نوع مقاله : مقاله پژوهشی

نویسندگان

• سجاد هاشمی ¹
• سعید صمدیان فرد ²
• علی اشرف صدرالدینی ³

¹ دانشجوی دکتری، گروه مهندسی آب، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران

² گروه مهندسی آب، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران

³ استاد، گروه مهندسی آب، دانشکده کشاورزی، دانشگاه تبریز، تبریز، ایران

چکیده

انرژی خورشیدی مهمترین منبع انرژی‌های تجدیدپذیر و به عبارتی منبع اصلی انرژی‌های موجود در زمین است. با در نظر گرفتن غیرقابل تجدیدپذیر بودن انرژی‌های فسیلی و آثار مخرب استفاده از سوخت‌های فسیلی بر محیط زیست، استفاده از انرژی‌های پاک و تجدیدپذیری مانند انرژی خورشیدی می‌تواند بهترین و مناسبترین جایگزین برای تأمین انرژی موردنیاز باشد. در این راستا، در تحقیق حاضر از داده‌های هواشناسی 3 ایستگاه هواشناسی استان اردبیل شامل مشگین شهر، گرمی و نیر در بازه زمانی 2 ساله (2018 -2017) در مقیاس روزانه استفاده شده و با به‌کارگیری روش‌های جنگل تصادفی و جنگل تصادفی-الگوریتم ژنتیک شدت تابش خورشیدی روزانه در هر یک از ایستگاه‌های مذکور برآورد شد. نتایج به‌ دست‌آمده با استفاده از پارامترهای آماری با یکدیگر مقایسه شده و مدل‌های برتر انتخاب گردید. با مقایسه کلی نتایج، مدل‌های ایستگاه‌های نیر، مشگین شهر و گرمی به ترتیب از بیشترین به کمترین دقت مدل‌سازی رتبه‌بندی شدند؛ به‌طوری‌که مدل GA-RF-V در ایستگاه نیر با دارا بودن جذر میانگین مربعات خطای 346/0 (مگاژول بر متر مربع در روز) و راندمان کلینگ-گاپتا 687/0 با کمترین خطا به‌عنوان برترین مدل در این مطالعه معرفی شد. همچنین نتایج به‌دست‌آمده نشان داد که الگوریتم ژنتیک به افزایش دقت همه مدل‌های مورد استفاده کمک شایانی کرده است.

کلیدواژه‌ها

• اردبیل
• انرژی خورشیدی
• بهینه‌سازی
• پارامترهای آماری
• مدل‌های هوشمند

موضوعات

• هیدرولوژی

عنوان مقاله [English]

Evaluation of Random Forest-Genetic Algorithm Hybrid Model in Estimating Daily Solar Radiation

نویسندگان [English]

• Sajjad Hashemi ¹
• Saeed Samadianfard ²
• Ali Ashraf Sadraddini ³

¹ Ph.D. Candidate, Department of Water Engineering, Faculty of Agriculture, Tabriz University, Tabriz, Iran

³ Professor, Department of Water Engineering, Faculty of Agriculture, Tabriz University, Tabriz, Iran

چکیده [English]

Extended Abstract

Introduction

The sources of fossil energy are running out and the use of this type of energy has disadvantages such as greenhouse gas emissions, air pollution and global warming. So, there is no doubt that the replacement and use of clean and renewable energy such as solar energy can be the best and most appropriate way for production, growth and economic development of developing countries such as Iran. In addition, solar energy and solar radiation parameter is one of the key factors in the fields of agriculture, hydrology and meteorology. Due to the fact that there are problems in using physical methods and meteorological data (requires complex calculations, high costs, etc.) in predicting solar radiation, statistical methods and intelligence learning models can be used as a complementary solution that requires much less cost and time. In recent years, researchers have used these methods to model solar radiation.

Methods

In the present study the meteorological data of average, minimum and maximum temperature, relative humidity and wind speed were utilized to estimate daily solar radiation in three meteorological stations of Ardabil province including Meshgin Shahr, Germi and Nir from in a period of 2 years (2017-2018) on a daily scale so that the mentioned parameters are used as input data in eight different combinations. The methods used in this research include random forest and random forest optimized by genetic algorithm. The results obtained from these models were compared using statistical indices such as correlation coefficient (CC), root mean square error (RMSE) and Kling-Gupta efficiency (KGE) and the best models were selected.

Results

The results showed that in the random forest method and in Germi station, RF-VI model with CC of 0.77, RMSE of 0.522 MJ/m2 d and KGE of 0.654, in Meshgin Shahr station, RF-VI model with CC of 0.813, RMSE of 0.417 MJ/m2 d and KGE of 0.529 and in Nir station RF-V model with CC of 0.778, RMSE of 0.363 MJ/m2 d and KGE of 0.693 had the best performance. Overall, by comparing the results between the three studied stations, the models of Nir station performed more accurately in the RF method than the other two stations, and the results of Meshgin Shahr and Germi stations were in the next ranks, respectively. In GA-RF models, Nir, Meshgin Shahr and Germi stations were ranked first to third, respectively. It is noteworthy from the obtained results that in the studied stations, the genetic algorithm has improved the performance of all the utilized models, so that all GA-RF models have more accurately estimated the intensity of solar radiation by reducing the error.

Coclusion

The overall conclusion showed that: 1) In Germi and Meshgin Shahr stations, the sixth combination models with the input parameters of minimum and maximum temperature, relative humidity and wind speed provided the most desirable results. 2) In Nir station, the fifth combination models with the input parameters of minimum and maximum temperature and relative humidity had the highest accuracy and the lowest error. 3) By comparing the results between mentioned stations, in both RF and GA-RF methods, Nir, Meshgin Shahr and Germi stations were ranked from more to less accuracy, respectively. 4) By examining RF models with GA-RF, it was concluded that the genetic algorithm improved the performance of the models and had a positive effect on all models.

کلیدواژه‌ها [English]

• Ardabil
• Intelligent Models
• Optimization
• Solar Energy
• Statistical Parameters