Autonomous control of heat pipes through digital twins: Application to fission batteries

Abstract

Fission batteries are envisioned as nuclear energy systems and associated technology that can be fully utilized in a battery-like operation, where the system is delivered as a ‘plug-and-play’ service. Several key attributes define the desired functionality of these systems: economic, standardized, installed, unattended, and reliable. The construction and operation of unattended, plug-and-play fission batteries will require sufficiently robust hardware and software technologies. Using a digital twin (DT) may reduce costs and risk associated with employing fission batteries through the integration of the disparate systems used in the design, construction, and operation of these nuclear energy systems. A DT employing machine-learning (ML) and physics-based representations to forecast future performance could potentially be used for anticipatory control. Before application of a DT in the fission domain, the DT technology should be validated in a non-fission environment. A DT of a single-heat-pipe test article in the Microreactor AGile Non-nuclear Experimental Testbed (MAGNET) was demonstrated with predictive, self-adjusting capability on 30 March 2022. The test plan stated that: (1) the operators will manually change the temperature set point of the heat pipe to an upper limit or lower limit; and (2) the DT will predict the temperature will go beyond this temperature threshold, and then will update the temperature set point to the baseline temperature without any human intervention. The DT used a two-step process including a least absolute shrinkage and selection operator (LASSO) for variable selection between the sensors followed by vector autoregressive (VAR) models for multivariate forecasting to predict future performance of the heat pipe within MAGNET. Additionally, a physics model was created within the Sockeye framework that would be applied in future tests. With controlled rates of temperature change, the DT successfully self-adjusted the heat pipe before reaching the lower limit under expected conditions. Ultimately, this DT could be leveraged as a foundational framework in future fission battery applications for continuously monitoring and actively self-adjusting a heat pipe.