Intelligent thermostatic cold storage design strategies based on Monte Carlo and graph neural networks

We first introduce four annular lattice systems with periodic boundaries as the basic model of a cold storage cooler. The four annular lattice systems represent the four subsystems of the refrigeration system and are located at the four corners of the cold storage. We introduce the energon as a self-driven particle in the system. The energy generated by the energon leaping through the subsystems ultimately serves as the source of cooling power for each subsystem, and the energon can be adsorbed into each subsystem from an external particle source, detached from each subsystem to an external particle source, moved from one subsystem to another by leapfrogging, or leapfrogged within a subsystem. At the same time, we introduce slow movers, which obey a Poisson distribution and enter each lattice point of the ystem with equal probability with parameter. When there are slow movers in the front lattice of the energetics, the leap rate of the energetics will be slowed down, thus preventing the system from being overpowered and fine-tuning the power of the system. We use Monte Carlo simulation to simulate the jumping process of energetons in the system and calculate the cooling power of the system with different parameters. By adjusting the particle adsorption rate, desorption rate, and the entry rate of slow movers, the temperature can drop to the target temperature quickly and temperature. Finally, we use CNN convolutional neural network to perform supervised learning image classification of the spatio-temporal evolution diagrams of closed system, evanescent system and absorbing system, which are three kinds of systems.