Small sample learning and LSTM (Long Short Term Memory)-based runoff prediction method for areas lacking data

Abstract

The invention discloses a runoff prediction method for a region lacking data based on small sample learning and LSTM. The method comprises the steps of firstly obtaining time series data of influence factors related to runoff intensity in a drainage basin, then performing feature extraction, and constructing a time series data set corresponding to features and predicted runoff; establishing a runoff prediction model based on small sample learning prototype network and LSTM fusion, and carrying out model training and verification; and then, inputting the influence factors to predict the runoff. In consideration of the current situation of data shortage of an area lacking data and the dependence of traditional machine learning on a large amount of label data, a data-driven deep learning strategy is used, a small sample learning algorithm is combined, a prototype network and LSTM are fused to reduce the dependence of the model on the data, and it is ensured that the model is more accurate under the condition that few training samples are input. The model can also obtain high prediction precision, and a new method is provided for runoff prediction in areas lacking data.

Description

technical field [0001] The invention belongs to the field of hydrology and water resources, and more specifically, relates to a method for predicting runoff in data-deficient areas based on small sample learning and LSTM. Background technique [0002] Runoff prediction is one of the most important scientific issues in hydrological research. Due to the intensification of global warming and the increasing frequency of extreme weather, heavy rain and flood disasters have caused huge economic losses and ecological damage on a global scale. In the changing environment, for runoff forecasting, improving its forecasting accuracy and prolonging the forecast period is an urgent problem to be solved, which has great social and ecological value. [0003] At present, most of the runoff prediction research at home and abroad is aimed at watersheds with available data, and based on the existing measured data to establish empirical relationships or models for prediction. However, there a...

AI Technical Summary

Problems solved by technology

However, LSTM needs to compare a large amount of labeled data for training in order to have a high enough prediction accuracy.

Images