Production Forecasting Using Transfer Learning of Pretrained Deep Model
Fluid rate measurement and forecasting is crucial for the field development. The goal of this project is to develop a virtual-based flow meter to forecast the production rate of wells from the Norwegian Volve field using deep learning models. This project introduces a novel idea showing the implementation of transfer learning and pretrained deep Neural Basis Expansion Analysis for Interpretable Time Series Forecasting (N-BEATS) model in production forecasting. Prior work was limited to applying feature-based linear regression algorithms and traditional sequence deep learning models, mainly RNNs, LSTMs, and Long Short Time Series Network (LSTnet), to predict the pressure response of a single well with a single fluid. In this project -- we extended the application of deep learning research by introducing two methods, an attention-based model using the Temporal Fusion Transformer (TFT) and a transfer learning approach using pretrained N-BEATS on M4 series. Both TFT and the pretrained N-BEATS models outperformed the traditional LSTM model, increasing the test score (MAE) by 0.04 and 0.08, respectively, demonstrating excellent matching results by N-BEATS. We concluded that using transfer learning and pretrained N-BEATS model eliminates the previous disadvantages of LSTM models requiring multivariate features and a large training history. This research shows promising results for using transfer learning and N-BEATs model, especially for new or green fields with limited historical data.
Detection of Stylolite Zones in Reservoirs Using Machine Learning Methods
Stylolite is a special geopattern that can occur in both sedimentary rocks and deformed zones. Such rough surfaces could change porosity of the matrix and modify the permeability either in a positive way by producing microcracks, or a negative way by dissolving soluble minerals and deposit in the near vicinity, which sometimes could even form to act as horizontal permeability barriers. Multiple fields in Abu Dhabi such as Upper Zakum oil field have encountered big production losses from such barriers in the past. As a potential approach to supplement traditional methods, machine learning methods can be used to locate a stylolite zone in a reservoir and further determine if it is barrier.
Scale Buildup Detection and Characterization in Production Wells by Deep Learning Methods
Inorganic scale results from brine solution deposition onto the surfaces of the reservoir, wellbore and other production facilities. In this project, we focused on studying scale deposition in the wellbore, which can introduce decrease in tubing ID and induce abnormal production declines and pressure drops. There are limited recent studies focusing on scale deposition detection, while most of them only managed to detect the problem after considerable production declines. Therefore, as the objective of this work, we tried to implement deep learning methods to detect the scale deposition immediately after its occurrence, and to estimate the magnitude of the scaling to help with the further design of its removal.
Inferring Subsurface Information using Tidal Analysis
Earth, ocean, and atmospheric tides are naturally occurring and offer an opportunity to infer subsurface information. The downhole pressure in closed wells or water level in open wells may include periodic signals with dominantly diurnal and semidiurnal periods induced by
earth tides, ocean tides, and atmospheric tides. Using the oscillation signal, reservoir properties can be evaluated by calculating the phase difference and amplitude ratio between the recorded pressure/water level fluctuations and the corresponding theoretical tides.
An Integrated Framework for Production Data Analysis Using Machine Learning and Wavelets
The modeling framework introduced the Maximum Overlap Discrete Wavelet Transform
Multiresolution Analysis (MODWT-MRA) as a useful transform for decomposing production time series data. Moreover, the research proved that applying the MODWT-MRA is equivalent to decomposing a single well's data into a set of virtual wells that present simpler behavior when compared to the original flowrate and pressure readings). These virtual wells decomposition is then leveraged with the use of machine learning and deep learning models to capture the reservoir response.
A New Look into Nonlinear Regression
Well testing provides a set of useful tools to estimate reservoir parameters. Based on the estimates, it is possible to forecast the reservoir behavior for future production and hence a reservoir can be utilized in an optimal way. Nonlinear regression was introduced to well testing more than two decades ago and revolutionized well testing with new methods and substantially improved the accuracy of the prediction.
Analyzing Fractures with Resistivity Data
As a potential approach to augment and improve on the existing methods, resistivity measurements can be used to characterize fractures. Fractures saturated with water effectively reduces the resistivity of rock matrix and create opportunities to extract fracture characteristics by monitoring the change in resistivity distribution.
Automated Analysis of DTS Data
The research is focused in applying Machine Learning to aid with the interpretation of DTS data. The goal is to automatically detect the presence of fractures from the data and give an estimation of flow from them. Having the capacity to automate some of the interpretation would allow for the use of fiber optic data to real time decision making.
Distributed Temperature Sensing
Distributed Temperature Sensing is a newly developed measurement technique, and was introduced into the oil industry in recent years. DTS can provide high resolution, real-time and continuous temperature information along a wellbore. With the increasing need for monitoring reservoir production, more and more permanent downhole gauges are installed in oilfields. Compared with traditional measurement tools (e.g. PLT), DTS has many advantages, making it more suitable for permanent installation.
Flow Behavior of Gas-Condensate Wells
Gas-condensate reservoirs experience reductions in productivity by as much as a factor of 10 due to the dropout of liquid close to the wellbore. The liquid dropout blocks the flow of gas to the well and lowers the overall energy output by a very substantial degree. As heavier components separate into the dropped-out liquid while the flowing gas phase becomes lighter in composition, the overall composition of the reservoir fluid changes due to the combined effect of the condensate phase behavior and the rock relative permeability.
Large Volume Data Processing for Permanent Downhole Gauges
Permanent Downhole Gauge (PDG for short) is a newly developed tool for well testing in the petroleum industry. In traditional well testing, pressure and flow rate transient data are collected for a short period which leads to a large uncertainty. Due to long time continuous data acquisition, a PDG may provide measurements for several years or longer. However, at the same time, this brings a new problem--a large volume of noise together with large volume of measurement.
Machine Learning Applied to Multiwell Data
This research focuses on solving multiwell problems using machine learning approaches with Python. So far, the study has identified important features in a single-phase multiwell problem. This study will explore several different machine learning algorithms and compare them in terms of accuracy and performance. Python is very helpful for this purpose as it has a lot of machine learning libraries to choose from and is handier in solving machine learning related problems.
Machine Learning in a Full-Physics Analysis
The preliminary objective of this research is to investigate the possibility to represent and/or replace the numerical reservoir simulation model using a proxy based on machine learning. The approach is envisioned to take advantage of the recent progress in machine learning and data mining approaches to help complement or replace parts of the functionality that numerical reservoir simulation models provide.
Modeling reservoir temperature transients, and matching to permanent downhole gauge (PDG) data for reservoir parameter estimation
Over the last decade, permanent download gauges (PDGs) have been used to provide a continuous source of downhole data in the form of pressure, temperature and sometimes flow rate. The tools provide access to data acquired continuously over a large period of time and containing reservoir information at a much larger radius of investigation than conventional wireline testing.
Monitoring and Control of Multilateral Wells
A smart (or intelligent) well is a nonconventional well that is completed with downhole instruments such as pressure and temperature sensors coupled with donwhole control devices. The smart well can be a segmented, single horizontal well where every segment is controlled by independent control valves or a multi-lateral well where each lateral is controlled separately.
Relating Time Series in Data to Spatial Variation in the Reservoir Using Wavelets
Accurate description of the reservoir is crucial to reservoir management. Yet, due to the complex nature of reservoir heterogeneity, obtaining accurate description of the reservoir poses a big challenge.