Data-Driven Subspace Predictive Control of a Nuclear Reactor

Abstract

This paper introduces a methodology of designing subspace predictive reactor core power control during load-following mode of operation. The central idea is to implement predictive control law directly from the preprocessed input–output data set without using any explicit process model. The controller is designed to include design constraints, feedforward control, and integral control action effectively. Furthermore, time variations in the process are taken into account by recursively updating control parameters with the arrival of new data set. The efficacy of the proposed technique is demonstrated for tracking various load rejection as well as load-following transients for a pressurized water nuclear reactor. A detailed parameter sensitivity analysis is carried out to analyze the controller performance.