مدل‌سازي جريان روزانه رودخانه با استفاده از فرامدل‌هاي شبيه‌ساز (مطالعه موردي:رودخانه گاماسياب)

Modeling Daily River Flow Using Simulator Meta-Models (Case study: Gamasiab River)

زﻣﯿﻨﻪ و ﻫﺪف: ﻫﺪف در اﺑﺘﺪا ﺑﯿﺎن ﻧﻤﻮدن ﺗﻔﺎوتﻫﺎ و ﺷﻨﺎﺳﺎﯾﯽ3 ﻣﺪل ﺑﻪ ﻧﺎمﻫﺎي، ﺑﺮﻧﺎﻣﻪ رﯾﺰي ﺑﯿﺎن ژن )GEP(، ﺷﺒﮑﻪ ﻋﺼﺒﯽ ﻓﺎزي )ANFIS( و ﺷﺒﮑﻪ ﺑﯿﺰﯾﻦ )BN( اﺳﺖ و ﻣﻘﺎﯾﺴﻪ آﻧﻬﺎ ﺑﺎ ﯾﮑﺪﯾﮕﺮ و ﺳﻮال اﺳﺎﺳﯽ ﺗﺤﻘﯿﻖ اﯾﻦ اﺳﺖ ﮐﻪ اﯾﺎ ﻓﺮاﻣﺪل ﺷﺒﯿﻪﺳﺎز ﺑﺮﺗﺮ در اﯾﻦ ﻣﻄﺎﻟﻌﻪ ﻣﯽﺗﻮاﻧﺪ در ﺷﺮاﯾﻂ ﮐﻤﺒﻮد داده و اﻃﻼﻋﺎت، ﺟﺎﯾﮕﺰﯾﻦ ﻣﻨﺎﺳﺒﯽ ﺑﺮاي ﻣﺪلﻫﺎي ﻣﻔﻬﻮﻣﯽ ﺑﺎﺷﺪ. روش ﺑﺮرﺳﯽ: دادهﻫﺎي ﻣﻮرد اﺳﺘﻔﺎده ﺑﺮاي اﯾﻦ ﭘﮋوﻫﺶ، دادهﻫﺎي ﺑﺎرش و ﺟﺮﯾﺎن روزاﻧﻪ رودﺧﺎﻧﻪ ﮔﺎﻣﺎﺳﯿﺎب ﻧﻬﺎوﻧﺪ در ﯾﮏ دوره 10 ﺳﺎﻟﻪ 1381-1391 ﻣﯽﺑﺎﺷﺪ. ﺑﺮاي ﻣﺮﺣﻠﻪ ﭘﯿﺶﺑﯿﻨﯽ ﯾﺎ ﺷﺒﯿﻪﺳﺎزي از دادهﻫﺎي ﺳﺎل آﺑﯽ 1390-1391 اﺳﺘﻔﺎده ﺷﺪه اﺳﺖ. ﯾﺎﻓﺘﻪﻫﺎ: در ﻣﺮﺣﻠﻪ آﻣﻮزش و ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺿﺮﯾﺐ ﺗﺒﯿﯿﻦ و ﭘﺎراﻣﺘﺮ ﺟﺬر ﻣﯿﺎﻧﮕﯿﻦ ﻣﺮﺑﻌﺎت ﺧﻄﺎ و ﻣﻌﯿﺎر AIC، ﻣﺸﺎﻫﺪه ﻣﯽ ﺷﻮد ﮐﻪ در ﻫﺮ 3 ﻣﺪل، ﻫﻢ در ﻣﺮﺣﻠﻪ آﻣﻮزش و ﻫﻢ در ﻣﺮﺣﻠﻪ ﺗﺴﺖ ﺷﺎﻫﺪ اﺧﺘﻼف ﺑﺴﯿﺎر اﻧﺪك در ﻣﻘﺪار اﯾﻦ ﭘﺎراﻣﺘﺮﻫﺎ ﻫﺴﺘﯿﻢ و ﻧﺘﺎﯾﺞ ﻫﺮ 3ﻣﺪل ﺗﻘﺮﯾﺒﺎ ﺑﺎ اﺧﺘﻼف ﺑﺴﯿﺎر اﻧﺪك، ﻧﺰدﯾﮏ ﺑﻪ ﻫﻢ اﺳﺖ و ﺗﻘﺮﯾﺒﺎ ﺑﺮﺗﺮي ﻧﺴﺒﯽ ﻣﺪل GEP را ﻣﯽﺗﻮان ﻣﺸﺎﻫﺪه ﮐﺮد. ﺑﺤﺚ و ﻧﺘﯿﺠﻪ ﮔﯿﺮي: ﻧﺘﺎﯾﺞ ﺑﯿﺎﻧﮕﺮ آن اﺳﺖ ﮐﻪ ﻓﺮاﻣﺪل5 ﺷﺒﯿﻪ ﺳﺎز ﺑﯿﺎن ژن ﺗﻮاﻧﺎﯾﯽ ﺧﻮﺑﯽ ﺑﺮاي ﺷﺒﯿﻪ ﺳﺎزي و ﭘﯿﺶ ﺑﯿﻨﯽ ﺟﺮﯾﺎن روزاﻧﻪ رودﺧﺎﻧﻪ دارد و اﯾﻦ ﻓﺮاﻣﺪل ﺷﺒﯿﻪﺳﺎز، ﻣﯽﺗﻮاﻧﺪ در ﺷﺮاﯾﻂ ﮐﻤﺒﻮد داده و اﻃﻼﻋﺎت، ﺟﺎﯾﮕﺰﯾﻦ ﻣﻨﺎﺳﺒﯽ ﺑﺮاي ﻣﺪلﻫﺎي ﻣﻔﻬﻮﻣﯽ ﺑﺎﺷﺪ. ﻋﻼوه ﺑﺮ اﯾﻦ ﺳﺮﻋﺖ اﺟﺮاي ﻣﺪل ﺑﺮﻧﺎﻣﻪرﯾﺰي ﺑﯿﺎن ژن ﻧﺴﺒﺖ ﺑﻪ ﺑﻘﯿﻪ ﻣﺪلﻫﺎ ﺑﯿﺸﺘﺮ ﺑﻮده و در زﻣﺎن ﮐﻮﺗﺎﻫﯽ ﻗﺎدر ﺑﻪ اراﺋﻪ ﻧﺘﺎﯾﺞ ﺑﻮده اﺳﺖ.

Background and Aim: The aim is first to express the differences and identify three models, namely, Gene Expression Programming (GEP), Neural-Fuzzy Network (ANFIS), and Bayesian Network (BN), and compare them with each other. Furthermore, the research's central question is whether the superior simulation meta-modal in this study can be a suitable alternative to conceptual models in the conditions of lack of data and information. Methods: The data used for this study are the daily rainfall and flow data of the Gamasiab Nahavand River in 10 years from 2002 to 2012. For the prediction or simulation stage, the data of the blue year 2012-2011 have been used. Results: In the training phase and according to the coefficient of explanation and the square root of the mean squares error and the AIC criterion, it is observed that in all three models, both in the training phase and in the test phase, we see a minimal difference in the amount of these parameters. Moreover, all three models' results are close to each other with almost a minimal difference, and almost the relative superiority of the GEP model can be seen. Discussion & Conclusion: The results indicate that the simulator meta-model of gene expression has an excellent ability to simulate and predict the river's daily flow, this simulation meta-model can be a suitable alternative to models in the absence of data and information. Be conceptual. Also, the speed of implementation of the gene expression programming model was faster than other models and was able to provide results in a short time.