Page 200 - Intelligent Digital Oil And Gas Fields

156 Intelligent Digital Oil and Gas Fields

Repetitive tasks that require engineering support and include smart analytics are as
follows:
• Forecast (prediction) analysis:
intelligent alarming by exception-based surveillance
event recognition and diagnosis
identification and tracking of well production opportunities
short-term failure detection of artificial lift (ESP, GL, RP) including
early detection of unexpected water production (after water
injection)
predictive advisory short-term forecasts
production optimization and long-term forecasts (>30days)
production and injection management.

5.2.3 Software Components of an E&P Workflow
Automated workflows are a series of processes programmed using a comput-
ing language that is capable of executing logical and calculation instructions
with minimum human intervention. Saputelli et al. (2013) in a classical
paper have showed the best practices in DOF in the last 10years. They
showed a series of technologies which gave origin to the modern integrated
analysis. Fig. 5.5 shows the main steps of an automated engineering
workflow which include: (1) store information in a database, (2) filter and
condition (cleanse) data and extract high-frequency and average data, and
(3) send data to the various models and applications in the workflow, which
can include both data-driven models and physical models. Fig. 5.5 highlights
the importance of these key elements.
Database. Accessibility to data through a database is the starting point of
any automated workflow. For production workflows, a database is config-
ured to link data from different data sources (which include data from dif-
ferent time frames and/or frequencies), select and organize data, and send it
to technology for filtering and conditioning.
Filtering technology. Mathematical algorithms programmed either in
stand-alone or Web-based applications are required to perform data filtering
and cleansing (which was discussed in Chapter 3). Many factors, such as sig-
nal problems, data transfer, weather/environmental problems, instrument
errors, and human interruptions, contribute to errors in the raw data.
The filtering process is programmed to remove erroneous data and outliers
and provide representative values of the data during the time period of eval-
uation. When the data is received, cleaned up, and post-processed, the main
output is an average value of raw data—that is, production rate (Q), pressure