Multi-input LSTM for water level forecasting in Black River at the border of Vietnam-China

Black River is an important section in the river system in Northern Vietnam with three of the Vietnamese biggest hydroelectric dams along its bank. Additionally, it plays an indispensable role in the area’s irrigation systems and agricultural activities. Hence, water level prediction on the river is highly necessary. In this paper, we apply the multi-input LSTM model with two datasets (rainfall and water level) to enhance the accuracy of forecasting the water level of the Lai Chau dam on Black River. The usage of additional rainfall data is essential in that rainwater pours straight down into the river due to the surrounding rocky mountainous terrain. The proposed model is compared to RNN model and LSTM model with one input used in 2019 by Le et al. [1], who also conducted a study on the Black River. The results have indicated that our multi-input model surpasses the compared ones with MSE of 0.036687 and MAPE of 0.068576.