Runoff Forecasting using Convolutional Neural Networks and optimized Bi-directional Long Short-term Memory

Junhao Wu, Zhaocai Wang (), Yuan Hu, Sen Tao and Jinghan Dong
Additional contact information
Junhao Wu: Shanghai Ocean University
Zhaocai Wang: Shanghai Ocean University
Yuan Hu: Shanghai Ocean University
Sen Tao: East China University of Science and Technology
Jinghan Dong: Shanghai Ocean University

Water Resources Management: An International Journal, Published for the European Water Resources Association (EWRA), 2023, vol. 37, issue 2, No 18, 937-953

Abstract: Abstract Water resources matters considerably in maintaining the biological survival and sustainable socio-economic development of a region. Affected by a combination of factors such as geographic characteristics of the basin and climate change, runoff variability is non-linear and non-stationary. Runoff forecasting is one of the important engineering measures to prevent flood disasters. The improvement of its accuracy is also a difficult problem in the research of water resources management. To this end, an ensemble deep learning model was hereby developed to forecast daily river runoff. First, variational mode decomposition (VMD) was used to decompose the original daily runoff series data set into discrete internal model function (IMF) and distinguish signals with different frequencies. Then, for each IMF, a convolutional neural network (CNN) was introduced to extract the features of each IMF component. Subsequently, a bi-directional long short-term memory network (BiLSTM) based on an attention mechanism (AM) was used for prediction. A Bayesian optimization algorithm (BOA) was also introduced to optimize the hyperparameters of the BiLSTM, thereby further improving the estimation precision of the VMD-CNN-AM-BOA-BiLSTM model. The model was applied to the daily runoff data from January 1, 2010 to November 30, 2021 at the Wushan and Weijiabao Hydrological Stations in the Wei River Basin, and the RMSEs of 3.54 and 15.23 were obtained for the test set data at the two stations respectively, which were much better than those of EEMD-VMD-SVM and CNN-BiLSTM-AM models. Additionally, the hereby proposed model is proven to have better peak flood prediction capability and adaptability under different hydrological environments. Based on this sound performance, the model becomes an effective data-driven tool in hydrological forecasting practice, and can also provide some reference and practical application guidance for water resources management and flood warning.

Keywords: Attention mechanism; Bayes optimization algorithm; Bi-directional long short-term memory; Convolutional neural networks; Daily runoff prediction; Variational mode decomposition (search for similar items in EconPapers)
Date: 2023
References: View references in EconPapers View complete reference list from CitEc
Citations: View citations in EconPapers (2)

Downloads: (external link)
http://link.springer.com/10.1007/s11269-022-03414-8 Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:waterr:v:37:y:2023:i:2:d:10.1007_s11269-022-03414-8

Ordering information: This journal article can be ordered from
http://www.springer.com/economics/journal/11269

DOI: 10.1007/s11269-022-03414-8

Access Statistics for this article

Water Resources Management: An International Journal, Published for the European Water Resources Association (EWRA) is currently edited by G. Tsakiris

More articles in Water Resources Management: An International Journal, Published for the European Water Resources Association (EWRA) from Springer, European Water Resources Association (EWRA)
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().