2025 Spring Meeting and 21st Global Congress on Process Safety

(156a) Maximizing LNG Throughput: Real-Time Optimization with High-Fidelity Models

Authors

Udit Gupta, Siemens
With recent concerns over lower-than-expected growth in LNG demand and an expected wave of new export capacity, it is more important than ever for LNG operators to maximize their plants’ availability, efficiency, and throughput rates without resorting to design modifications to remain competitive. While digital tools based on machine learning offer some value, they are limited by the data available and often fail to achieve optimal operation. This presentation will share our latest experience in using high-fidelity physics-based digital solutions, combined with real-time data, to optimize operational setpoints and achieve additional throughputs of 1-3% while minimizing the cost of production.

Liquefaction plant operators face several challenges in optimizing process performance. The highly integrated nature of liquefaction processes, coupled with numerous decision variables, makes it difficult to identify marginal improvements without reliable decision support tools. Additionally, technology licensors protect their intellectual property, and major process equipment like cryogenic exchangers remain black boxes to operators, complicating the development of effective tools. Our approach leverages high-fidelity models based on first principles, which provide a robust framework for optimization. These models are validated using historical plant data, ensuring high predictive reliability for operational decisions.

A data reconciliation module reconciles plant data before optimization to address data uncertainty and random errors. Finally, any operational performance optimization tool must be available online as a digital solution suitable for real-time optimization. Liquefaction process performance is continuously affected by changes to the richness of the gas processed, seasonal weather changes, and gas availability. The bottleneck within a train can shift between the pre-treatment, refrigeration, liquefaction, fractionation, or utilities, and the tool should provide solutions considering all these operational constraints.

In conclusion, this presentation will delve into the challenges and solutions for optimizing liquefaction plant performance in the face of rising LNG demand. We will explore the limitations of current machine learning techniques and highlight the advantages of high-fidelity physics-based digital solutions. By leveraging real-time data and advanced modeling, we will demonstrate how operators can achieve sustained additional throughput of 1-3% without design modifications. Additionally, we will discuss the importance of data reconciliation and the need for robust, online optimization tools that adapt to changing operational conditions. Join us to discover how these innovative approaches can unlock significant value and enhance the efficiency of liquefaction trains.