Design of Basin Irrigation System using Multilayer Perceptron and Radial Basic Function Methods

The common use of an artificial neural network model has been in water resources management and planning. The length, width, and discharge of a basin were measured in this study utilizing field data from 160 Dashti Hawler existing projects. Multilayer Perceptron (MLP) and Radial Basic Function (RBF) networks were employed in the basin irrigation assessment. Input factors included the soil type, the conveyance system effectiveness, and the root zone depth. 130 projects were used for calibration, while the remaining 30 were used for validation. When developing the basin irrigation system, the models’ aforementioned indicators’ performance was evaluated using the coefficient of determination (R2), mean absolute error (MAE), relative error (RE), and Nash Sutcliff efficiency (NSE). For the basin's length, width, and discharge, the (R2) values for the MLP model were determined to be 0.97, 0.97, and 0.96, respectively, whereas the corresponding values for the RBF model were 0.88, 0.89, and 0.89. Compared to the RBF model, the values of (MAE) for basin length, width, and discharge for the MLP model were determined to be 8.99, 8.52, and 42.58, respectively. However, the (NSE) values for the models mentioned above were 0.95, 0.96, and 0.94, as well as 0.65, 0.66, and 0.66 for the basin’s length, width, and discharge, respectively. When it comes to building the irrigation system for the basin, the MLP is more precise than RBF depending on the values of (R2), (MAE), and (NSE). Finally, the ANN approach uses additional design options quickly examine which model is computationally efficient.