Optimization of Multi‐Stage CO 2 Liquefaction and Recirculation Processes Using Artificial Neural Networks

ABSTRACT This study presents an integrated framework for optimizing energy‐efficient CO 2 liquefaction in carbon capture and storage (CCS) systems, combining thermodynamic modelling in Engineering Equation Solver (EES) with artificial intelligence (AI). The simulated liquefaction pathway includes five‐stage compression, intercooling, a counter‐flow heat exchanger and a Joule–Thomson valve. A key performance driver is the inclusion of a CO 2 gas recovery loop, which recycles unliquefied vapour. This configuration reduces specific power consumption by 18%–22% and increases liquid production by 23%–26% compared to baseline operations, with only marginal increases in energy input. Further parameter optimization of compressor pressure ratios and cooling capacities resulted in producing 30.83 kg/s of liquid CO 2 at 20.13 MW. This represents an additional 12%–15% reduction in specific energy consumption (SEC) and a 28.96% improvement in overall process effectiveness. To facilitate rapid performance assessment, a deep artificial neural network (ANN) was developed using the Levenberg–Marquardt algorithm. The model, featuring a 64‐32 neuron hidden layer architecture, achieved exceptional predictive accuracy ( R 2 = 0.992968) with mean absolute errors below 0.99. This integrated EES and AI approach provides a robust, high‐speed tool for the design and optimization of large‐scale CCS infrastructure.