Current events have once again highlighted the vital importance of natural gas. Demand for the fuel will only continue to grow as decarbonisation efforts gain momentum. But reliance on pipeline supplies is not always a viable—or perhaps desirable—option, so more and more countries are coming to rely on LNG imports. And as LNG takes centre-stage, it becomes more and more important to maximise production and energy efficiency at existing liquefaction facilities, particularly as many forecasts predict that medium- and long-term demand will exceed production, perhaps by as much as 700mn t by 2040. The panellists on PE’s recent webinar, Increasing production and improving the energy efficiency of LNG operations with AI, examined the potential role AI can play in that vital process of optimising and raising efficiency levels.
Gas has an important role in the energy transition due to its reliability, availability and relatively low emissions, says Nikhil Gulati, head of applied machine learning at BakerHughesC3.AI. Demand is rising, and a supply-demand gap will emerge over the next 3-4 years, so the industry is facing additional pressure to optimise its existing production capacity. Binu Mathew, CTO for oil and gas at C3.AI, agrees, saying there will be a significant LNG demand increase with the energy transition, necessitating the industry finds more than just traditional efficiency gains.
But LNG plants are complex and have to accommodate changeable inputs—feed quality, for example—and traditional process optimisation is highly labour-intensive, Gulati explains. Non-linear relations between inputs and production are often missed by traditional process-control systems. Machine learning can give extra value, and operators can take action based on the resulting recommendations. AI for optimisation offers “significant opportunity” across the liquefaction process, granting increased output, reduced variability in production rates and lower emissions intensity, says Gulati.
There has been a lot of progress recently on the data collection side, but making sense of the sheer amount of data generated by an LNG facility needs techniques such as AI, Matthew says. C3.AI solutions are unique because of their open framework and their focus on integration with all the existing systems in the plant, all with one unified interface. C3.AI, Baker Hughes, Shell and Microsoft have formed a consortium called the Open AI Energy Initiative, Mathew explains, the first open ecosystem of AI-based solutions to address these challenges, with the goal of combining the strengths of each participant. For instance, optimisation for LNG combines BHC3 Process Optimisation with Shell’s LNG-specific process optimiser and Baker Hughes’ TPS iCenter. Deep knowledge of the process is vital, and these solutions are integrated to function as a single application and fully supported by Baker Hughes and C3.AI, Matthew says.
Connie Griffin, E2E optimisation product owner at Shell, sees an opportunity for process optimisation in most businesses since information is often segregated. This application allows different departments and teams to work together more effectively, and collaboration with Baker Hughes and C3.AI allows us to bring additional value from aggregate data to drive optimal business performance goals, Griffin explains, adding that being able to visualise handles, gap-to-potential and closed-loop performance monitoring in one place is key to open loop-value realisation.
Shell’s operational experience
Shell developed an optimiser for the LNG space, based on its extensive experience as an operator, explains Fabio Mazzocchetti, senior solution leader at C3.AI. One of the key lessons learned is that a common framework is needed to exchange information, share insight and reach solutions much faster. And that is made possible by the system’s ability to integrate disparate data sets. Baker Hughes, C3.AI and Shell own the underlying code and model pipelines/structure for the optimisation systems, but the actual data and the actual model belong to the customer, Mathew says.
Mazzocchetti gave a demonstration of the system from the perspective of a production engineer, explaining that Shell’s LNG module has already been deployed across multiple facilities. The system calculates additional revenue saved through optimisation and shows “gap-to-potential”, displaying the difference between actual production and optimised production and comparing recommendations with implementations. The system generates recommendations, the frequency and receivers of which can be customised. Most customers use the technology in an open-loop system, Mazzocchetti says, but closed loop—where recommendations are pushed through directly—is also a possibility.
The webinar, Increasing production and improving the energy efficiency of LNG operations with AI, held in association with BakerHughesC3.AI, is now available to view on demand.