Machine learning is great, but effective well management means getting the right data to the right humans – where and when they need it.
By Grant Eggleton
Drilling and production engineers are well-acquainted with the constant grind of firefighting that defines so much of well management. Determining the artificial needs for an entire lifecycle of a well is an ongoing challenge and so is the pinpointing of integrity issues in the tree, wellhead, annulus, tubing and valves.
Historically, the upstream industry has largely invested in development projects and drilling activities. However, at the current 50$ per barrel price, exploration is less attractive. Capital projects have either been killed off or put on hold and operational headcount reduced.
That puts even more pressure on engineers to anticipate problems while having a clear, up-to-date view of the current and future profitability of production assets.
But how to go about getting it? In a time when operations are awash in disconnected data and engineers are continually asked to do more with less, emerging technologies like artificial intelligence and machine learning are being touted as a panacea for addressing all inefficiency.
While AI can be a brilliant source of information, it's only of many info resources available to harried engineers. Rather than stuffing another gadget into an already bulging toolkit, they need to simplify and consolidate what they have to achieve something new: operational intelligence.
With so many wells under management, it can be hard for staff in the field and in the office to constantly monitor every single one. The tedious and time-consuming tasks around monitoring are perfect candidates for a predictive monitoring, machine learning or artificial intelligence solution. With the increase in the amount of data available from the well head through the broader adoption of SCADA (supervisory control and data acquisition) systems, computers now have the ability to monitor wells around the clock and alert engineers to possible signs of failure.
However SCADA is only one of the data sources. There are also multiple production and reservoir models and performance equations. This is why access to a self-service environment with all the data available to monitor the wells on behalf of the engineers.
A single-data environment and exception-based-surveillance process
Data is everywhere. SCADA data, production accounting information, drilling and completions information, maintenance and reliability data, well header data, information from spreadsheets – and more. Not only is there a ton of data, but it's also disjointed in most cases because disparate systems are used to capture it. Knowing what information to look for and how best to use it become immediate challenges for subject matter experts like production engineers and technicians.
That's why it is important to help engineers find and interpret that data by giving them the tools to go beyond standard industrial engineering calculations entered into a spreadsheet. Adding value here means enabling engineers to self-serve on data so that they don't have to go back to IT to obtain and understand the data. They need to be able to understand and interpret data themselves, to see and identify leading indicators to mitigate risk, or reduce its duration in the worst case – and to do it faster.
AI and machine learning are really just new sources of data, and all too often the people selling those solutions skip over the ned to include the human intelligence that exists in the business. AI models can be too highly static and time-consuming to update. Context, nuance, experience and yes intuition all still count. Even when an AI solution is appropriate the subject matter experts need to be part of the equation.
Here's where a single data environment and exception-based-surveillance processes come in. A single data environment brings together data from disparate sources and connects it to digital representations of the physical assets, giving production engineers and technicians a one-stop, self-service environment for configuring rules and performing key analysis. Exception-based-surveillance processes provide a 24/7 watch over company assets and trigger alerts and tasks when predefined events occur, allowing personnel to manage more wells and fix problems in less time.
Here's a practical example
Foreign materials – salt, scale, paraffin and others – have a tendency to build up in produced water (PW) wells. This build-up occurs slowly over time and can hinder production if an acid treatment isn't mobilised in due course.
This is a scenario where you could use field-data-capture information, an injection curve (PQ) and SCADA data to monitor the pressure-flow correlation against the curve over time. Using the derived operational intelligence, you could then create a rule that would provide prioritised notifications long before material build-up became a serious issue, giving you plenty of time to plan the intervention activities needed to keep production rates at their highest.
The operational decisions you make are only as good as the data you have. Being able to view data from multiple disparate sources while managing oil and gas operations is critical.
There are only a limited number of hours in the day for technical staff to deal with complications, yet effectively designed systems can alert personnel and give them the vital information to quickly and efficiently solve problems. The use of predictive analytics can help oil and gas management to maximise their staff's workdays, their production and ultimately their companies' overall performance. In today's challenging upstream industry environment, embracing the potential benefits heralded by the rise of the machines obviously makes sense.
But think of them as companions not replacements for your experienced engineers. Improving profitability will happen when data from across the enterprise is brought together and presented to the right people at the right time.
-- Grant Eggleton is Vice President, Global Production Solutions at P2 Energy Solutions (http://www.p2energysolutions.com).
Visit source siteaimachine learningArtificial intelligenceengineer