The integration of data-driven, expertise-based and physics-based approaches for artificial lift optimisation can be a differentiator in a prolonged low oil price environment, says Dr Venkat Putcha, PhD in Petroleum Engineering, director of data science at OspreyData
Oil prices have been operating in the US$30-US$65 per barrel range for the majority of time since 2016. The US Energy Information Administration projects Brent oil price will average US$53 per barrel for both 2021 and 2022. The oil and gas industry has been further motivated to innovate to improve efficiency and reduce costs as a result of low oil prices. According to IHS CERA, digital oilfield implementations have led to a 2-8% production increase, 5-25% reduction in operating expenditures, and 1-10% reduction in capital expenditures. The popular strategy for advancing towards a digital oilfield entails the automation of processes and productionised workflows through a centralised system for data collection, integration, processing and analytics.
The Digital Oilfield Outlook Report by GE, Accenture and JWN lists the top four priorities among oilfield digitalisation as fleet management, field productivity, production asset optimisation and predictive maintenance. Currently, 85-90% of oil wells utilise some type of artificial lift. Misconfigured artificial lift systems directly translate to reduced asset production and sub-optimal field productivity. Traditionally, production operators, engineers and managers have used their expertise and engineering intuition, along with physics-based models, to identify inefficiencies and opportunities to improve. These may include simple examples, such as using fundamental physical equations and empirical rules of thumb for decision making, or more sophisticated examples such as employing nodal analysis software for modelling, design and optimisation.
Over the past decade, large volumes of oilfield data have been gathered. Such data includes sensor signals, production data, asset metadata such as deviation, location and completion, event data such as logs of various equipment failures, downtime and maintenance activity. This has enabled the development of data-driven models that help better manage systems through systematically detecting anomalies, identifying inefficiencies and automating the classification of worst offenders
An expertise-based approach: pros and cons
Expertise-based, physics-based and datadriven approaches have their benefits and limitations when implemented exclusively. Experts employ their experience to provide quick recommendations on systems. This is a simple, fast and hands-on approach when experts are available to monitor wells. Expertdefined rules are a good fall-back when data is not available or sufficient. However, expert supervision is not a scalable solution, as it is not possible to manually monitor and optimise all wells at all times. Furthermore, expertdefined rules may be limited to a specific range of wells, types of wells, fields or plays. Expert-based rules that operate in the North Sea may not be as effective in the Ghawar oilfield of Saudi Arabia
A physics-based approach: pros and cons
Physics-based models are much more widely applicable than expertise-based systems. However, these models pose a different challenge. Physics-based models must be calibrated with field data in order to make the models meaningful. This process may need updates as well characteristics change, without which the model would fail to simulate and predict well behaviour accurately. In some cases the detailed information or detailed physics-based models required to accurately represent the physical system and study certain problems may not be available. For example, an advanced simulation model of an electrical submersible pump (ESP) well can estimate the liquid production as a function of the pump head boost, however, it may not be possible to simulate the well’s behaviour while the ESP is undergoing gas locking….Continue reading.
