An Oil&Gas production system is based on the pressure balance between the reservoir and the delivery point. Since a deviation of the flowing conditions in a single point of the production system has consequences in the entire asset, a big challenge for flow assurance engineers is to preserve a stable pressure in the system, with a particular attention at the separator level. Given that the description of pressure oscillations in the systems provided by the available simulators, based on first principle, is not always enough accurate and reliable as needed for practical applications, it is fundamental to develop models and tools to help plant operators in understanding the phenomenon and properly managing it. In this context, the objective of the present thesis work is to use large datasets, containing historical measurements of hundreds of plant signals, for identifying the most critical components of the production system with respect to the pressure oscillation phenomenon and extracting knowledge from it. To this aim, a novel indicator of the intensity of the pressure oscillation phenomenon in the separators has been firstly defined combining Discrete Short Time Fourier Transform and Principal Component Analysis. Then, a method to extract information on the causes of the oscillation phenomenon has been developed. It is based on (i) the prioritization of the plant signals importance with respect to the pressure oscillation by using the Maximum Information Coefficient and the moment-independent Kolmogorov-Smirnov distance; (ii) the aggregation of the signal for the identification of the most critical plant components; (iii) the extraction of rules describing the phenomenon by developing a Classification And Regression Tree model whose inputs are the most important signals in (i). This method has been verified considering a production plant operated by Eni S.p.A. where the oscillations in the separator’s pressure are frequent. The results show the effectiveness of the developed indicator of the intensity of oscillation and provide hints about the physical causes.
Un impianto di estrazione di Oil&Gas è basato su un unico bilancio di pressione tra il giacimento e il punto di consegna. Dal momento in cui una variazione nelle condizioni di flusso in un singolo punto ha conseguenze sull’intero sistema, una grande sfida per gli ingegneri di flow assurance è il mantenimento di una pressione stabile nell’impianto, con una particolare attenzione all’altezza del separatore. Poiché la descrizione delle oscillazioni di pressione nell’impianto fornite dai simulatori disponibili, basati su modelli fisici, non è né sufficientemente accurata né affidabile quanto necessario, è fondamentale sviluppare modelli e strumenti in grado di aiutare gli operatori nel comprendere e controllare il fenomeno. In questo contesto, l’obbiettivo del presente lavoro di tesi è di sfruttare il dataset, contenente misurazioni storiche di centinaia di segnali dell’impianto, per identificare i componenti più critici del sistema di produzione rispetto al fenomeno di oscillazione della pressione ed estrarre informazioni da esso. A questo scopo, come prima cosa è stato sviluppato un innovativo indicatore dell’intensità di oscillazione della pressione nel separatore combinando una Trasformata di Fourier Discreta a Tempo Breve e l’Analisi delle Componenti Principali. Successivamente, è stato sviluppato un metodo per l’estrazione di informazioni sulle cause del fenomeno di oscillazione. Questo è basato su (i) la prioritizzazione dei segnali dell’impianto in base all’importanza rispetto al fenomeno di oscillazione di pressione usando il Coefficiente di Massima Informazione and la distanza Kolmogorov-Smirnov; (ii) raggruppando i segnali per l’identificazione delle componenti più critiche dell’impianto; (iii) con l’estrazione di regole che descrivano il fenomeno sviluppando un modello di Classificazione e Regressione ad Albero i cui input sono i segnali più importanti in (i). Questo metodo è stato verificato su un impianto di produzione operato da Eni S.p.A. dove le oscillazioni di pressure nel separatore sono frequenti. I risultati mostrano l’efficacia dell’indicatore di oscillazione sviluppato e fornisce suggerimenti riguardo alle cause fisiche del problema.
Sensitivity analysis and decision trees for the identification of the causes of pressure oscillations in oil & gas production systems
BORGHI, RICCARDO
2017/2018
Abstract
An Oil&Gas production system is based on the pressure balance between the reservoir and the delivery point. Since a deviation of the flowing conditions in a single point of the production system has consequences in the entire asset, a big challenge for flow assurance engineers is to preserve a stable pressure in the system, with a particular attention at the separator level. Given that the description of pressure oscillations in the systems provided by the available simulators, based on first principle, is not always enough accurate and reliable as needed for practical applications, it is fundamental to develop models and tools to help plant operators in understanding the phenomenon and properly managing it. In this context, the objective of the present thesis work is to use large datasets, containing historical measurements of hundreds of plant signals, for identifying the most critical components of the production system with respect to the pressure oscillation phenomenon and extracting knowledge from it. To this aim, a novel indicator of the intensity of the pressure oscillation phenomenon in the separators has been firstly defined combining Discrete Short Time Fourier Transform and Principal Component Analysis. Then, a method to extract information on the causes of the oscillation phenomenon has been developed. It is based on (i) the prioritization of the plant signals importance with respect to the pressure oscillation by using the Maximum Information Coefficient and the moment-independent Kolmogorov-Smirnov distance; (ii) the aggregation of the signal for the identification of the most critical plant components; (iii) the extraction of rules describing the phenomenon by developing a Classification And Regression Tree model whose inputs are the most important signals in (i). This method has been verified considering a production plant operated by Eni S.p.A. where the oscillations in the separator’s pressure are frequent. The results show the effectiveness of the developed indicator of the intensity of oscillation and provide hints about the physical causes.File | Dimensione | Formato | |
---|---|---|---|
2018_12_Borghi.pdf
non accessibile
Descrizione: Testo della tesi
Dimensione
4.15 MB
Formato
Adobe PDF
|
4.15 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/144258