Machine Learning for Mixed Refrigerant Composition in LNG Plants

The patent implements dynamic adaptation by training a machine learning model with historical operation data that captures varying operating conditions (feed gas composition, ambient temperature) and their corresponding optimal mixed refrigerant compositions. The model dynamically determines the optimal refrigerant composition based on current operating conditions, enabling the system to adapt to changing conditions while maintaining production efficiency.

The patent replaces complex mechanical control systems with a machine learning-based information processing system. Instead of using multiple sensors and complex control algorithms to determine optimal refrigerant composition, the system uses a trained model that processes operating condition data and outputs optimal composition recommendations, simplifying the overall control architecture while improving efficiency.

The patent uses historical operation data as a virtual copy of actual plant operations to train the machine learning model. This training data captures the relationships between operating conditions and optimal refrigerant compositions without requiring real-time complex measurements. The model then uses this learned knowledge to determine optimal compositions in real-time based on readily available operating condition data.