Prediction of Natural Gas Price with Hybrid Model Based on Elman Neural Networks and Dragonfly Optimization Algorithm

Seçkin Karasu

Article Details

Prediction of Natural Gas Price with Hybrid Model Based on Elman Neural Networks and Dragonfly Optimization Algorithm

Authors:

Seçkin Karasu

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Elektrik Mühendisliği (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.8 Issue 1, pp.102-114

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001331043

pISSN:

2619-8991

Keywords:

Abstract:

 With the increasing in the world population, the use of various fossil and renewable energy resources is increasing. Compared to coal and oil, which are among the fossil energy sources, natural gas has found use at individual and institutional levels due to its features such as lower carbon dioxide emissions, high efficiency, easy access and low storage costs. The price of natural gas is not only economically important but also strategically important. In particular, the prediction of the future value of natural gas prices provides guidance for energy producers, consumers, investors and governments when making strategic decisions. In this study, the one step ahead natural gas close price is predicted by using Elman Neural Networks ENN and Dragonfly Optimization Algorithm DOA approaches. The analysis utilized a dataset spanning from June 01, 2009, to May 31, 2024, encompassing 3986 close prices. ENN, one of the artificial intelligence approaches, is used to predict the next closing price. ENN, which is among the feedback neural networks, has the ability to predict future values by taking into account past data and is especially used in time series forecasting. During the model training phase, the weight and bias values of the ENN are found with the DOA method, which is a heuristic optimization algorithm developed inspired by the hunting and migration behavior of dragon flies. In the evaluations of the model, the generalization capacity of the model is measured more reliably by dividing the data set into training, validation and test sets. Model performance is evaluated using various statistical error criteria and the results are found to be satisfactory. The use of artificial intelligence approaches is critical to increase forecast accuracy in dynamic and complex systems such as energy markets. The combination of ENN and DOA provides a powerful and flexible solution for such problems. This study demonstrates the effectiveness of artificial intelligence methods in predicting natural gas prices and reveals the usability of these approaches in practical applications.

 With the increasing in the world population, the use of various fossil and renewable energy resources is increasing. Compared to coal and oil, which are among the fossil energy sources, natural gas has found use at individual and institutional levels due to its features such as lower carbon dioxide emissions, high efficiency, easy access and low storage costs. The price of natural gas is not only economically important but also strategically important. In particular, the prediction of the future value of natural gas prices provides guidance for energy producers, consumers, investors and governments when making strategic decisions. In this study, the one step ahead natural gas close price is predicted by using Elman Neural Networks ENN and Dragonfly Optimization Algorithm DOA approaches. The analysis utilized a dataset spanning from June 01, 2009, to May 31, 2024, encompassing 3986 close prices. ENN, one of the artificial intelligence approaches, is used to predict the next closing price. ENN, which is among the feedback neural networks, has the ability to predict future values by taking into account past data and is especially used in time series forecasting. During the model training phase, the weight and bias values of the ENN are found with the DOA method, which is a heuristic optimization algorithm developed inspired by the hunting and migration behavior of dragon flies. In the evaluations of the model, the generalization capacity of the model is measured more reliably by dividing the data set into training, validation and test sets. Model performance is evaluated using various statistical error criteria and the results are found to be satisfactory. The use of artificial intelligence approaches is critical to increase forecast accuracy in dynamic and complex systems such as energy markets. The combination of ENN and DOA provides a powerful and flexible solution for such problems. This study demonstrates the effectiveness of artificial intelligence methods in predicting natural gas prices and reveals the usability of these approaches in practical applications.

 With the increasing in the world population, the use of various fossil and renewable energy resources is increasing. Compared to coal and oil, which are among the fossil energy sources, natural gas has found use at individual and institutional levels due to its features such as lower carbon dioxide emissions, high efficiency, easy access and low storage costs. The price of natural gas is not only economically important but also strategically important. In particular, the prediction of the future value of natural gas prices provides guidance for energy producers, consumers, investors and governments when making strategic decisions. In this study, the one step ahead natural gas close price is predicted by using Elman Neural Networks ENN and Dragonfly Optimization Algorithm DOA approaches. The analysis utilized a dataset spanning from June 01, 2009, to May 31, 2024, encompassing 3986 close prices. ENN, one of the artificial intelligence approaches, is used to predict the next closing price. ENN, which is among the feedback neural networks, has the ability to predict future values by taking into account past data and is especially used in time series forecasting. During the model training phase, the weight and bias values of the ENN are found with the DOA method, which is a heuristic optimization algorithm developed inspired by the hunting and migration behavior of dragon flies. In the evaluations of the model, the generalization capacity of the model is measured more reliably by dividing the data set into training, validation and test sets. Model performance is evaluated using various statistical error criteria and the results are found to be satisfactory. The use of artificial intelligence approaches is critical to increase forecast accuracy in dynamic and complex systems such as energy markets. The combination of ENN and DOA provides a powerful and flexible solution for such problems. This study demonstrates the effectiveness of artificial intelligence methods in predicting natural gas prices and reveals the usability of these approaches in practical applications.

