AI to Manage Agricultural Reservoirs... Pukyong National University Student Develops Reservoir Storage Prediction Model

Park Jaeseong Achieves Accurate Three-Day Forecasts
Published in the SCIE-Indexed Journal "Water"

Park Jaeseong, a senior majoring in Satellite Information Convergence Engineering at Pukyong National University, has developed an artificial intelligence (AI) model that can accurately predict the reservoir storage rate of agricultural reservoirs in Korea up to three days in advance.

Park Jaeseong, a student in the Artificial Intelligence Remote Sensing Laboratory (supervised by Professor Lee Yangwon), conducted this research and achieved the distinction of publishing a paper as the first author in an SCIE-indexed international journal as an undergraduate.

Currently, there are approximately 17,000 agricultural reservoirs in Korea, but most do not provide inflow and outflow data, making accurate reservoir storage prediction difficult. During the summer, rapid changes in water levels due to heavy rainfall increase the risk of flood damage, while water shortages for agriculture recur during droughts. As a result, the need for reservoir storage prediction technology has been consistently raised.

To address these issues, Professor Lee Yangwon and Park Jaeseong used a rainfall-runoff hydrological model. They proposed a time-series AI ensemble framework in their paper, which simulates reservoir inflow and outflow and uses these as additional input data for the model.

By applying this AI ensemble framework to predict reservoir storage rates, the model achieved top-level accuracy, with mean absolute errors (MAE) of 0.820 percentage points for one-day predictions, 1.339 percentage points for two-day predictions, and 1.766 percentage points for three-day predictions.

This framework enables highly accurate predictions even for reservoirs with irregular seasonal rainfall patterns and insufficient inflow and outflow data, making it useful for effective reservoir management.

This research was recently published in an SCIE-indexed international journal under the title "AI-Based Time-Series Ensemble Approach Coupled with a Hydrological Model for Reservoir Storage Prediction in Korea."

This content was produced with the assistance of AI translation services.