This study aims to compare the performance of different models (multiple linear regression (MLR), artificial neural network (ANN), generalized regression neural network (GRNN), and long short-term memory (LSTM)) in predicting monthly evaporation rates. The data used evaporator measurements from 2000 to 2019, including air and water temperature, wind speed, relative humidity and precipitation at reservoir-dam Brezina. The models were evaluated using statistical indexes such as Nash–Sutcliffe Efficiency (NSE), Root Mean Squared Error (RSME), Mean Absolute Error (MAE), Coefficient of Determination (R2), Root Mean Square Ratio (RSR), and Willmott Index (WI). The GRNN model, trained with 70% of the available data, exhibited excellent performance compared to other models, with NSE = 0.99, RSR < 0.5 (0.08), WI = 0.998, and the minimum RSME of 11.25. Even after 15% validation and 15% testing, the GRNN model still shows very good results, with NSEs ranging from 0.95 to 0.96, RSR below 0.50, and WI ranging from 0.990 to 0.997. The generalized regression neural network model proves to be highly suitable for estimating evaporation rates, outperforming traditional methods such as evaporator pans or empirical formulas, which often lead to over or underestimation of evaporation from reservoir dams.