Analysis of compositions of flowback water from marcellus shale wells by utilizing data mining techniques

Yuewei Pan, John Yilin Wang

Research output: Contribution to conferencePaperpeer-review

2 Scopus citations

Abstract

Volume and salt concentrations in Marcellus flowback water depend on geology, drilling and completions, stimulation and flowback operations. Recent studies include evaluations of geochemical origins based on the compostition concentrations, flowback sampling analysis and numerical studies. However, an in-depth understanding of chemical compositions as well as the changes of compositions is still needed. In this paper, we will first review the literature related to flowback water in Marcellus shale gas wells to fully understand the chemistry, geochemistry, and physics governing a fracture treatment, shut-in, and flowback. We will then gather all public and in-house flowback data, named as 3-week or 3-month flowback in this work, to build a data set of flowback water compositions. After data screening, we will then analyze this composition database using four different methods: geographical, changes over time, linear regression, clustering and multi-variable analysis. New understandings such as the magnitude and prevailing trends of concentrations for target constituents as well as the correlations among flowback compositions, the differentiation between early and late time flowback water were obtained and explained on the basis of geochemistry and physics. Guidelines for a comprehensive sampling protocol will be provided based on our analysis.

Original languageEnglish (US)
StatePublished - 2017
EventSPE Eastern Regional Meeting 2017 - Lexington, United States
Duration: Oct 4 2017Oct 6 2017

Other

OtherSPE Eastern Regional Meeting 2017
Country/TerritoryUnited States
CityLexington
Period10/4/1710/6/17

All Science Journal Classification (ASJC) codes

  • General Engineering

Fingerprint

Dive into the research topics of 'Analysis of compositions of flowback water from marcellus shale wells by utilizing data mining techniques'. Together they form a unique fingerprint.

Cite this