measures for flood prevention and disaster reduction. The accurate reliable runoff mainly depends on development science technology, many machine learning models have been proposed in recent years. Considering non-linearity real-time hourly rainfall data. In this study, two were proposed, which combination bidirectional gated recurrent unit backpropagation (BGRU-BP) neural network long short-term memory (BLSTM-BP) network. compared with (GRU), (LSTM), (BGRU), (BLSTM) models. research methods applied to simulate Yanglou hydrological station, Northern Anhui Province, China. results show that superior unidirectional model, (BP) based propagation was conducive improving generalization ability BP could better guide model find optimal nonlinear relationship. also BGRU-BP performs equally well as BLSTM-BP model. has few parameters a short training time, so it may be preferred method forecasting.