Abstract:

To understand future conditions and plan for the optimal allocation of water resources across various sectors such as drinking water, agriculture, and others, predicting inflows to water resource systems is essential for water resource management. The objective of this study is to predict the monthly inflow rates to Golestan Dam for the future. Hydrometric data from four stations—Ghoochamaz, Pol Kooseh, Ghareh Shur, and Oghan—were used, along with two models: Neural Network and Support Vector Machine (SVM), to make predictions. The results of different models were compared.

Based on the evaluation criteria, the SVM model outperformed the Neural Network model at three stations: Ghareh Shur, Pol Kooseh, and Oghan. Only at Ghoochamaz station did the Neural Network model show better performance. According to the findings, the best model for predicting monthly inflows to Golestan Dam is the Support Vector Machine model.