Pressure decline data from high temperature and high pressure tests of gas transport into the oil phase are often divided into three regions, representing different fluid transport mechanisms. Clustering of these regions -early-time, transition and late-time regions- is typically based on changes in slopes of pressure decline data with time. For more details on the context of the experiment and dataset used in this notebook, please refer to Chapter 6 of my Ph.D. dissertation and article SPE-200341-PA.
https://era.library.ualberta.ca/items/db2f5685-6375-40aa-8301-e581caadf7ca
This notebook presents two unsupervised machine learning algorithims -K-Means and K-Plane- for clustering pressure decline vs. time data. The structure of this Jupyter Notebook is as follows:
- Prepare problem
a) Load libraries
b) Load dataset
- Dataset summary
a) Descriptive statistics
b) Data visualizations
- Evaluate and compare algorithms
a) K-Plane clustering
b) K-Means clustering
- Conclusions