Research on LSTM Water Level Prediction Model Based on Improved Attention Mechanism
Ma Fei 1, Tu Zhenyu 1*, Zhu Songting 2, Xiang Xinyue 1, Sun Yifei 1, Fang Qiang 1
(1. School of Information Engineering, Nanchang Engineering College, Nanchang, Jiangxi, 330099; 2. Jiangxi Flood Control Information Center, Nanchang, Jiangxi, 330009)
In order to further improve the accuracy of water level prediction, this paper proposes a prediction model that integrates an improved attention mechanism with a Long Short-Term Memory (LSTM) network. The model splits the input sequence into time series and feature sequences, introduces an attention mechanism before the LSTM network model to compute attention for the two sequences separately, and then fuses them. The LSTM network can adaptively select the most important input features based on their significance. The parameters of the attention mechanism layer are obtained through a competitive random search algorithm, which further enhances the robustness of the model. Finally, prediction experiments are conducted on the water level data of Poyang Lake. The experimental results show that compared to models such as Support Vector Regression (SVR) and LSTM, the proposed LSTM model based on the improved attention mechanism has better prediction accuracy, providing technical support for water level prediction and precise scheduling of water resources.
Keywords:Water Level Prediction; Prediction Model; Long Short-Term Memory Network; Attention Mechanism; Competitive Random Search
Classification Number:TP391.9
Document Identification Code:A
Article Number:1004-4701（2023）03-0162-06
Source: Jiangxi Water Conservancy Science and Technology, June 2023, Vol. 49, No. 3
Disclaimer:The content published by this public account (including original and reprinted text, images, and other materials from the internet) is copyrighted by the author. This public account is only for learning and reference purposes. If other media, websites, or individuals reprint or use it, they must retain the “source of the manuscript” indicated by this service and bear legal responsibility.
