Water level forecasting at Hanoi station using transformer-based AI models

Tóm tắt

Flood forecasting is the main task to mitigate the damage caused by flooding in the Red River, Vietnam. Many reservoirs have been operating in the Red River to regulate flood. This research aims at developing a method for rapidly forecasting Hanoi’s water levels under various reservoir operation scenarios and river system conditions that will facilitate the assessment of multiple reservoir operation scenarios to provide effective and reliable real-time operational advice. A deep learning model based on the Transformer architecture was used to forecast the 24-hour lead time at Hanoi station. The dataset was divided into three subsets (Training set from 2015 to 2022, validation set in 2023 and Test set in 2024). The results showed that the Mean Absolute Error (MAE) was within an acceptable range, with MAE values of 24.1 cm, 26.1 cm, and 30.7 cm for the training, validation, and testing phases, respectively. The model demonstrated a significant ability to capture historical patterns and achieve high accuracy on the validation dataset, emphasizing the effectiveness of the Transformer architecture in forecasting water levels under normal conditions. Hydraulic models can be used to simulate additional data to improve the quality of flood forecasts for these extreme cases.