Performance comparison of gene expression programming and artificial neural network methods to estimate water distribution uniformity in sprinkler irrigation

Uniformity of sprinkler irrigation is an important technical parameter in the design of sprinkler irrigation systems and its value has a significant impact on the quality and the efficiency of investment in irrigation projects. With sprinkler irrigation, determining CU from a single sprinkler is time consuming due to the overlap of adjacent sprinklers and in different amounts of operating pressures (P), riser heads (Hr), sprinkler spacing on laterals (Sl) and the distance between laterals (Sm). In this research, CU quantities of two model sprinklers (AQ-20 and KA-6) were measured in slow wind velocity (0-2 m/s) , at Hashemabad cotton research station of Gorgan city under 4 different operating pressures (2, 2.5, 3 and 3.5 at), 16 distances of sprinklers (including 9×18, 12×18, 15×18, 18×18, 9×15, 12×15, 15×15, 18×15, 9×12, 12×2, 15×12, 18×12, 9×9, 9×12, 9×15, 9×18) , 4 riser heads (60, 90, 120 and 150 cm) and 3 arrangements of sprinklers (square, rectangular and triangular). For estimating CU based on gene expression programming (GEP) and artificial neural network (ANN) methods P, Hr, Sl and Sm were selected as the input variables. By statistical comparison of results, root mean squared error (RMSE) for AQ-20 sprinkler in GEP and ANN methods were obtained as 0.06 and 0.062 and for KA-6 sprinkler as 0.067 and 0.064, respectively. The results indicate the high accuracy of the two methods for modeling since GEP is capable of estimating an explicit equation for estimating CU, it incorporates a more practical feature.