基于GRU-CNN双网络输出构建BP模型的径流预测方法OACSTPCD

Improving the accuracy of runoff prediction is an important means to avoid flood disasters.However,the lack of known effective modeling samples during the prediction phase poses difficulties for the prediction.This paper proposes using dual network outputs to provide data reference for the prediction stage,combining the relationship between the dual network outputs and the true values in the training stage,and using quadratic multivariate modeling to achieve the runoff prediction in prediction stage.Firstly,the GRU and CNN deep learning networks are constructed to synchronously output two runoff prediction sequences.Secondly,within a known time period,a multivariate BP model between two predicted results and the measured values is constructed.Finally,based on the dual network output prediction values,the runoff prediction results are output through the determined BP model.After testing,the proposed method can provide reliable prior samples for predicting time periods,efficiently learn the relationship between network output and true values,and significantly improve prediction accuracy.