Enhanced performance evaluation and operational regulation of a novel combined cooling and power system using machine learning

The combined power and cooling system based on the S-CO₂ Brayton cycle is a proven solution for meeting the multi-energy needs of distributed energy systems. By reusing the working medium from the refrigeration system for further power generation, energy utilization efficiency is markedly improved. This paper proposes a combined cooling and power system with high-pressure mixing, which facilitates the reuse of the working medium and reduces the mass flow rate of the main compressor in the S-CO₂ recuperation Brayton cycle. Machine learning models, utilizing two-layered feedforward neural networks, are judiciously developed and employed to predict the off-design performance of turbomachines. The operational characteristics and regulation of the high-pressure mixing (HPM) and low-pressure mixing (LPM) systems are evaluated and compared using multi-objective optimization with a genetic algorithm. The results indicate that the HPM system excels in converted thermal efficiency, while the LPM system is superior in refrigeration performance. The optimal converted thermal efficiencies are 47.6 % and 32.3 % for HPM and LPM systems under constant turbomachine performance. Based on the machine learning model, corrected optimal converted thermal efficiencies of 48.02 % and 32.88 % are achieved for the HPM and LPM systems, respectively. This research presents an innovative concept for distributed energy systems with diverse energy requirements.