Hourly Water Level Forecasting in an Hydroelectric Basin Using Spatial Interpolation and Artificial Intelligence.

Area of Science:

• Hydrology
• Environmental Modeling
• Water Resource Management

Background:

• Accurate hydroelectric basin water level prediction is crucial for efficient water resource management.
• Existing models face challenges with data availability and prediction accuracy.
• Integrating diverse data sources can enhance hydrological forecasting.

Purpose of the Study:

• To develop and validate a new cascaded modeling approach for hydroelectric basin water level prediction.
• To compare spatial interpolation methods for environmental variable estimation.
• To assess the performance of a neural network and NARX model for hydrological forecasting.

Main Methods:

• Utilized weather stations, satellite observations, reanalysis data, and basin data.
• Compared Kriging, Radial Basis Function, and Natural Neighbours for spatial interpolation.
• Employed a neural network for soil moisture prediction and a NARX model for water level simulation.

Main Results:

• Achieved accurate hydroelectric basin water level predictions 1 to 6 hours ahead.
• Reported mean absolute errors (MAE) between 2 cm and 10 cm for the respective prediction horizons.
• Successfully integrated diverse data sources to improve prediction accuracy.

Conclusions:

• The proposed cascaded modeling approach demonstrates high accuracy in predicting hydroelectric basin water levels.
• The integration of spatial interpolation, neural networks, and NARX models offers a robust solution for hydrological forecasting.
• This approach can aid in optimizing hydroelectric power generation and water resource management.