You are Home   »   News   »   View Article

Using graph databases in operations

Monday, July 23, 2018

There are many places in the oil and gas industry where graph databases could prove better than spreadsheets and databases, said Joe Chesak of FabLabs.

There are many places in the oil and gas industry where graph databases could prove much better than both spreadsheets and relational databases for data which is highly inter-dependent, said Joe Chesak of Norwegian software company FabLabs. This includes managing production, linking production with reservoirs, or building a computer system from an engineer's mental model.

He was speaking at the Finding Petroleum forum in Stavanger on Nov 29, 'Transforming offshore operations'.

The graph database management system, 'holds' the same data as a relational database, but with completely flexible data structures and gives you tools to model and monitor your data. It can also hold all data in tabular format.

In the oil business graph databases have a lot to offer because of their ease of use, and ability to directly represent the real world. They should be used in upstream oil and gas much more, Mr Chesak said.

As an example, the graph database Neo4j can be used for anything from 'throw-away' resource allocation simulations, to massive all-topside optimization projects., he says. Further, using graph databases 'allow you to think differently, without constraints purely about your domain. In the end it helps you go directly to what is important, understand what is going on with your wells and topside, and ultimately make better decisions".

Engineers are given large amounts of different types of data, and expected to synthesise it all in their minds. "There is a professional pride in that," he said. But the systems are not designed with much consideration about how much detail a person is even capable of parsing and synthesizing.

Databases aside, Excel is really the engineer's go-to tool for working with data, he says. But that means putting data into rigid boxes again, and making a system which does not naturally map to the real world, he said. The graph database can store data in a way much closer to the real world, with a collection of nodes, or things, and relationships between those things.

Graph databases can also be used for time series data, structuring it according to the real-world connections between these data streams.

Mr Chesak's research on how the oil and gas industry operates is based on his experience as 'embedded tech' in a production engineering department of a major oil and gas company, for about 5 years.

The original role was to streamline data reporting, but he found that he could make a bigger impact by moving up-stream data-wise, to work with data flows starting from the raw data arriving from offshore. Developing a single source of truth available make it that much easier for people to use tools like Neo4J 'and identify patterns nobody knew existed'.

Engineers do their own modelling

Consider several engineers having a discussion, and drawing something on a whiteboard. Typically you see labelled boxes, circles, flows and connections, to show how something works.

'This is really how you model something in a graph database,' he said. If you can draw something on a whiteboard with your team and you feel you have everything of importance represented, 'you are 90 per cent of the way to having a working model in a database that you can populate with real data and query endlessly."

'Engineers, in my experience, model what they can, using the industry grade fit-for-purpose software that is sold to the company. They make decisions based on that, plus a lot of other stuff in their heads that they cannot shoehorn into the tools,' he said.

'If the engineers could extend the fit for purpose tools in the same way they extend their own spreadsheet they'd be in a much better place. They would be in control of what questions they can ask.'

It is necessary to have a model in which the entire team can accept as 'close enough' to how their real world works, he says. It is important for engineers to get involved in doing the modelling, rather than expect the system to do it all for them. They know the personality of their wells, they are closer to the problem than big industry suppliers can ever be.

Tuning production

Mr Chesak's Norwegian company, Fablabs AS, has built a software solution for production engineers, called Production Tuner, which can be used to build flow models about how wells and topsides are operating. The aim is to remove bottlenecks.

This way it can help companies make sure they are making the most of the available production capacity, at minimum operational cost, while staying within safety limits. The model accounts for all wells and all constraints to flow, such as limited capacity of a separator.

To imagine it, consider that if you had 2 or 3 wells, it might be easy to work out in your head how to maximise production while not needing more separator capacity than you have available. But if you had hundreds of wells, all with different proportions of oil, water, gas, CO2 and H2S in their flows, it quickly gets too hard to do work it out in your head or in a spreadsheet. With flows changing by the hour or minute, it is practically infeasible.

The ProductionTuner software monitors what is happening in real time system wide, and calculates forward looking scenarios. So if something changes in the production system, you immediately have a viable solution at hand, including when you are in an 'edge case' that you didn't know might happen until it does.

This is very different to the usual production engineering set-up, where most decisions are made by an engineer periodically checking different screens displaying data from a production historian.

Production teams typically group their wells in some logical way, and assign each group to a different engineer. This means each engineer can act as a mini silo, and competing for resources on topsides with other production engineers, without the knowledge or intent to maximise the flow from the export pipe, he says.

Associated Companies
» Fablabs AS
comments powered by Disqus


To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.


Latest Edition Oct-Nov 2020
Oct 2020

Download latest and back issues


Learn more about supporting Digital Energy Journal