عنوان

بررسی کارآیی مدل‌های هیبریدی داده‌کاوی در پیش‌بینی فرآیندهای هیدرولوژیکی

Number

‭۱۷۷۲۴پ‬

.Language of Text, Soundtrack etc

per

Title Proper

بررسی کارآیی مدل‌های هیبریدی داده‌کاوی در پیش‌بینی فرآیندهای هیدرولوژیکی

First Statement of Responsibility

/علی دوانلو تاج بخش

Name of Publisher, Distributor, etc.

: فنی و مهندسی عمران

Date of Publication, Distribution, etc.

، ‮‭۱۳۹۶‬

Name of Manufacturer

، افشاری

Text of Note

چاپی

Text of Note

Due to the importance of hydrological models in the climate study and its impact on the water management, the hybrid Wavelet-M5 model was introduced, applied on Aji Chai and Murrumbidgee catchments in both daily and monthly scales, and compared its results to the other models in the current study. At first, the rainfall and runoff time series were decomposed using the wavelet transform to several sub-time series to overcome the non-stationary. Then, the obtained sub-time series were applied to M5 model tree as input data. At last, the related regression of each cluster was presented after classifying the input data. As it was expected, the performance of Wavelet-M5 model was better 3. on the daily scale, 50. on the monthly scale for Aji Chai catchment, 37. and 3.21times on the daily and monthly scales for the Murrumbidgee catchment, respectively rather than M5 tree model because of the appropriate preprocessing and eliminating the available trend in the main time series. It was also investigated the performance of Wavelet-M5 model in the case of forecasting multi-step ahead in comparison with other models. The results showed the acceptable efficiency of Wavelet-M5 model and improved 9. and 1.71 times on Aji Chai watershed in the daily and monthly scales, respectively and 1.73 and 14.4 times in the Murrumbidgee basin on the daily and monthly scales rather than M5 tree model, respectively. In addition, the ability of the proposed hybrid model was analyzed in the case of different data division strategies (60-40. and 50-50. ), as Wavelet-M5 encountered the small number of data for training. The results indicated the insignificant percentage of changes in the diverse data division strategies (-1. and 0 on the daily scale and no change in the monthly scale of Aji Chai watershed, no change on the daily scale, 0 and -1

Dissertation or thesis details and type of degree

کارشناسی ارشد

Discipline of degree

عمران گرایش مهندسی و مدیریت منابع آب

Date of degree

‮‭۱۳۹۶/۰۶/۲۰‬

Body granting the degree

تبریز

Text of Note

با توجه به اهمیت مدل‌صهای هیدرولوژیکی در مطالعه اقلیم و تاثیر به سزای آن در مدیریت و مهندسی آب، در این پایان‌صنامه تلاش شد تا ضمن معرفی مدل جدید‮‭M۵- Wavelet‬، عملکرد آن بر روی فرآیند بارش-رواناب دو حوضه آبریز آجی‌صچای و ‮‭Murrumbidgee‬ مطالعه شود .در ابتدا سری-های زمانی بارش و رواناب با استفاده از تبدیل موجک به چندین زیرسری تجزیه گشته تا بر ناایستایی آن غلبه گردد .سپس زیرسری‌صهای حاصله به عنوان ورودی به مدل ‮‭M۵‬ درختی اعمال شده تا پس از طبقه‌صبندی داده‌صها، رگرسیون مربوط به هر خوشه ارائه شده و نتایج با سایر مدل‌صها مقایسه گردد .مطابق انتظار با اعمال پیش‌صپردازش مناسب و حذف روند موجود در سری‌صهای زمانی اصلی، مدل‮‭M۵ - Wavelet‬به میزان ‮‭۳‬ در مقیاس روزانه، ‮‭۵۰‬ در مقیاس ماهانه برای حوضه آجی‌صچای، و به میزان ‮‭۳۷‬ و ‮‭۲۱/۳‬ برابر به ترتیب در مقیاس‌صهای روزانه و ماهانه برای حوضه ‮‭Murrumbidgee‬ نسبت به مدل ‮‭M۵‬ درختی بهبود یافته است .همچنین عملکرد مدل‮‭M۵ - Wavelet‬در پیش‌صبینی افق چند گام جلوتر با سایر مدل‌صها بررسی گردید و نتایج جاکی از عملکرد مناسب آن در حوضه آجی‌صچای به میزان ‮‭۹‬ و ‮‭۷۱/۱‬ برابر به ترتیب در مقیاس‌صهای روزانه و ماهانه، ‮‭۷۳/۱‬ و ‮‭۴/۱۴‬ برابر به ترتیب در مقیاس‌صهای روزانه و ماهانه در حوضه ‮‭Murrumbidgee‬ نسبت به مدل ‮‭M۵‬ درختی بود .علاوه بر آن توانایی مدل هیبریدی پیشنهادی در استراتژی‌صهای گوناگون تقسیم داده‮‭۶۰ - ( ۴۰‬ و‮‭۵۰) - ۵۰‬ به هنگام مواجه با تعداد کم داده‌صها جهت آموزش سنجیده شد و نتایج حاصل بیانگر تغییر درصد ناچیزی در این تقسیم‌صبندی‮‭( ۱‬ - و ‮‭۰‬ در مقیاس روزانه و عدم تغییر در مقیاس ماهانه حوضه آجی‌صچای، عدم تغییر در مقیاس روزانه، ‮‭۰‬ و‮‭۱‬ - در مقیاس ماهانه حوضه ‮‭Murrumbidgee)‬ در استراتژی‌صهای گوناگون بود .

Text of Note

Due to the importance of hydrological models in the climate study and its impact on the water management, the hybrid Wavelet-M5 model was introduced, applied on Aji Chai and Murrumbidgee catchments in both daily and monthly scales, and compared its results to the other models in the current study. At first, the rainfall and runoff time series were decomposed using the wavelet transform to several sub-time series to overcome the non-stationary. Then, the obtained sub-time series were applied to M5 model tree as input data. At last, the related regression of each cluster was presented after classifying the input data. As it was expected, the performance of Wavelet-M5 model was better 3 on the daily scale, 50 on the monthly scale for Aji Chai catchment, 37 and 3.21times on the daily and monthly scales for the Murrumbidgee catchment, respectively rather than M5 tree model because of the appropriate preprocessing and eliminating the available trend in the main time series. It was also investigated the performance of Wavelet-M5 model in the case of forecasting multi-step ahead in comparison with other models. The results showed the acceptable efficiency of Wavelet-M5 model and improved 9 and 1.71 times on Aji Chai watershed in the daily and monthly scales, respectively and 1.73 and 14.4 times in the Murrumbidgee basin on the daily and monthly scales rather than M5 tree model, respectively. In addition, the ability of the proposed hybrid model was analyzed in the case of different data division strategies (60-40 and 50-50 ), as Wavelet-M5 encountered the small number of data for training. The results indicated the insignificant percentage of changes in the diverse data division strategies (-1 and 0 on the daily scale and no change in the monthly scale of Aji Chai watershed, no change on the daily scale, 0 and -1 on the monthly scale of the Murrumbidgee catchment)

دوانلو تاج بخش، علی

نورانی، وحید، استاد راهنما

حسن زاده، یوسف، استاد مشاور

Public note

سیاه و سفید

نمایه‌سازی قبلی