 Dünya nüfusunun artışı ile çeşitli fosil ve yenilenebilir enerji kaynaklarının kullanımı giderek artmaktadır. Doğal gaz, fosil enerji kaynakları arasında yer alan kömür ve petrolle karşılaştırıldığında, daha düşük karbondioksit emisyonu, yüksek verimlilik, kolay erişim ve düşük depolama maliyeti gibi özellikleri nedeniyle bireysel ve kurumsal düzeyde kullanım alanı bulmuştur. Doğal gaz fiyatı ekonomik açıdan önemli olduğu kadar stratejik öneme de sahiptir. Özellikle doğal gaz fiyatının gelecekte alacağı değerin tahmini, enerji üreticilerine ve tüketicilerine, yatırımcılara ve hükümetlere stratejik kararlar alırken yol gösterici olmaktadır. Bu çalışmada, Elman Sinir Ağları ENN ve Yusufçuk Optimizasyon Algoritması DOA yaklaşımları kullanılarak bir adım sonraki doğal gaz kapanış fiyatının tahmini yapılmıştır. Çalışma 01,06,2009-31,05,2024 tarihleri arasında 3986 adet kapanış fiyatı içeren veri seti kullanılarak yapılmıştır. Bir adım sonraki kapanış fiyatının tahmini için yapay zekâ yaklaşımlarından ENN yöntemi kullanılmıştır. Geri beslemeli sinir ağları arasında yer alan ENN, geçmiş verileri dikkate alarak gelecekteki değerleri tahmin etme yeteneğine sahiptir ve özellikle zaman serisi tahmininde kullanılmaktadır. Model eğitim aşamasında yusufçukların avlanma ve göç etme davranışlarından ilham alınarak geliştirilmiş bir sezgisel optimizasyon algoritması olan DOA yöntemiyle ENNnin ağırlık ve bias değerleri bulunmuştur. Modelin değerlendirilme aşamasında veri setinin eğitim, doğrulama ve test setlerine bölünmesiyle modelin genelleme kapasitesi daha güvenilir bir şekilde ölçülmektedir. Model başarımı, çeşitli istatistiksel hata kriterleri kullanılarak değerlendirilmiş ve elde edilen sonuçlar tatminkâr bulunmuştur. Yapay zekâ yaklaşımlarının kullanımı, enerji piyasaları gibi dinamik ve karmaşık sistemlerde tahmin doğruluğunu artırmak için kritik önem taşımaktadır. ENN ve DOAnın birleşimi, bu tür problemler için güçlü ve esnek bir çözüm sunmaktadır. Bu çalışma, doğal gaz fiyatlarının tahmininde yapay zekâ yöntemlerinin etkinliğini göstermekte ve bu yaklaşımların pratik uygulamalarda kullanılabilirliğini ortaya koymaktadır.

Article Details

Prediction of Natural Gas Price with Hybrid Model Based on Elman Neural Networks and Dragonfly Optimization Algorithm

Authors:

Seçkin Karasu

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Elektrik Mühendisliği (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.8 Issue 1, pp.102-114

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001331043

pISSN:

2619-8991

Natural gas prediction

Elman neural networks

Dragonfly optimization algorithm

Time series

ForecastingDoğal gaz tahmini

Elman sinir ağları

Yusufçuk optimizasyon algoritması

Zaman serisi

Tahmin

Issues

Article Details

Authors: Seçkin Karasu

Publisher: Karyay Karadeniz Yayımcılık

Main Topics: Elektrik Mühendisliği (Diğer)

Journal: Black Sea Journal of Engineering and Science Vol.8 Issue 1, pp.102-114

Document Type: Article

Publication Date: 1/2025

Document Language: Turkish

Document ID (Org No.): 2001331043

pISSN: 2619-8991

Issues

Authors:

Seçkin Karasu

Publisher:

Karyay Karadeniz Yayımcılık

Main Topics:

Elektrik Mühendisliği (Diğer)

Journal:

Black Sea Journal of Engineering and Science Vol.8 Issue 1, pp.102-114

Document Type:

Article

Publication Date:

1/2025

Document Language:

Turkish

Document ID (Org No.):

2001331043

pISSN:

2619-8991