AI and Surge Prediction: Cutting LNG Boil-Off Gas Losses

Liquefied natural gas (LNG) liquefaction terminals are dramatically reducing boil-off gas (BOG) energy losses by integrating mixed-refrigerant compressor surge prediction with thermodynamic artificial intelligence. This article explores how these technologies work together to stabilize the liquefaction process, prevent costly equipment trips, and optimize thermal efficiency in real time. By merging physics-based thermodynamic principles with machine learning algorithms, operators can maintain ultra-low temperatures consistently, thereby minimizing the vaporization of LNG during processing and storage.

The Challenge of Boil-Off Gas in LNG Terminals

Liquefied natural gas must be kept at approximately -162°C (-260°F) to remain in a liquid state. Any heat ingress into the system—whether from ambient temperatures, pipe friction, or process inefficiencies—causes the liquid to warm and vaporize into boil-off gas. While BOG can be reliquefied or used as fuel gas, excessive generation represents a massive energy loss. Managing this requires precise control over the refrigeration cycles, which represent the most energy-intensive aspect of an LNG terminal.

Mixed-Refrigerant Compressor Surge Prediction

At the heart of the LNG cooling process is the mixed-refrigerant (MR) loop, driven by massive centrifugal or axial compressors. These compressors are vulnerable to a destructive aerodynamic instability known as "surge." Surge occurs when the flow rate drops too low for a given pressure ratio, causing a rapid flow reversal that can severely damage the compressor.

Traditionally, operators prevent surge by opening anti-surge valves to recycle gas back to the compressor suction. However, this safety measure comes at a high energy cost:

  • Thermal Inefficiency: Recycling hot gas introduces heat back into the system, instantly reducing cooling efficiency and driving up BOG generation.
  • Wide Safety Margins: Conventional control systems keep compressors operating far away from the surge line to be safe, forcing the equipment to run in a less efficient zone.
  • Unplanned Trips: If a sudden surge causes a compressor to trip, the entire refrigeration loop stalls, leading to a massive spike in BOG as the LNG warms.

Predictive surge technology uses high-frequency sensor data—monitoring pressure differentials, rotor vibrations, and acoustic signatures—to detect the micro-instabilities that precede a surge event. By predicting a surge milliseconds to minutes before it happens, the control system can make micro-adjustments to the compressor speed and guide vanes rather than dumping large volumes of gas through the anti-surge valve. This allows the compressor to operate safely much closer to its peak efficiency zone, maintaining steady cooling and preventing temperature spikes that create BOG.

The Role of Thermodynamic Artificial Intelligence

While standard machine learning looks only at historical data correlations, thermodynamic AI (often called physics-informed AI) embeds the fundamental laws of thermodynamics—such as mass conservation, heat transfer, and fluid dynamics—directly into the neural network's algorithms.

The mixed-refrigerant used in LNG plants is a complex blend of nitrogen, methane, ethane, propane, and butane. The exact ratio of these gases must be continually adjusted based on the incoming feed gas composition and ambient weather conditions. Thermodynamic AI optimizes this process in two critical ways:

  1. Dynamic Refrigerant Blending: The AI continuously calculates the ideal mixture of the refrigerant in real time. It ensures the cooling curve of the refrigerant closely matches the cooling curve of the natural gas in the main cryogenic heat exchanger, minimizing wasted energy.
  2. Subcooling Optimization: Before LNG is sent to storage tanks, it is slightly subcooled below its boiling point. Thermodynamic AI calculates the exact degree of subcooling required to offset the specific heat ingress of the transfer pipes and tanks, virtually eliminating flash gas (a primary source of BOG) upon storage entry.

Synergizing Surge Prediction and Thermodynamic AI

The true reduction in BOG energy losses occurs when compressor surge prediction and thermodynamic AI operate as a unified closed-loop system.

As the thermodynamic AI adjusts the composition of the mixed refrigerant to maximize cooling efficiency, the molecular weight and density of the gas flowing through the compressors change. This shift alters the compressor's surge line dynamically. The surge prediction system immediately maps this new operating envelope, ensuring the compressor adapts to the new refrigerant blend without recycling gas or losing pressure.

By uniting these technologies, LNG terminals achieve a stable, deeply optimized state. Compressors run at peak aerodynamic efficiency without tripping, the heat exchangers operate with minimal thermal stress, and the LNG remains safely subcooled, drastically cutting the energy lost to boil-off gas.