Dynamic Prediction Model for Short-Term Reservoir Carryover Storage Forecasting

Accurate short-term forecasting of carryover storage is crucial for effective reservoir management and water resource decision-making. Real-time information plays a vital role in reservoir operation decisions, highlighting the need to incorporate such information in carryover storage forecasting. In this study, we propose a dynamic prediction model (DPM) that integrates static and dynamic factors, including runoff forecasting, real-time water level, and remaining time from carryover level, to improve short-term reservoir operation. The model is applied to three reservoirs in China, and its performance is evaluated using statistical measures. The results demonstrate that the DPM surpasses the traditional static prediction model, yielding enhanced accuracy in reservoir carryover storage prediction. The inclusion of weakly correlated real-time information contributes to the improvement of forecasting accuracy. Moreover, we observe variations in the importance of dynamic factor combinations across different seasons. Notably, in the wet season, the combination of runoff forecasting and real-time water level factors significantly enhances forecast accuracy. This study highlights the potential of employing DPMs that incorporate real-time information for short-term reservoir operation, leading to improved reservoir management and decision-making. The findings emphasize the importance of considering real-time information and seasonality in carryover storage forecasting, thereby facilitating more effective water resource utilization and reducing the risks associated with floods and droughts.