Multi-Objective Optimization of an SOFC-GT-ORC Integrated Multi-Generation System With Liquefied Natural Gas As Cold Source Based on Intelligence Method

Abstract Solid oxide fuel cells (SOFCs), as a power generation technology that can efficiently convert chemical energy into electrical energy without any pollutant emissions, have attracted widespread global attention in recent years. While conventional SOFC systems use gas turbines (GT) for the initial utilization of exhaust gases, the temperature of the emissions remains comparatively high. Therefore, further waste heat recovery for the SOFC-GT system is crucial for improving energy utilization efficiency. This paper proposes an integrated multi-generation system that can produce electricity, heating, and cooling based on SOFC, GT, and organic Rankine cycle (ORC), utilizing liquefied natural gas (LNG) as a cold source. The system not only recovers waste heat from the power generation process but also captures the significant cold energy released during LNG vaporization. A mathematical model of the system is constructed from the energy, exergy, and economic perspectives, and key thermodynamic parameters that significantly affect the system performance are analyzed and determined. To enhance the optimization efficiency of the system, a backpropagation neural network (BPNN) intelligent algorithm is used to train a performance prediction model for the system. This intelligent model serves as a surrogate model for the thermodynamic model to conduct the multi-objective optimization design of the system. Finally, the reliability of this intelligent method is validated. The results indicate that through the multi-objective optimization design based on the BPNN model, the system can achieve a system efficiency of 72.33%, exergy efficiency of 50.92%, and levelized cost of energy of 0.0127$/kWh. By comparing the multi-objective optimization results based on the intelligent model with those based on the thermodynamic model, it can be observed that the use of the intelligent method not only accelerates the optimization process but also provides relatively accurate results.