Eventos Anais de eventos
ENCIT 2022
19th Brazilian Congress of Thermal Sciences and Engineering
A LSTM Neural Network Approach for the Rock Formation Consolidation Inference of Brazilian Sandstone Reservoirs
Submission Author:
Fabio Silva , RJ
Co-Authors:
Fabio Silva, Andre Leibsohn Martins, Victor Carriço, Alexandre Pereira
Presenter: Fabio Silva
doi://10.26678/ABCM.ENCIT2022.CIT22-0362
Abstract
The objective of this work is to present a novel methodology based on data science to infer if a reservoir rock formation is well consolidated or not. Nowadays this type of analysis is performed by different specialists, depending on the seniority level of these geologists this subjective activity can lead to different opinions. Taking into consideration 48 cases from different drilled wells, the model was trained to learn how to tag a reservoir rock formation, between well consolidated or not. Once the well is drilled, the model analyzes 23 engineering variables plus geological data to reach a conclusion. This work proposes a classification statistical model and the usage of a memory based neural network, known as LSTM network. This type of model explores time series characteristics of the problem and it is validated using a cross validation strategy. The dataset is partitioned by groups of wells and its evaluation is done by F1 score, which is a metric for equilibrate precision and recall generally used when the dataset is unbalanced. After training the model, tests were performed and results shown a high identification efficiency: around 90% of accuracy. It is the first time in literature that this approach is used for this specific objective and its results show that this kind of model has a potential to be applied for real-time decision-making, specifically to guarantee if a gravel pack or even any kind of sand control is indeed necessary.
Keywords
completion, Well Completion, Drilling, gravel pack, sandstone, inference, Data analysis, rock formation, real time data
