Forecasting for Haditha reservoir inflow in the West of Iraq using Support Vector Machine (SVM)

Area of Science:

• Hydrology and Water Resources Engineering
• Environmental Science
• Artificial Intelligence in Environmental Management

Background:

• Accurate inflow forecasting is crucial for effective flood management and optimizing water supply systems.
• Reservoir design and operational strategies heavily rely on precise inflow predictions.
• The Euphrates River, upstream of the Haditha Dam, is a vital water resource requiring reliable flow forecasting.

Purpose of the Study:

• To develop and generalize a machine learning model for predicting Euphrates River discharge.
• To evaluate the performance of Support Vector Regression (SVR) with different kernel functions for inflow forecasting.
• To identify the optimal SVR model configuration for daily, monthly, and seasonal flow prediction.

Main Methods:

• Utilized time series data of river flow from 1986-2024.
• Applied Support Vector Regression (SVR), a machine learning technique.
• Tested various SVR kernel functions: linear, Quadratic, and Gaussian (Radial Basis Function - RBF).
• Analyzed daily, monthly, and seasonal flow data.

Main Results:

• Daily flow prediction demonstrated superior performance compared to monthly and seasonal scales.
• The linear kernel SVR model, with a one-day time delay, achieved the highest accuracy.
• Performance was validated by a coefficient of determination (R2) of 0.95 and a root mean square error (RMSE) of 53.29 m3/sec for daily flow.
• The machine learning model proved effective in predicting daily river flow.

Conclusions:

• The developed machine learning model shows significant promise for accurate inflow forecasting of the Euphrates River.
• Effective daily flow prediction can substantially improve water resource management and dam operational efficiency.
• The study highlights the potential of SVR in hydrological forecasting applications.