Dự báo mực nước sông Cấm, thành phố Hải Phòng bằng mô hình mạng Nơ-ron LSTM

Abstract

The Cam River is a big river in Hai Phong, holding an important position related to economy, national defense and culture not only of Hai Phong but also of Northern Vietnam. Recently, many large and modern urban centers have been built on the banks of the Cam River. Therefore, accurately forecasting the water levels in the Cam River will make an important contribution to flood prevention, ensuring the safety of people's lives and socio-economic development. Accordingly, the author of this article has set up a Long Short-Term Memory Neural Network (LSTM) model, a special type of the Recurrent Neural Network (RNN), to predict the water levels of the Cam River at Cua Cam station in Hai Phong. The input data of the forecast model is only the water levels measured at the hydrological stations in the study area. Rainfall at stations: Cao Kenh, Kien An, Phu Lien, Cua Cam have low correlation coefficients, so these data series are not used for the model. Nash Sutcliffe Efficiency, Root Mean Squared Error, Mean Absolute Error were used to evaluate the errors of the forecast values. The forecast results are highly accurate, predictive quality is sufficiently reliable. Therefore, this model can be applied to forecast the water levels of the Cam River and other rivers in Hai Phong.