Surrogate-assisted constrained hybrid particle swarm optimization algorithm for propane pre-cooled mixed refrigerant LNG process optimization

Abstract

The propane pre-cooled mixed refrigerant (C3MR) process is one of the most widely used and efficient natural gas liquefaction processes. However, optimization of this process involves various design and operational constraints, complex thermodynamics, and hence nonconvex in nature. Therefore, it’s computationally expensive, often involving trial and error approaches. Existing optimization algorithms often possess flaws such as insufficient accuracy and premature convergence, leading to suboptimal solutions that may violate essential process constraints. This article proposes a novel method to optimize the C3MR process that handles the computational complexity, meets the constraints, and achieves feasible solutions quickly. The proposed method includes modified feasibility-based constraint handling and radial basis function network-assisted surrogate modeling and optimization, where the power consumption is optimized using a hybrid of the particle swarm optimization (PSO) and social learning PSO algorithm. The proposed algorithm achieves the optimum power consumption (121109.31 kW), which is 21.5% less than the base case (154200 KW). The optimization results are compared with similar optimization algorithms, where the proposed algorithm outperformed the other algorithm in terms of optimal solution, convergence, and speed. The results from this study is compared to previous studies from the literature, which validates the accuracy and applicability of the proposed method.