Salinity Time Series Prediction Based on LSTMs Neual Network

Abstract

Almost all the waterworks of Hangzhou will be subjected to saltwater intrusion during dry season and spring tide every year, which will have a great influence on the lives of local residents. Therefore, it is of great significance to accurately predict saltwater intrusion. In view of the long-term correlation in the salinity time series data, continuous sampling and large amount of data, and the factors affecting salinity data are mostly non-linear and non-stationary. The Long Short-Term Memory (LSTMs) neural networks was used to modeling the salinity time series observed at Cangqian hydrological station for analysis and forecast. The prediction results of the model show that prediction of salinity time series by memory network is quite consistent, and it is very suitable for modeling, analysis and prediction non-linear and non-stationary hydrological time series.