Estimation of Natural Gas Pipeline Construction Costs Using Multiple Linear Regression and K-Nearest Neighbor Methods

Year 2024, , 1327 - 1337, 15.11.2024

Coşkun Çakmak , Mürsel Erdal

https://doi.org/10.34248/bsengineering.1525230

Abstract

In this study, models were developed using machine learning algorithms for the preliminary estimation of natural gas pipeline (NGP) costs within the borders of Türkiye. For this purpose, data obtained from NGP projects completed in Türkiye between 1997 and 2022 were used. Variables such as pipe diameter, line length, number of line valves, number of take-off valves and number of pigging stations of the projects were determined as independent variables in the cost estimation. Since the data set was quantitatively insufficient and the data quality was at an average level, classical machine learning estimation processes could not be carried out. For this reason, the existing data set was studied using the entire data without dividing it into training and test sections, and it was examined whether the model performed appropriately when positioned in Multiple Linear Regression (MLR) and K-Nearest Neighbor (KNN) algorithms. This study was carried out to provide a preliminary idea about whether classical machine learning estimation processes can be carried out in the future if the data quality and number increase. Similar and average coefficients of determination (R²) were obtained in both different method trials. As a result, in this study, the effectiveness of the MLR and KNN methods was compared to improve the accuracy of preliminary cost estimates in NGP projects and it was evaluated that it will make a significant contribution to the sector. It is anticipated that future studies can increase the accuracy of cost estimates by using larger data sets and different model techniques and can guide the sector stakeholders.

Keywords

Natural gas pipelines, Machine-learning, Cost estimate

Doğal Gaz Boru Hattı İnşaatı Maliyetlerinin Çoklu Doğrusal Regresyon ve K-En Yakın Komşuluk Yöntemleri İle Tahmini

Abstract

Bu çalışmada, Türkiye sınırları içerisinde yapılacak olan doğal gaz boru hattı (DGBH) maliyetlerinin ön tahmini için makine öğrenmesi algoritmaları kullanılarak modeller geliştirilmiştir. Bunun için, 1997-2022 yılları arasında Türkiye'de tamamlanmış DGBH projelerinden elde edilen veriler kullanılmıştır. Projelerin boru çapı, hat uzunluğu, hat vanası sayısı, take-off vana sayısı ve pig istasyonu sayısı gibi değişkenleri, maliyet tahmininde bağımsız değişkenler olarak belirlenmiştir. Veri setinin nicel anlamda yetersiz ve veri kalitesinin ortalama bir seviyede olmasından dolayı, klasik makine öğrenmesi tahmin süreçleri yürütülememiştir. Bu nedenle, mevcut veri seti eğitim ve test bölümlerine ayrılmadan, bütün veri kullanılarak çalışılmış ve Çoklu Doğrusal Regresyon (ÇDR) ile K-En Yakın Komşu (KNN) algoritmalarına konumlandırıldığında modelin uygun bir şekilde performans gösterip göstermediği incelenmiştir. Bu çalışma, ileride veri kalitesinin ve sayısının artması durumunda, klasik makine öğrenmesi tahmin süreçlerinin yürütülüp yürütülemeyeceği konusunda ön fikir vermesi amacıyla gerçekleştirilmiştir. Her iki farklı yöntem denemesinde de benzer ve ortalama düzeyde belirleme katsayıları (R²) elde edilmiştir. Sonuç olarak, bu çalışmada, DGBH projelerinde ön maliyet tahminlerinin hassasiyetini iyileştirmek için ÇDR ve KNN yöntemlerinin etkinliği karşılaştırılmış ve sektöre önemli bir katkı sağlayacağı değerlendirilmiştir. Gelecekte yapılacak çalışmaların daha geniş veri setleri ve farklı model teknikleri kullanarak maliyet tahminlerinin doğruluğunu artırabileceği ve sektör paydaşlarına yol gösterici olabileceği öngörülmektedir.

