کاربرد مدل‌های هوشمند عصبی در پیش‌بینی پارامترهای کیفی آب مخازن سدها (مطالعه موردی: سد اکباتان همدان)

نوع مقاله : مقاله پژوهشی

نویسندگان

1 کارشناس ارشد مهندسی عمران،گرایش مهندسی و مدیریت منابع آب

2 دانشجو کارشناس ‌ارشد، مهندسی عمران گرایش مهندسی و مدیریت منابع آب، دانشگاه یاسوج

3 استادیار گروه مهندسی عمران،مهندسی و مدیریت منابع آب، دانشگاه یاسوج

4 ستادیار مهندسی منابع آب، دانشکده کشاورزی، دانشگاه ملایر

10.22034/jess.2022.352332.1827

چکیده

چکیده
بررسی کیفیت آب مخازن سدها به‌منظور تامین تقاضا منابع آب جهت کاربری‌های مختلف مانند: تامین آب شرب شهر و حق‌آبه کشاورزی در مدیریت منایع آب ضروری می‌باشد. هدف از این پژوهش پیش‌بینی پارامترهای کیفی سد کباتان همدان به تفکیک دوره گرم و سرد می‌باشد.در این مطالعه با تکیه بر توانمندی مدل‌های هوشمند عصبی از 4 مدل شبکه عصبی مصنوعی، عصبی-فازی، عصبی- موجک و عصبی-فازی- موجک برای پیش‌بینی پارامترهای کیفیت آب سد اکباتان همدان استفاده شده است. بدین منظور از داده های پارامترهای BOD5، DO، pH، دما، کل جامدات و کدورت آب طی سال‌های 1388و 1389 جهت برآورد فسفات، نیترات، کلیفرم مدفوعی و کلیفرم a استفاده شد. جهت بررسی شرایط محیطی بر دقت نتایج، پیش‌بینی‌ها در دو دوره گرم و سرد سال صورت گرفت. براساس نتایج به‌دست آمده، مدل تلفیقی شبکه عصبی با نظریه موجک به‌عنوان ساختار بهینه در برآورد هر چهار پارامتر کیفی در هر دو دوره معرفی شد. در بین پارامترهای مورد بررسی در دوره گرم، کمترین خطای جذور میانگین مربعات خطای نرمال (NRMSE) و بیشترین ضریب همبستگی به‌ترتیب 905/0 و 999/0و در دوره سرد کمترین مقدار NRMSE و بیشتربن ضریب همبستگی به‌ترتیب 75/2 و 905/0 برای پارامتر کیفی نیترات مشاهده گردید. در مجموع کاربرد نظریه موجک منجر به بهبود نتایج پیش‌بینی شده پارامترهای کیفی سد اکباتان گردیده است. همچنین پیش‌بینی پارامترهای کیفی در دوره گرم دقت بیشتری نسبت به دوره سرد داشت. این امر اهمیت کاربرد مدل‌های هوشمند عصبی در برآورد پارامترهای کیفی آب در فصول گرم را نشان می‌دهد.

کلیدواژه‌ها


عنوان مقاله [English]

Application of Intelligent Neural Models in Determining the Water Quality Parameters of Dams Reservoirs (Case study: Ekbatan dam of Hamadan)

نویسندگان [English]

  • reza khalili 1
  • Mohammad Rostami 2
  • Hossein Montaseri 3
  • Mohammad Parvin Nia 3
  • Maryam Bayat Verkeshi 4
1 Master of Civil Engineering, Engineering and Water Resources Management
2 M.Sc. Student of Civil Engineering of Water Resources Management, Engineering Faculty, Yasuj University
3 Assistant Professor of Civil Engineering, Engineering Faculty, Yasuj University
4 3Assistant Professor of Water Resources Engineering, Faculty of Agriculture, Malayer University
چکیده [English]

Abstract
Introduction:
Identifying the quantitative and qualitative problems in water resources monitoring systems is one of the most important steps in formulating the structure of water resources systems management plans and implementing pollution reduction environmental plans. It is considered to use some indirect methods to simulate qualitative parameters in high volume in order to reduce cost, time and high accuracy. In the field of simulating water quality models, many models have been developed that require a lot of input parameters such as hydrological, meteorological data, etc., which require spending time and money to access them. The increasing expansion of computers and the use of artificial intelligence and the use of artificial neural network methods have been widely used in the estimation of qualitative parameters.

Methodology:
In the present study, based on the ability of intelligent neural models of four models such as artificial neural network, neuro-fuzzy, neural-wavelet, and neuro-fuzzy-wavelet were used to predict the water quality parameters of the Hamadan Ekbatan dam. For this purpose, BOD5, DO, pH, temperature, total solids and water turbidity were measured during 1388and 1389 to estimate phosphate, nitrate, fecal coliform and total chlorophyll a. in order to evaluation of the environmental conditions on the accuracy of the results , predictions were made in the last two warm and cold periods of the year .

Conclusion:
Based on the results, the combined model of neural network with wavelet theory was introduced as the optimal structure for estimating all four qualitative parameters in both periods. Among the parameters studied during the warm period, the lowest normal root mean square error (NRMSE) and the highest correlation coefficient were 0.990 and 0.999, Furthermore, in the cold period, the least amount of NRMSE and the most correlation coefficient was 2.75 and 0,905 have seen for the nitrate quality parameter

In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than in cold period.
In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than in cold period.In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than in cold period.In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than in cold period.In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than in cold period.In this research, 4 smart models were used to predict the quality parameters of phosphate, nitrate, faecal coliform and coliform a of Ekbatan Dam in Hamedan. The error measurement criteria of accuracy and agreement with the measured value were examined. In general, the combined method of neural network and wavelet theory was introduced as an optimal structure with a high correlation coefficient and a lower error rate than other methods for both cold and warm periods. The predicted value using this method has the most agreement with the measured value. Using smart models reduces cost and time and has high accuracy. Also, the accuracy of forecasting of qualitative parameters is more in hot period than

Keywords:
"Neural Intelligence", "Wave Theory", "Cold and Hot Period", "Ekbatan Dam"

کلیدواژه‌ها [English]

  • "Neural Intelligence"
  • "Wave Theory"
  • "Cold and Hot Period"
  • "Ekbatan Dam"