پیش‌بینی نوسانات ماهانه سطح آب دریاچه ارومیه با استفاده از مدل هوش مصنوعی مرکب نظارت‌شده

نویسندگان

1 دانشجوی دکتری هیدروژئولوژی، دانشکده علوم طبیعی، دانشگاه تبریز

2 استاد گروه علوم زمین، دانشکده علوم طبیعی، دانشگاه تبریز

3 استادیار دانشکده زمین‌شناسی، پردیس علوم، دانشگاه تهران

چکیده

در سال­های اخیر کاهش سطح آب دریاچه ارومیه سبب ایجاد بحران آبی و زیست­محیطی در منطقه گردیده است. بنابراین ضروری است که مدیریت و برنامه­ریزی صحیح و قابل­اعتماد در این زمینه صورت گیرد که لازمه آن مدل­سازی سطح آب دریاچه برای آینده می­باشد. در این تحقیق از روش­های شبکه عصبی مصنوعی (ANN)، سیستم استنتاج فازی- عصبی تطبیقی (ANFIS) و ماشین بردار پشتیبان (SVM) برای پیش­بینی سطح آب یک [h (t+1)]، دو [h (t+2)] و سه  [h (t+3)] ماه آینده دریاچه ارومیه استفاده گردید و در نهایت از یک مدل هوش مصنوعی مرکب نظارت شده (SCMAI) برای رسیدن به یک عملکرد بهتر از مدل­های منفرد به­کار گرفته شده، استفاده شد. برای مدل­سازی، اطلاعات سطح آب ماه جاری  [h (t)]و یازده ماه گذشته [h (t-1),…, h (t-11)] به­عنوان ورودی و سطح آب یک، دو و سه ماه آینده به­عنوان خروجی مدل­ها در نظر گرفته شدند. داده­ها به دو دسته داده­های آموزش/ اعتبارسنجی (90 درصد کل داده­ها) و داده­های آزمایش (10 درصد کل داده­ها) تقسیم­بندی گردیدند و پس از مدل­سازی، عملکرد مدل­ها بر اساس پارامترهای ضریب تعیین (R2)، جذر میانگین مربعات خطا (RMSE) و میانگین مطلق خطا (MAE) ارزیابی شدند. نتایج نشان دادند که مدل­های بردار پشتیبان عملکرد بهتری نسبت به مدل­های شبکه عصبی مصنوعی و عصبی- فازی دارند. مدل هوش مصنوعی مرکب نظارت شده به­منظور ترکیب نتایج مدل­های مختلف به­کار گرفته شد و نشان داد که مدل­های هوش مصنوعی مرکب نظارت شده قادرند کارایی مدل­های منفرد هوش مصنوعی را بهبود بخشند. نتایج معیارهای عملکرد مدل­ هوش مصنوعی مرکب نظارت شده بیان می­کند که مدل­سازی یک ماه آینده سطح آب با مقادیر R2، RMSE و MAE به­ترتیب برابر با 9896/0، 0547/0 متر و 0421/0 متر در مقایسه با مدل بردار پشتیبان عملکرد بهتری دارد که این عملکرد برای پیش­بینی­های دو و سه ماه آینده سطح آب دریاچه نیز صادق می­باشد.

کلیدواژه‌ها


عنوان مقاله [English]

Forecasting Monthly Water Level Fluctuations of Lake Urmia Using Supervised Committee Machine Artificial Intelligence Model

نویسندگان [English]

  • R Barzegar 1
  • A Asghari Moghaddam 2
  • E Fijani 3
1 Ph.D student of Hydrogeology, Dept. of Earth Sciences, Faculty of Natural Sciences, Univ. of Tabriz, Iran
2 Prof., Dept. of Earth Sciences, Faculty of Natural Sciences. Univ. of Tabriz, Iran
3 Assist. Prof., School of Geology, University College of Science, University of Tehran, Tehran, Iran
چکیده [English]

In recent years, declining the water level of Lake Urmia has caused water and environmental crisis in the area. Therefore, it is urgent to carry out an accurate and reliable management and planning which requires modeling the lake's water level for the future. In this research, the artificial neural network (ANN), adaptive neuro fuzzy inference system (ANFIS) and support vector machine (SVM) models were used to forecast the Lake Urmia water level fluctuations for one, two and three months ahead forecast horizons and finally, a supervised committee machine artificial intelligence (SCMAI) model was used to obtain a better performance than the used individual models. To develop the models, the current month [h (t)] and eleven months water level lags [h (t-1),…, h (t-11)] were introduced as input variables to forecast one, two and three steps ahead water levels. The datasets were divided into two subsets of training/validation (90%) and testing (10%). The performances of the models were evaluated based on the coefficient of determination (R2), the root mean square error (RMSE) and the mean absolute error (MAE). The results showed that the SVM models had better performance than the ANN and ANFIS models. The SCMAI model was applied to combine the used models’ outputs and illustrated that the SCMAI models are able to improve the performance of the individual artificial intelligence models. The results of the performance criteria for SCMAI model indicated that the one month step ahead water level modeling with R2, RMSE and MAE equal to 0.9896, 0.0547 m and 0.0421 m, respectively outperformed in comparison with SVM model which this performance is reliable for the two and three months step ahead lake's water level.