Keywords

Doğal gaz boru hatları, Makine öğrenmesi, Maliyet tahmini

References

• Adeli H, Wu M. 1998. Regularization Neural Network for construction cost estimation. J Constr Eng Manag, 124(1): 18-24.
• Arage SS, Dharwadkar NV. 2017. Cost estimation of civil construction projects using machine learning paradigm. International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud)(I-SMAC), February 13-15, Tamil Nadu, India, pp: 594-599.
• Bentley JL. 1975. Multidimensional binary search trees used for associative searching. Commun ACM, 18(9): 509-517.
• Birgönül TM, Dikmen İ. 1996. İnşaat projelerinin risk yönetimi. IMO Tek Derg, 97: 1305-1326.
• Buitinck L, Louppe G, Blondel M, Fabien P, Mueller A, Olivier G, Niculae V, Prettenhofer P, Gramfort A, Grobler J, Layton R, VanderPlas J. 2011. Scikit-learn: Machine learning in Python. J Mach Learn Res, 12: 2825-2830.
• Çakmak C, Erdal M. 2022. Preliminary estimation of natural gas pipeline construction costs with regression analysis. 7th International Project and Construction Management Conference, October 10-12, İstanbul, Türkiye, pp: 328-337.
• Erdal H. 2021. Prediction of pipeline projects construction costs utilizing machine learning techniques. International Marmara Sciences Congress (Spring 2021), May 12-14, Kocaeli, Türkiye, pp: 218-223.
• Govan P, Reinschmidt K. 2013. Benchmarking natural gas pipeline projects. Pipelines 2013 Conference, June 5-7, Texas, US, pp: 1532-1542.
• Hubert M, Veeken SVD. 2008. Outlier detection for skewed data. J Chemom, 22(3-4): 235-246.
• Ibrahim AH, Elshwadfy LM. 2021. Factors affecting the accuracy of construction project cost estimation in Egypt. Jordan J Civ Eng, 15(3): 329-344.
• Kaiser MJ. 2021. A review of onshore and offshore pipeline construction and decommissioning cost in the USA - Part 1: Specifications, cost estimation and onshore construction. Int J Oil Gas Coal Technol, 27(3): 247-285.
• Kim GH, An SH, Kang KI. 2004. Comparison of construction cost estimating models based on regression analysis, neural networks, and case-based reasoning. Build Environ, 39(10): 1235-1242.
• Parker NC. 2004. Using natural gas transmission pipeline costs to estimate hydrogen pipeline costs. URL= https://www.researchgate.net/publication/254396811 (accessed date: July, 15, 2024).
• Rui Z, Metz P, Wang X, Chen G, Zhou X, Reynolds D. 2012. Inaccuracy in pipeline compressor station construction cost estimation. SPE Annual Technical Conference and Exhibition, October 20-22, Texas, US, pp: 4219-4234.
• Rui Z, Metz PA, Reynolds DB, Chen G, Zhou X. 2011a. Regression models estimate pipeline construction costs. Oil Gas J, 109(14): 120-127.
• Rui Z, Metz PA, Reynolds DB, Chen G, Zhou X. 2011b. Historical pipeline construction cost analysis. Int J Oil Gas Coal Technol, 4(3): 244-263.
• Seabold S, Perktold J. 2010. Statsmodels: econometric and statistical modeling with Python. Proceedings of the 9th Python in Science Conference, June-July, Texas, US, pp: 92-96.
• Sueri M, Erdal M. 2022. Early Estimation of sewerage line costs with regression analysis. Gazi Univ J Sci, 35(3): 822-832.
• Thaduri RK. 2012. Oil and gas pipeline construction cost analysis and developing regression models for cost estimation. MSc Thesis, Texas A&M University, Institute of Technology & Science, Texas, US, pp: 34-35.
• Ugur LO, Kanit R, Erdal H, Namli E, Erdal HI, Baykan UN, Erdal M. 2019. Enhanced predictive models for construction costs: A Case study of Turkish mass housing sector. Comput Econ, 53(4): 1403-1419.
• Ulvestad M, Overland I. 2012. Natural gas and CO2 price variation: Impact on the relative cost-efficiency of LNG and pipelines. Int J Environ Stud, 69(3): 407-426.
• Yılmaz NF. 2005. Petrol ve doğal gaz boru hatları üzerine genel bir değerlendirme. Tesisat Müh Derg, 87: 4-14.
• Zhao J. 2000. Interim report IR-00-054 diffusion, costs and learning in the development of international gas transmission_lines. URL= https://pure.iiasa.ac.at/id/eprint/6192/_(accessed date: July, 15, 2024).