Abstract
Advanced nuclear reactors play an important role in the energy future of the United States and the rest of the world. They are designed and operated based on a different model than that of the current operating fleet, thus enabling deployment in remote locations and allowing for safe semi-autonomous or autonomous operations. Such characteristics require the development of a new reactor control paradigm. A significant factor in the development of control technologies and methods is integration of the various technologies and methods with each other and with hardware (both reactor system hardware and control hardware). A recent workshop on control of advanced reactors identified the lack of a flexible, expandable software/hardware infrastructure to enable such integration as a key gap. A previous phase of the current effort involved developing and demonstrating the Control and Optimization Modular Modeling Application for Nuclear Deployment (COMMAND) platform, which is capable of integrating autonomous-control-enabling technologies and methods, without the constraints imposed by existing software solutions. This platform will enable advanced reactor developers to deploy and test any developed technologies and methods by employing a common framework, and to couple them with their own models and hardware. The present phase of this effort entails using the Microreactor Automated Control System (MACS) platform, which was developed by the U.S. Department of Energy (DOE) Microreactor Program, to serve as a control method testbed. MACS can be used by advanced reactor developers to integrate their control-related research activities with any reactor system. For the present effort, MACS was customized to mirror Idaho National Laboratory (INL)’s Microreactor Applications Research Validation and Evaluation (MARVEL) microreactor, and COMMAND was leveraged to enable MACS to emulate the physics of MARVEL, thus positioning MACS as a physical twin of MARVEL. To create this physical twin, MARVEL models were integrated into COMMAND. Previously, the MARVEL Reactor Excursion and Leak Analysis Program (RELAP5 3D) model was modified to allow simulations to be run with interactive process control (using COMMAND), and in this effort was further customized to accurately represent asymmetric drum rotations. The MARVEL Monte Carlo N Particle (MCNP) model was also integrated to reflect the excess reactivity generated by moving the MACS drums, as well as to capture the spatial power distribution of the MACS core. As MCNP is computationally expensive and takes a great deal of time to run, machine learning (ML) was utilized to develop a surrogate modeling approach capable of estimating the relevant simulation results in milliseconds. An ML block was integrated within COMMAND to allow for interfacing with the surrogate models, each of which captures a specific MCNP parameter of interest. To enable COMMAND to send signals to and receive feedback from MACS, the generic Remote Procedure Calls (gRPC) communication protocol was integrated within COMMAND for interfacing with MACS hardware. Integration of MACS and the MARVEL physics models via the COMMAND software platform was successfully demonstrated. The next phase of this effort will introduce and demonstrate a control scenario utilizing MACS and COMMAND. Both COMMAND and MACS are expected to continue to expand and evolve, thus closing the gap as an integrated software/hardware infrastructure and fostering advancement toward autonomous advanced reactor operations.
Funding Information
  • USDOE Office of Nuclear Energy (NE)
  • DOE (AC07-05ID14517)