Publication: Henry Hub doğal gaz spot fiyat tahminlemesinde makine öğrenmesi ve istatiksel modellerin karşılaştırmalı analizi
Date
Authors
Authors
Akbıyık, Feyzanur Soyal
Advisor
Aksoy, Aslı
Language
Type
Publisher:
Bursa Uludağ Üniversitesi
Journal Title
Journal ISSN
Volume Title
Abstract
Doğal gaz, temiz yanan ve son derece verimli bir enerji kaynağı olduğu için sera gazı emisyonlarını azaltmada bugün mevcut olan en önemli araçlardan biridir. Jeopolitik riskler genellikle öngörülemez olsa da, doğal gaz üreten ülkelerdeki bölgesel çatışmalar veya politika değişiklikleri önemli fiyat dalgalanmalarına neden olmaktadır. Enerji piyasası katılımcıları ve politika yapıcılar doğru doğal gaz fiyat tahminlerine ihtiyaç duymaktadırlar. Enerji şirketleri, doğal gaz fiyat tahminlerine dayalı tedarik stratejileri ve fiyatlandırma politikaları oluşturarak rekabet güçlerini artırıp, son kullanıcıyı etkileyen doğal gaz satış fiyatı politikalarını belirleyebilmektedirler. Doğal gaz fiyatlandırma dinamikleri, üretim maliyetleri, altyapı geliştirme ve genel piyasa istikrarı üzerinde önemli yankıları olan ekonomik stratejileri ve enerji politikası çerçevelerini belirlemek için çok önemlidir. Bu çalışmada, Kuzey Amerika'daki en etkili doğal gaz ticaret merkezi olan Henry Hub spot doğal gaz fiyatlarının yapay sinir ağları, otoregresif entegre hareketli ortalama (ARIMA) ve TimeGPT yöntemleri aracılığıyla tahmin edilmesi hedeflenmiştir. Çalışmada elde edilen bulgulara göre tüm yöntemlere ilişkin en iyi sonuçlar MAE, MSE ve RMSE kriterlerine göre karşılaştırılmıştır. Çalışma kapsamında yapay sinir ağı modellerinin performans metriklerine göre daha iyi sonuçlar verdiği tespit edilmiştir. Bu çalışma, doğal gaz ihtiyacının büyük ölçüde ithalat yoluyla karşılandığı ülkemizde doğal gaz arz kaynaklarının planlanmasına önemli katkı sunacaktır. Doğru planlama stratejisiyle birlikte ülkemizin doğal gaz ithalat maliyeti azaltılabilecek ve son kullanıcıya yapılan doğal gaz satış fiyatlandırma politikası belirlenebilecektir.
Natural gas is one of the most important tools available today to reduce greenhouse gas emissions because it is a clean-burning and highly efficient energy source. While geopolitical risks are often characterized by unpredictable, regional conflicts or policy changes in natural gas producing countries cause significant price fluctuations. This underscores the necessity for energy market participants and policymakers to possess precise natural gas price forecasts. Energy companies can increase their competitiveness by creating supply strategies and pricing policies based on natural gas price forecasts and determine natural gas sales price policies that affect end users. Natural gas pricing dynamics are crucial for determining economic strategies and energy policy frameworks, which have significant repercussions on production costs, infrastructure development and overall market stability. This study aims to estimate spot natural gas prices in Henry Hub, the most effective natural gas trading center in North America, using artificial neural networks, autoregressive integrated moving average (ARIMA) and TimeGPT methods. According to the findings obtained in the study, the best results of all methods were compared according to MAE, MSE and RMSE criteria, and it was determined that artificial neural network models gave better results according to performance metrics. The findings of this study will be of significant value in the planning of natural gas supply sources in our country, where natural gas demand is largely met by imports. With the implementation of an effective planning strategy, it is anticipated that our country's natural gas import costs can be reduced, and the natural gas sales pricing policy to the end user can be determined.
Natural gas is one of the most important tools available today to reduce greenhouse gas emissions because it is a clean-burning and highly efficient energy source. While geopolitical risks are often characterized by unpredictable, regional conflicts or policy changes in natural gas producing countries cause significant price fluctuations. This underscores the necessity for energy market participants and policymakers to possess precise natural gas price forecasts. Energy companies can increase their competitiveness by creating supply strategies and pricing policies based on natural gas price forecasts and determine natural gas sales price policies that affect end users. Natural gas pricing dynamics are crucial for determining economic strategies and energy policy frameworks, which have significant repercussions on production costs, infrastructure development and overall market stability. This study aims to estimate spot natural gas prices in Henry Hub, the most effective natural gas trading center in North America, using artificial neural networks, autoregressive integrated moving average (ARIMA) and TimeGPT methods. According to the findings obtained in the study, the best results of all methods were compared according to MAE, MSE and RMSE criteria, and it was determined that artificial neural network models gave better results according to performance metrics. The findings of this study will be of significant value in the planning of natural gas supply sources in our country, where natural gas demand is largely met by imports. With the implementation of an effective planning strategy, it is anticipated that our country's natural gas import costs can be reduced, and the natural gas sales pricing policy to the end user can be determined.
Description
Source:
Keywords:
Keywords
Henry Hub, Artificial neural networks, ARIMA, Price estimation, Fiyat tahminleme, Yapay sinir ağları