Reinforcement learning-based exergetic analysis and assessment of a novel cogeneration system for smart urban energy applications

This paper presents a novel Kalina-based geothermal cogeneration system for smart urban energy applications, integrated with a reinforcement-learning-based exergetic analysis and control framework implemented in a digital twin environment coupling EES, TRNSYS and COMSOL. The configuration exploits absorber-integrated internal heat regeneration to raise the working-fluid temperature above the geothermal source without auxiliary energy input. The optimisation results indicated that the reinforcement learning (RL)-based control strategy improved overall energy and exergy efficiencies by 0.066 and 0.058, respectively while simultaneously reducing CO2 emissions by 0.126, levelised cost of electricity by 0.191, and the levelised cost of heating by 0.133.