Comparative thermodynamic analysis of China Spallation Neutron Source second phase (CSNSII) SRF system cooling Scheme

Abstract

The China Spallation Neutron Source second phase (CSNS-II) accelerator upgrades consist of three major components: linac energy upgrade from 80 MeV to 300 MeV, injection system upgrade, and new magnetic alloy dual harmonic cavities. There are ten double spoke cavities cryomodules and eight elliptical cavities cryomodules in total. A comparative thermodynamic analysis of the superconducting radio frequency (SRF) systems can be of great importance to select a reliable, cost-effective and easy-maintenance cooling schemes. This is particularly critical for SRF systems that require long-term operation. A comparative thermodynamic analysis of both types of cooling schemes for SRF systems, referred to as Scheme 1 (a cryogenic distribution system with distributed heat exchangers) and Scheme 2 (a cryogenic distribution system with central heat exchangers), was presented in this paperThe length and size of process transfer lines, as well as the heat loads subjected to their components, of cryogenic distribution systems were introduced. In addition, the unsteady state cooldown process for the CSNS-II SRF cryomodules was investigated. The analysis results that included temperature and pressure profiles along the linac distance, as well as distributions of JT expansion effectiveness, mass flow rate, and cryomodule exegetic efficiency. Furthermore, the exergy analysis and refrigeration power are investigated using a neural network optimized by a novel sparrow search algorithm genetic algorithm (SSA-GA). Results show that, Scheme 1 is more advantageous than Scheme 2 because it can save maintenance time. From 300 K to 4.5 K, the total cooldown time of SRF cryomodules is approximately 28 h. The maximum ΔP is around 1.1 bara, which can be considered in the acceptable range. The prediction of ηex is stable at the 7th epoch after being optimized by SSA-GA, and the MSE is 1 × 10-5. And the error between the predicted and true values of the BP neural network after SSA optimization is very small (less than 0.1%), approaching zero, indicating that the BP optimized by SSA achieves good prediction. This paper's work can serve as a solid foundation for the construction and operation of the CSNS-II.