Page 142 - Intelligent Digital Oil And Gas Fields

Components of Artificial Intelligence and Data Analytics 105

of data integration to examine when an authorization for expenditure
(AFE) was created, and then perform history matching against previous
costs and revenue to see if any added value has been created.

4.1.1 Artificial Intelligence: Overview of State of the Art in E&P
Artificial intelligence (AI) techniques have been used in the E&P industry
since the early 1970s (Bravo et al., 2014). After several decades of R&D
and focused implementation—through smart wells, intelligent fields, expert
systems and real-time analysis, and interpretation of large-scale data for pro-
cess optimization—AI is now maturing in E&P. A literature search indicates
that there is no unique consensus on AI techniques commonly used in the
E&P community; however, artificial neural networks (ANNs), fuzzy-cluster
analysis, evolutionary (genetic) algorithms, genetic optimization, and fuzzy
inference analysis appear to have had a predominant role in applications in
reservoir modeling and simulation, production and drilling optimization,
drilling automation, and process control (Braswell, 2013). For example,
Mohaghegh (2005) combines most of the aforementioned AI techniques
under integrated intelligent systems, dividing them into four main catego-
ries: fully data driven (e.g., developing synthetic well logs), fully rule based
(e.g., well-log interpretation), optimization (e.g., history matching), and
data/knowledge fusion (e.g., candidate-well selection).
However, with the recent expansion of intensively harvesting the
hyper-dimensional, complex, fast/streaming, and Big Data from oil and
gas assets, AI techniques are increasingly seen as compatible with the
methods of predictive data analytics. Recently, the term artificial intelligence
and predictive analytics (AIPA) was coined (Bravo et al., 2014), which puts AI
techniques into a broader context of techniques for data and business ana-
lytics, data mining, process control, automation and optimization, and
advanced visualization. Bravo et al. (2014) provide a comprehensive sum-
mary of AIPA families and techniques, captured in Table 4.3. Selected AIPA
techniques are described later in this chapter.
While the E&P industry is systematically heading toward comprehensive
model integration between static, dynamic, surface and the entire produc-
tion systems, AI is already being deployed to identify model inconsistencies,
narrow model, and process uncertainties; improve forecasts and option
assessment; mitigate risks; and support better decision-making. Moreover,
with the worldwide implementation of DOF programs, the application of
AIPA techniques is also increasing.