Download PDFOpen PDF in browser

Ensemble Models Based on System Identification and Machine Learning for Downhole Pressure Simulations

EasyChair Preprint no. 11115

6 pagesDate: October 23, 2023


The digital era has come, the use of data analytics, cloud applications and environments is growing faster than years ago in oil and gas subject. The industry own a huge quantity of data storage from Instrumentation, logging and sensors. In the other side reservoir simulation has a vital role in petroleum engineering due it application to fluids flow rates forecasting specially for exploitation strategic definition. The complexity of using sophisticated mathematical models can define a specific simulation study scenario and even with simplifications it needs the domain knowledge in different areas such as reservoir, production, engineering and geosciences. Some ideal assumptions can simplify the differential equations making not representative of the complex behavior of the underground fluid and without these assumptions the reservoir complexity increase greatly. In this paper, we use a public data from Volve field corresponding to the production years 2008 to 2018. We build an assemble model based on system identification, with a Non Linear Auto-Regressive (NARX) model and Artificial Neural Network (ANN) architecture to simulate reservoir model and forecast real time the downhole pressure for short-term decisions. The ensemble method did not provide better prediction compared to standalone short-term simulations techniques. We use the initial dynamic data to build a consistency model able to downhole pressure forecasting helping the oil production optimization with significant error decrease.

Keyphrases: Artificial Neural Network, Ensemble, hybrid model, NARX, production data, production simulation, Volve field

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
  author = {Daniel de Marins and Helon Ayala},
  title = {Ensemble Models Based on System Identification and Machine Learning for Downhole Pressure Simulations},
  howpublished = {EasyChair Preprint no. 11115},

  year = {EasyChair, 2023}}
Download PDFOpen PDF in browser