Selection of Similar Basin Area Sites and Model Inflow Characteristics for Dam Inflow Prediction Using Neural Network and Application to Training Data

Key Points

• The study aims to enhance inflow prediction for dams using neural networks by selecting similar basin sites as training data.
• Developed a neural network model for predicting inflow levels and flow rates over a 24-hour period.
• Identified dams with characteristics similar to the target dam to create a relevant training dataset.
• Evaluated the model's performance using real measurement data and Nash-Sutcliffe efficiency coefficient.
• Predicted inflow values showed similar variability to observed levels.
• Achieved a Nash-Sutcliffe coefficient of over 0.7 during inflow events, indicating good prediction accuracy.
• Identified that including data from dissimilar dams negatively affected prediction accuracy.