River flood prediction method based on deep learning model interpretable research strategy

Abstract

The invention discloses a riverway flood prediction method based on a deep learning model interpretable research strategy, and the method comprises the steps: building a more visual and reliable interpretable research strategy of a deep learning model Net-T, i.e., employing an interpretable assembly to construct a model Net-S, and achieving the prediction of the riverway flood on the premise of guaranteeing the minimum prediction capability difference of the two types of models. Carrying out substitution research on the deep learning model Net-T, and carrying out distillation training on the model Net-S so as to extract knowledge contained in the trained model Net-T to the model Net-S; selecting the trained model Net-S with the optimal prediction result according to the statistical indexes of the prediction results of the plurality of trained models Net-S, and determining the optimal input step length and the optimal prediction step length of the trained model Net-S; based on the optimal input step length and the optimal prediction step length, the trained model Net-T is adopted to predict the water level or flow of the river section.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and more specifically relates to a river flood prediction method based on a deep learning model that can explain research strategies. Background technique [0002] River flood forecasting is an important non-engineering measure to control flood disasters. Using the river flood forecasting method can not only directly provide the magnitude and occurrence time of flood disasters, but also provide accurate flood information for reservoir flood control scheduling, and then through the flood control function of the reservoir Reduce downstream flood disasters. Therefore, studying how to improve the reliability and accuracy of river flood prediction and constructing a river flood prediction model with high prediction accuracy has gradually become one of the important ways to control flood disasters in the river basin in the future. [0003] In recent years, deep learning models have bee...

AI Technical Summary

Problems solved by technology

[0005] In view of the above defects or improvement needs of the prior art, the present invention provides a river flood prediction method based on a deep learning model that can explain the research strategy, thereby solving the problem that the existing deep learning model for river flood prediction has low prediction accuracy technical issues

Images