Data Science
Well-log Based Reservoir Property Estimation with Machine Learning
Well logs are interpreted to estimate the in-situ reservoir properties (petrophysical, geomechanical, and geochemical) which is essential for reservoir modeling, reserve estimation, and production forecasting. The modeling is often based on multi-mineral physics or empirical formulae. When sufficient amount of training data is available, machine learning solution provides an alternative approach to estimate those reservoir properties based on well log data and is usually with less turn-around time and human involvements. The goal of this project was to develop a data-driven model to estimate reservoir properties including shale volume, porosity, and fluid saturation, based on historical set of well logs including gamma ray, bulk density, neutron porosity, resistivity, and sonic. Using Xgboost Regressor, a model was developed to predict the reservoir properties for other wells drilled within the study area. The model has an accuracy of 80%
Product Recommendation System with Machine Learning
Sapa.com is one of the leading eCommerce platforms in Nigeria with millions of daily complete transactions. The goods and services on sapa.com cater for both the elite and the masses which makes it the first choice for almost everybody in Nigeria. Due to the COVID-19 pandemic that struck the entire world in all areas of living, the companies' daily complete transactions have dropped drastically to thousands. My goal as a data scientist in this project was to build a robust intelligent model capable of recommending products and services to Users based on their activities on sapa.com. Using the principle of feature engineering, I selected relevant features form the several features provided by Sapa data engineer to develop a model capable of recommending products to their clients.
Automating Breccia Rock Logging using Convolutional Neural Network
The Arturo district within the Carlin trends possess a lots of historical data (geologic core) that needs to be re-logged for the purpose of rebuilding the geological model of the area. Manually identifying and logging these cores takes a lot of time. Hence the process was automated. I developed a model that can be used to automatically identify these geologic cores based on their physical features using machine learning. Currently, a success rate of 65% have been achieved, and more work is currently ongoing to reach an accuracy of 99%.
Lone Star Reserve Model
As part of the exploration team process in developing an improved approach in quantifying and estimating copper reserve within the Safford district, I integrated data science techniques/libraries (python) with geological domain knowledge through Resource Modelling Solutions library to rebuild lone star reserve model. This approach has helped to automate some redundant procedures in the current modelling workflow.
Morenci Long Range Model
The Morenci copper mine is the largest copper producer in North America and also one of the largest copper producer in the World. The copper deposit is hosted within the basin-and-range province of the U.S. southwest, an area that is controlled by tectonic forces which led to the development of large blocks separated by faults. The copper deposit found in this region is the porphyry deposits. As part of the effort to ensure that we utilize recent development in technology to ensure smooth production in this region, I build the long range reserve model for this district using python. This approach helped us to automate some of our workflow as well as improved turn around time.
Geo Science
Application of Monte Carlo Simulation in Quantifying Geological Uncertainties Associated with Prospect Characterization and Evaluation
Every measurement contains some level of uncertainty due to statistical and random error. Recognizing data ambiguity is an important part of conveying scientific research findings. Calculating the effect of unknown parameters on the best design concept is entailed by Uncertainty analysis.
In hydrocarbon exploration, uncertainties are introduced from the empirical formulas used to evaluate petrophysical parameters, the ones established during data acquisition, the ones factored in from engineering parameters, and the ones introduced from the business. Most of the uncertainties introduced during hydrocarbon production can be associated with the geologic concepts established and delineated during exploration. This research therefore tends to estimate the geological uncertainties of reservoir X based on the static structure of the reservoir using Monte Carlo Simulation.
Impact of Depositional Environment on CO2 Storage
In delineating a potential carbon storage sites, much attention is placed on the physical properties (porosity, permeability, saturation) and sealing ability of the geological formation to be used for storing this gas. Little or no detailed attention has been placed on how the depositional environments will affects the fluid (CO2) to be stored in the formation. It is true that the depositional environment has been previously studied since they have previously acts as reservoirs or aquifers for some fluids (hydrocarbon) or mineralization host (coal, gold, etc.) but the fact remains that the behavior of CO2 when placed in this environment will differ since there are distinct physical and chemical properties between these fluids. This research focused inferring depositional environments from core and full imager log, incorporating it into geologic models to see how it affects CO2.
Automating Breccia Rock Logging using Convolutional Neural Network
The Arturo district within the Carlin trends possess a lots of historical data (geologic core) that needs to be re-logged for the purpose of rebuilding the geological model of the area. Manually identifying and logging these cores takes a lot of time. Hence the process was automated. I developed a model that can be used to automatically identify these geologic cores based on their physical features using machine learning. Currently, a success rate of 65% have been achieved, and more work is currently ongoing to reach an accuracy of 99%.
Lone Star Reserve Model
As part of the exploration team process in developing an improved approach in quantifying and estimating copper reserve within the Safford district, I integrated data science techniques/libraries (python) with geological domain knowledge through Resource Modelling Solutions library to rebuild lone star reserve model. This approach has helped to automate some redundant procedures in the current modelling workflow.
Morenci Long Range Model
The Morenci copper mine is the largest copper producer in North America and also one of the largest copper producer in the World. The copper deposit is hosted within the basin-and-range province of the U.S. southwest, an area that is controlled by tectonic forces which led to the development of large blocks separated by faults. The copper deposit found in this region is the porphyry deposits. As part of the effort to ensure that we utilize recent development in technology to ensure smooth production in this region, I build the long range reserve model for this district using python. This approach helped us to automate some of our workflow as well as improved turn around time.