Short-term prediction of dissolved oxygen in a river system based on a hybrid deep-learning prediction methodology

Yue YANG, Yi LIU, Yan YU, Zhuoying ZHANG

Journal of Systems Science and Information ›› 0

PDF(2015 KB)
PDF(2015 KB)
Journal of Systems Science and Information ›› 0 DOI: 10.21078/JSSI-E2022009

Short-term prediction of dissolved oxygen in a river system based on a hybrid deep-learning prediction methodology

  • Yue YANG, Yi LIU, Yan YU, Zhuoying ZHANG
Author information +
History +

Abstract

This paper proposes a hybrid deep-learning prediction methodology that integrates rolling variational mode decomposition (VMD) and long short-term memory (LSTM) neural network optimized by tunicate swarm algorithm (TSA) for the short-term prediction of dissolved oxygen (DO) in a river system. We use a rolling VMD method at first to extract the variation characteristics of different frequencies in the previous period for each time’s prediction. The decomposition results, the history data of DO, other water quality parameters, and some climatic parameters are served as features to construct a prediction model based on the LSTM optimized by TSA. The water quality data of the Dongyang River is used to examine the effectiveness of this prediction methodology. The experimental results demonstrate that the model has better effectiveness and robustness compared to other benchmark models in terms of accuracy and correlation, which illustrates the proposed prediction method can be recommended as a promising method for water quality forecasting, especially in some small tributaries.

Cite this article

Download Citations
Yue YANG, Yi LIU, Yan YU, Zhuoying ZHANG. Short-term prediction of dissolved oxygen in a river system based on a hybrid deep-learning prediction methodology. Journal of Systems Science and Information, 0 https://doi.org/10.21078/JSSI-E2022009

Funding

National Natural Science Foundation of China (71988101, 71874183) and The 14th Five-Year Plan special program of Wuxi Municipal Public Utility Industry Group.
PDF(2015 KB)

430

Accesses

0

Citation

Detail

Sections
Recommended

/