An LSTM and Encoder-Decoder-Based Approach for Runoff Prediction

Runoff time series are characterized by nonlinearity, complexity, and variability. Effective runoff prediction can improve the information and intelligence of local water resources management. To improve the accuracy of daily runoff prediction, this paper addresses the multidimensional data of hydro...

Full description

Saved in:
Bibliographic Details
Main Authors: Lin, Xiuxiu, Yang, Shuang
Format: Conference Proceeding
Language:English
Subjects:
Convolutional neural networks
Data models
Encoder-Decoder
Jiulongjiang
LSTM
Meteorological factors
Predictive models
Recurrent neural networks
Runoff prediction
Support vector machines
Time series analysis
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Runoff time series are characterized by nonlinearity, complexity, and variability. Effective runoff prediction can improve the information and intelligence of local water resources management. To improve the accuracy of daily runoff prediction, this paper addresses the multidimensional data of hydrological data and multidimensional correlation meteorological factors to make multi-step prediction, and this paper proposes a sequence-to-series daily runoff prediction model based on encoder-decoder of long and short term memory network. To verify the validity of the model, six years of daily runoff data (2012-2017) and related meteorological factor data and three evaluation indexes (root mean square error, mean absolute error, and coefficient of determination) from the Punan hydrological station site in the Jiulong River basin of Fujian Province are selected for testing. The prediction results show that the proposed model has higher accuracy than the existing runoff prediction models such as gradient boosting (XGBoost), decision tree (DTA), support vector machine (SVR), convolutional neural network (CNN), recurrent neural network (RNN), and long-short time neural network (LSTM).
ISSN:2688-0938
DOI:10.1109/CAC57257.2022.10054711