Reliability analysis of core power optimization control using Kalman filter for accelerator driven system based on reconfigurable computing

•Stochastic differential equation model has been developed to analyze the random behavior in ADS.•Kalman filter has been adopted to achieve the optimal core power control in operation process.•Closed-loop control structure has been established to describe the mapping relation between proton beam intensity and core power.•All the algorithms have been implemented on the ZYNQ using reconfigurable computing technology to meet the real time design requirements. Accelerator Driven System (ADS) is a promising nuclear facility for the transmutation of minor actinides and long-lived fission products, which is constructed by proton accelerator, spallation target and sub-critical reactor. For the industrial implementation, the reliability of core power control system in the operation process has to be analyzed and verified in considering the dynamic characteristics and real time response for the safety operation strategy. Therefore, a digitalization of Instrumentation and Control (I&C) system for ADS that can predict the changing value of proton beam intensity has been suggested in this research to get the appropriate variation of reactor core power without arousing the unexpected impulse to the main structures. In order to accurately analyze the random behavior of the precursor concentrations and the neutron density in sub-critical reactor, the stochastic differential equation model has been developed, and the closed-loop control system in describing the mapping relationship between the proton beam intensity and the corresponding core power has been established. The Kalman filter has been adopted for the optimal design of proposed analysis method to achieve the competitive predicting performance in operation process, and all the algorithms have been implemented on the XILINX ZYNQ System-on-Chip (SoC) using reconfigurable computing technology to meet the real time design requirements for embedded applications. Finally, the reliability of core power control model has been tested against Monte Carlo calculations and all the results demonstrate the good performance in predicting the core power with acceptable efficiency.