Modelling of the factors affecting the working speed of turbo compressors using machine learning methods

Sema Atasever; Uğur Sorgucu

Article Details

Modelling of the factors affecting the working speed of turbo compressors using machine learning methods

Authors:

Sema Atasever | Uğur Sorgucu

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Bilgisayar Yazılımı

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.34-40

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309711

pISSN:

2564-6605

Keywords:

Abstract:

 Due to the increasing demand for natural gas in industries and households caused by population growth, various projects have been planned both in our country and globally to store natural gas underground during periods of low demand and utilize it from storage during high-demand periods. It is aimed to prevent seasonal fluctuations in natural gas demand through these planned projects. Turbo compressors are used for underground natural gas storage in salt caves. Turbo compressors are significant machines capable of raising natural gas to high pressures. Within the scope of this study, machine learning methods were used to model the factors affecting the operating speed of the mentioned turbo compressors for predicting the compressor speed, including flow, line pressure, and wellhead pressure. According to the experimental results obtained from the study, where 42 machine learning models were tested, the Decision Tree Regressor DT algorithm provided the best results for predicting the compressor speed with an R2 value of 0.99, RMSE of 11.29, and a runtime of 0.01 seconds.

 Due to the increasing demand for natural gas in industries and households caused by population growth, various projects have been planned both in our country and globally to store natural gas underground during periods of low demand and utilize it from storage during high-demand periods. It is aimed to prevent seasonal fluctuations in natural gas demand through these planned projects. Turbo compressors are used for underground natural gas storage in salt caves. Turbo compressors are significant machines capable of raising natural gas to high pressures. Within the scope of this study, machine learning methods were used to model the factors affecting the operating speed of the mentioned turbo compressors for predicting the compressor speed, including flow, line pressure, and wellhead pressure. According to the experimental results obtained from the study, where 42 machine learning models were tested, the Decision Tree Regressor DT algorithm provided the best results for predicting the compressor speed with an R2 value of 0.99, RMSE of 11.29, and a runtime of 0.01 seconds.

 Due to the increasing demand for natural gas in industries and households caused by population growth, various projects have been planned both in our country and globally to store natural gas underground during periods of low demand and utilize it from storage during high-demand periods. It is aimed to prevent seasonal fluctuations in natural gas demand through these planned projects. Turbo compressors are used for underground natural gas storage in salt caves. Turbo compressors are significant machines capable of raising natural gas to high pressures. Within the scope of this study, machine learning methods were used to model the factors affecting the operating speed of the mentioned turbo compressors for predicting the compressor speed, including flow, line pressure, and wellhead pressure. According to the experimental results obtained from the study, where 42 machine learning models were tested, the Decision Tree Regressor DT algorithm provided the best results for predicting the compressor speed with an R2 value of 0.99, RMSE of 11.29, and a runtime of 0.01 seconds.

 Günümüzde nüfus artışı ile sanayi ve konutlara yönelik doğal gaz tüketim talebinin artması sonucunda, talebin az olduğu zamanlarda doğal gazın yeraltında depolanması ve çok olduğu zamanlarda ise depolardan kullanımının sağlanabilmesine yönelik hem ülkemizde hem de dünyada çeşitli projelerin planlaması yapılmıştır. Planlanan bu projeler vasıtasıyla doğalgaz talebindeki mevsimsel dalgalanmaların önlenmesi hedeflenmektedir. Tuz mağaralarında yeraltı doğal gaz depolaması için turbo kompresörler kullanılmaktadır. Turbo kompresörler doğalgazı yüksek basınçlara çıkarabilen önemli makinalardandır. Bu çalışma kapsamında bahsedilen turbo kompresörlerin çalışma hızını etkileyen faktörlerden akış, hat basıncı ve kuyu başı basıncının kompresör devrinin tahmini için makine öğrenmesi yöntemleri ile modellenmesi gerçekleştirilmiştir. 42 makine öğrenmesi modelinin denendiği çalışmadan elde edilen deneysel sonuçlara göre kompresör devrinin tahmininde Decision Tree Regressor DT algoritması 0.99 R2, 11.29 RMSE ve 0.01 sn çalışma zamanı ile en iyi sonucu veren algoritma olmuştur.

Article Details

Modelling of the factors affecting the working speed of turbo compressors using machine learning methods

Authors:

Sema Atasever | Uğur Sorgucu

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Bilgisayar Yazılımı

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.34-40

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309711

pISSN:

2564-6605

Machine Learning

Decision Tree Regressor

Underground natural gas storage

Turbo CompressorsMakine Öğrenmesi

Decision Tree Regressor

Yer altı doğalgaz depolaması

Turbo Kompresörler

Issues

Article Details

Authors: Sema Atasever | Uğur Sorgucu

Publisher: Nigde Omer Halisdemir University

Main Topics: Bilgisayar Yazılımı

Journal: Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.34-40

Document Type: Article

Publication Date: 1/2024

Document Language: Turkish

Document ID (Org No.): 2001309711

pISSN: 2564-6605

Issues

Authors:

Sema Atasever | Uğur Sorgucu

Publisher:

Nigde Omer Halisdemir University

Main Topics:

Bilgisayar Yazılımı

Journal:

Niğde Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi Vol.13 Issue 1, pp.34-40

Document Type:

Article

Publication Date:

1/2024

Document Language:

Turkish

Document ID (Org No.):

2001309711

pISSN:

2564-6605