نوع مقاله : مقاله پژوهشی
نویسندگان
1 گروه مهندسی آب، دانشگاه فردوسی مشهد
2 گروه آب، دانشکده مهندسی عمران و نقشه برداری، دانشگاه تحصیلات تکمیلی صنعتی و فناوری پیشرفته کرمان
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
In this research, a group data classification model was developed to model and estimate the groundwater level. For this purpose, daily data (twenty-one years ago) of two observation wells drilled in the Illinois plain of the United States were used. The development of the GMDH model was based on two approaches: the conventional GMDH model and the NF-GMDH data group neural fuzzy classification model. The training of the mentioned models was performed using the least squares error algorithm and also the particle swarm optimization (PSO) algorithm. In designing the pattern of input variables, time delays (up to five units) of data related to groundwater level in these wells were used. The results of this study showed that the GMDH model can measure the level of groundwater level in Bondville well with statistical error indicators including in the experimental stage are R ^ 2 = 0.98 and the root mean square error RMSE = 0.333 and the average absolute error percentage MAPE = 9.9% and Predict in medium filling well with R ^ 2 = 0.99, RMSE = 0.64 and MAPE = 12% in the validation modeling stage. The results of the development of the NF-GMDH model showed that the accuracy of the development of the GMDH model based on the fuzzy neural approach is also of the same accuracy as the GMDH model. The accuracy of both GMDH and NF-GMDH models increases with increasing groundwater level so that the maximum values of groundwater level are accurately predicted,
کلیدواژهها [English]