Inference of Properties of a Natural Gas Processing Plant Through the Application of Machine Learning to Time Series

Resumo


Context: Virtual analyzers or inferences are mathematical models widely used in the chemical process industry, as they allow real-time prediction of properties of interest from measurements of basic properties available at the appropriate frequency. Problem: Therefore, they are a key component of chemical process control and optimization structures. Inferences with superior prediction capabilities provide a better process control reducing losses and improving profits. Solution: This work seeks to establish whether Machine Learning (ML) algorithms and techniques could be used to generate inferences for the ethane content in Liquefied Petroleum Gas (LPG) produced in a Natural Gas Processing Unit (NGPU) with advantages over the inferences traditionally used. IS Theory: This work is associated with the Theory of the knowledge-based company, assisting in decision-making and efficiency in the application of resources. Method: In order to accomplish this, a dataset with five years of real data from a Petrobras UPGN has been obtained, and models with different types of algorithms were generated. The machine learning models are compared between them and against models similar to the inferences currently used in the plant, in order to verify whether the ML tools have real potential to provide better results. Summary of Results: The obtained results showed that the machine learning models have a much better representation capability when compared to the traditional inference models. Contributions and Impact in the IS area: The paper contributes to the use of machine learning techniques to generate better inferences for the ethane content in LGP produced in a NGPU, making the analytical process more agile and assertive to increase the generation of value for Petrobras.

Palavras-chave: Natural Gas, Time Series, Machine Learning, Modeling and simulation

Referências

Alexandre F. Cordeiro. 2011. Estudo de simulação e controle de uma unidade de processamento de gás natural. Dissertação de Mestrado em Engenharia Química, Escola de Química, Universidade Federal do Rio de Janeiro - UFRJ.

J. R. Parrish, C.B. Brosilow. 1985. Inferential Control Applications. Automatica. 21(5). Pergamon Press, International Federation of Automatic Control.

Campos, Mário César M. M., Gomes, Marcos V. C. , Perez, José M. G. T. 2013. Controle Avançado e Otimização na Indústria do Petróleo. 1ª ed. Interciência. Rio de Janeiro, Brasil.

Shima Khatibisepehr, Biao Huang, Swanand Khare. 2013. Design of inferential sensors in the process industry: A review of Bayesian methods. Journal of Process Control. Volume 23, Issue 10. Pages 1575-1596, ISSN 0959-1524, https://doi.org/10.1016/j.jprocont.2013.05.007. [link]

Morettin, Pedro. A.; Toloi, Clélia M.C. 2018. Análise de Séries Temporais Modelos lineares univariados. Volume 1 3ª edição. Blucher, São Paulo, Brasil.

Machine Learning Mastery, Jason Brownlee, May 8th 2017. How to Convert a Time Series to a Supervised Learning Problem in Python. Acessado Sept 10th, 2022 from [link].

A. Khotanzad, C. Chung. 1998. Application of multi-layer perceptron neural networks to vision problems. Neural Comput & Applic 7:249-259. Springer-Verlag London Limited

Machine Learning Mastery, Jason Brownlee, July 21st 2016. Time Series Prediction with LSTM Recurrent Neural Networks in Python with Keras. Acessado Sept 10th, 2022 from [link].

Ibomoiye D. Mienye, Yanxia Sun, Zenghui Wang. 2019. Prediction performance of improved decision tree-based algorithms: a review. Procedia Manufacturing 35. Pages 698-703. ISSN 2351-9789, https://doi.org/10.1016/j.promfg.2019.06.011, [link]

Lilian R. Canabarro. 2011. Inferência Dinâmica em Colunas de Destilação para uma Unidade de Processamento de Gás Natural. Dissertação de Mestrado em Engenharia Química, PEQ/COPPE, Universidade Federal do Rio de Janeiro - UFRJ.

Antonio A. C. Freitas. 2007. Previsão de séries temporais via seleção de variáveis, reconstrução dinâmica, ARMA-GARCH e redes neurais artificiais. Tese de Doutorado em Engenharia Elétrica, Faculdade de Engenharia Elétrica e de Computação, Universidade Estadual de Campinas - UNICAMP.

Cartwright, Hugh M. 2020. Machine Learning in Chemistry. 1st Edition. The Royal Society of Chemistry. Croydon, CR0 4YY, UK.

Towards Data Science, Michael Keith, June 17th 2022. Model Validation Techniques for Time Series. Acessado Sept 10th, 2022 from [link].

Korstanje, Joos. 2021. Advanced Forecasting with Python: With State-of-the-Art-Models Including LSTMs, Facebook's Prophet, and Amazon's DeepAR. Apress Berkeley, CA.

Antonio R. S. Parmezan, Vinicius M.A. Souza, Gustavo E.A.P.A. Batista. 2019. Evaluation of Statistical and Machine Learning Models for Time Series Prediction: Identifying the State-of-the-Art and the Best Conditions for the Use of Each Model. Information Sciences 484 (May 2019), Elsevier, 302-337, https://doi.org/10.1016/j.ins.2019.01.076, [link].

Medium, Soumya Shrivastava, January 17th 2020. Cross Validation in Time Series. Acessado Sept 10th, 2022 from [link].
Publicado
29/05/2023
SILVA, Luciana; MARQUES, Marcelo; GOMES, Marcos; ESCOVEDO, Tatiana; KALINOWSKI, Marcos. Inference of Properties of a Natural Gas Processing Plant Through the Application of Machine Learning to Time Series. In: SIMPÓSIO BRASILEIRO DE SISTEMAS DE INFORMAÇÃO (SBSI), 19. , 2023, Maceió/AL. Anais [...]. Porto Alegre: Sociedade Brasileira de Computação, 2023 .