Advanced Exergy Analysis and Optimization of Heat Exchanger Network under Uncertainty

Abstract

In this work, we utilized advanced exergy analysis in conjunction with an integrated framework that combines artificial neural network (ANN) with particle swarm optimization (PSO) and genetic algorithm (GA) to evaluate performance of heat exchangers within a crude oil distillation unit under uncertainty. At the beginning, we constructed an equilibrium-state Aspen model, which subsequently used to conduct both traditional and advanced exergy analyses. This exergy analysis enabled us to quantify exergy efficiency and the level of irreversibility within the heat exchanger network (HEN). Subsequently, we calculated four components of irreversibility—endogenous, exogenous, avoidable, and unavoidable—for the equipment exhibiting significant incompetence with advanced exergy analyses. The operational model was subsequently converted into a dynamic configuration by introducing ±10% fluctuation into system variables, such as temperature, and mass flow rate, resulting in the creation of a dataset comprising 600 samples. Five ANN models were developed using this dataset, each designed to predict various aspects, including overall exergy efficiency, exergy destruction, modified exergy efficiency, unavoidable exergy destruction and avoidable exergy destruction. ANN model served as a substitute within the PSO and GA environment to optimize HEN under uncertain condition. The optimized operational parameters obtained within the PSO and GA methods were further validated by feeding them into Aspen model for validation through crossreferencing. The exergy analysis revealed that HEN had 66.16% exergy efficiency, and the exergy destruction was 5403.166 kW. Advanced exergy analysis further revealed that avoidable exergy destruction amounted to 1,759.80 kW, while unavoidable exergy destruction stood at 3,643.35 kW. Seven heat exchangers, displaying the highest exergy destruction rates, were identified as priority candidates for intervention due to their significant impact on the network. The effectiveness of both the GA and PSO optimization methods exhibited similar results, and they notably enhanced the exergy performance of the facility when correlate to the standalone Aspen model of the process.