کلیدواژه‌ها [English]

  • Forecasting
  • Supervised committee machine artificial intelligence
  • Support vector machine
  • Lake Urmia
  • Water level
Barzegar R, Adamowski J, and Asghari Moghaddam A, 2016a. Application of wavelet-artificial intelligence hybrid models for water quality prediction: a case study in Aji-Chay River, Iran. Stochastic Environmental Research and Risk Assessment 30(7): 1797-1819.
Barzegar R, and Asghari Moghaddam A, 2016. Combining the advantages of neural networks using the concept of committee machine in the groundwater salinity prediction. Modeling Earth Systems and Environment 2: 26.
Barzegar R, Asghari Moghaddam A, Adamowski J, and Fijani E, 2016b. Comparison of machine learning models for predicting fluoride contamination in groundwater. Stochastic Environmental Research and Risk Assessment doi: 10.1007/s00477-016-1338-z.
Barzegar R, Asghari Moghaddam A, and Baghban H, 2016c. A supervised committee machine artificial intelligent for improving DRASTIC method to assess groundwater contamination risk: a case study from Tabriz plain aquifer, Iran. Stochastic Environmental Research and Risk Assessment 30(3): 883–899
Barzegar R, Sattarpour M, Nikudel MR, and Asghari Moghaddam A, 2016d. Comparative evaluation of artificial intelligence models for prediction of uniaxial compressive strength of travertine rocks, case study: Azarshahr area, NW Iran. Modeling Earth Systems and Environment 2:76.
Ebtehaj I, Bonakdari H, Shamshirband S, and Mohammadi K, 2016. A combined support vector machine-wavelet transform model for prediction of sediment transport in sewer. Flow Measurement and Instrumentation 47: 19–27.
Kadkhodaie-Ilkhchi A, Rezaee, M R, Rahimpour-Bonab H, and Chehrazi A, 2009. Petrophysical data prediction from seismic attributes using committee inference system. Computers & Geosciences 35: 2314–2330.
Kakahaji H, Dehghan Banadaki H,  Kakahaji A,  and Kakahaji, A 2013. Prediction of Urmia Lake water-level fluctuations by using analytical, linear statistic and intelligent methods. Water Resources Management 27(13): 4469–4492.
Karimi S, Shiri J, Kisi O, Makarynskyy O, 2012. Forecasting water level fluctuations of Urmieh Lake using gene expression programming and adaptive neuro-fuzzy inference system. International Journal of Ocean and Climate Systems 3(2): 109–125.
Khatami S, 2013. Nonlinear chaotic and trend analyses of water level at Urmia Lake, Iran. M.Sc. Thesis report: TVVR 13/5012, ISSN: 1101–9824, Lund University, Lund, Sweden.
Kisi O, Shiri, J, Karimi S, Shamshirband S, Motamedi S, Petković D, and Hashim R, 2015. A survey of water level fluctuation predicting in Urmia Lake using support vector machine with firefly algorithm. Applied Mathematics and Computation 270: 731–743.
Naftaly U, Intrator N, and Horn D, 1997. Optimal ensemble averaging of neural networks. Computation in Neural Systems 8: 283–296.
Salas JD, Deulleur JW, Yevjevich V, and Lane WL, 1980. Applied Modelling of Hydrologic Time Series. Water Resources Publ. Littleton, Colorado 484p.
Salat R, Osowski S, 2004. Accurate fault location in the power transmission line using support vector machine approach. Power Systems IEEE Transactions on 19: 879–886.
Sanikhani H, Kisi O, Kiafar H, and Ghavidel SZZ, 2015. Comparison of different data-driven approaches for modeling lake level fluctuations: The case of Manyas and Tuz Lakes (Turkey). Water Resources Management 29(5): 1557–1574.
Şen Z, Kadioğlu M, and Batur, E, 2000. Stochastic modelling of the Van Lake monthly level fluctuations in Turkey. Theoretical and Applied Climatology 65(1–2): 99–110.
Shamshirband S, Mohammadi K, Khorasanizadeh H, Yee PL, Lee M, Petkovic D, and Zalnezhad E, 2016. Estimating the diffuse solar radiation using a coupled support vector machine–wavelet transform model. Renewable & Sustainable Energy Reviews 56: 428–435.
Shiri J, Shamshirband S, Kisi O, Karimi S, Bateni SM, Hosseini Nejad SH, and Hashemi A, 2016. Prediction of Water-Level in the Urmia Lake Using the Extreme Learning Machine Approach. Water Resources Management 30(14): 5217–5229.
Talebizadeh M, and Moridnejad A, 2011. Uncertainty analysis for the forecast of lake level fluctuations using ensembles of ANN and ANFIS models. Expert Systems with Applications 38(4): 4126–4135.
Vaheddoost B, Aksoy H, and Abghari H, 2016. Prediction of water level using monthly lagged data in Lake Urmia, Iran. Water Resources Management 30 (13): 4951–4967.
Vapnik VN, 1995. The Nature of Statistical Learning Theory. Springer, New York, 314 p.
Zaji AH, Bonakdari H, Khodashenas SR, and Shamshirband S, 2016. Firefly optimization algorithm effect on support vector regression prediction improvement of a modified labyrinth side weir’s discharge coefficient. Applied Mathematics and Computation 274: 14–19