Gilmore—Cheniere has been working on LNG in the US since the 1970s, though the gas market was in the doldrums for decades. Today, increasing energy prices have revitalized the business. Cheniere saw this coming and develop a new terminal design and a new business from scratch—a top down design involving a business process team, engineering and finally IT design and implementation.
It seems like this kind of technology is fairly a propos regarding the digital oilfield of the future?
de Vries—Customers in all industries recognize the value of transferring information from the shop floor to the business and also in other direction. This is being recognized in power generation and other large plants. The secret is to make offline simulation tools that are reliable enough to be used online—a completely different situation from today. The idea is to provide engineers with shadow costs and KPIs in real time. Moving information out of the plant is getting to be a best practice. But there is a lack of business info flowing TO the floor. This is the problem we have addressed at Invensys by building an IT infrastructure layer adding in multiple simulators. This gets around the problem of instrument tag-based systems.
Can you explain?
de Vries The problem with tag-based systems is that within a plant, there may be a lot of equipment with identical naming systems for different machines—also the structure for naming can change across a plant. Our solution normalizes equipment nomenclature and centralizes operations. Our largest example of such a deployment is in Oman where we have deployed a system across 17 oil and gas fields—along with a whole phone book of equipment suppliers.
But that’s quite a high level of granularity.
de Vries Yes—we’ve not yet got all information to where it can be used to best effect. You need simulators to study what’s going on in wells, in gathering systems, for real time production allocation. During our different deployments we have accumulated a lot of information from customers’ sites. The next step is to ask, ‘what do we do about the business process?’ For Shell we have developed techniques to look for emerging patterns in data indicative of particular events like sand breakout or pressure drop. These can be addressed with appropriate remedial actions. Today it can be hard for upstream customers to take knowledge and associate it with a particular event because IT still lives in a ‘tag time value’ world.
So this is a data mining issue?
de Vries—Yes, but traditional data mining tools are not very good at time series analysis. We try to capture discrete events, when they start, when they end. The system is constantly looking for patterns in the data; when things go from good to bad, when production drops off, gas lift parameters change, flow from LNG terminal changes. We organize data so that business software can understand it.
Is this productized?
de Vries—It is the ArchestrA Collaborative Environment.
Does Cheniere use SAP on Sabine Pass?
Gilmore—Not SAP. But the ERP system is always key. Upstream businesses want to make enterprise finance visible to plant people and help operations make financially correct decisions. We offer real time accounting with our Dynamic Performance Measurement offering—this has been a big transformation.
de Vries—We are learning from customers. The old way of producing a report every 30 days just produces ‘ancient history,’ not actionable results. On the other hands, ‘targets’ can be counter productive. The answer is to provide real time accounting and make people aware how they are doing. Our tag line for the shop floor is ‘what did I make today?’.
In process control I get the impression that the data historian is both a huge focus and a barrier to information flow?
de Vries—Yes—the historian is both those things. It stores tag, time, value data. We are in the process of transforming information—not just capturing event history but providing trends and forecasts. Invensys is one of the three main providers of process historians. I say, let the historian do what it does. But use another data store to run your plant. Here you can associate measurement with derived values. But we have to recognize the strong culture within our clients that depends on the historian and Excel spreadsheets.
Gilmore—People used to complain that they didn’t have enough data about what was happening. Then came the Historian and folks suddenly has vast amounts of data and no time to look at it! So we have to help operations and business people filter and pre-process data—solving local potential problems. It’s a kind of ‘triage.’ Like a doctor asking, ‘Is this a cold or pneumonia?’
de Vries—In the past, technology workers knew individual tag names. People got familiar with their environment. Now we have fewer, more mobile people so you can no longer assume such familiarity—there is no longer the ability to noodle around with data in the historian to understand what’s happening. Today, ‘events’ need to be turned into actions.
What data store do you use?
de Vries—Any RDBMS will do—we often use Microsoft SQL Server unless customers want something else.
Gilmore—And we have our own Industrial SQL product based on SQL Server extended into real time data.
Is the RDBMS the right tool for time variant data?
de Vries—The RDBMS is horrible for time variant stuff! But other object datastores are not ready for prime time. Customers do ask, but at the end of the day, we need to put our data into a commercial database.
Gilmore—This is an evolving thing—a changing world.
de Vries—Indeed and speaking of change, these projects present huge change management issues involving IT, technologists and the business. This can create rivalries. So it is easier to start with white sheet of paper like Chiniere than to retrofit.
Speaking of the barrier of the Historian, do you encounter the ‘silo mentality’ when dealing with different parts of your clients’ businesses?
Gilmore—Sure. There are even completely different personalities. It takes a cowboy’s personality to bring an asset on stream and a dairyman’s to operate it. Some have tried to optimize handover by giving a project team operational responsibility for the first year of production. But this was not really effective. As Stan said, it’s a big change management issue.
© Oil IT Journal - all rights reserved.