Hybrid MPC-P controller for the core power control system at TRIGA reactor.

Abstract

The Training, Research, Isotopes, General Atomics (TRIGA) research nuclear reactor is utilized in a wide range nuclear applications and has three versions, Mark I, Mark II, and Mark III. The TRIGA reactor that is available in Malaysia is a Mark II type, Reaktor TRIGA PUSPATI (RTP). The core power control system in the RTP is designed based on many components including the Feedback Control Algorithm (FCA), and the control rod selection algorithm (CRSA). The CRSA is used in the RTP operation to provide external control to the reactor core reactivity. Even though the CRSA provide stability to the reactor operation, it has relatively complex software structure which produce delayed response time for the control rods selection. Thus, the Single Control Absorbing Rod (SCAR) algorithm is designed to improve the reactor performance by minimizing the time response for the rods selection. However, the SCAR yields a non-constant tracking performance due to the highly non-linear dynamic of the control rods. Thus, to compensate the drawbacks of the SCAR with the existing FCA controller, a hybrid controller is proposed based on the integration of a linear model predictive control (MPC) and Proportional (P) controller named as Hybrid MPC-P-SCAR. The MPC application in the SCAR enable the controller to handle multiple constraints through built-in prediction ability to select the best sequence of future control rod velocity during transient state. Whereas at the steady-state, the MPC is optimized by a scheduled P control switches within the allowable range of power variation. The performance of the proposed Hybrid MPC-P-SCAR, the MPC-SCAR and the existing Feedback Control Algorithm with SCAR (FCA-SCAR) is compared via simulation and evaluated in terms of tracking performance, workload of the control rod drive mechanism (CRDM), and the offset elimination in the unmeasured disturbance. Overall, the Hybrid MPC-P-SCAR controller reduces the settling time by 45% and the steady-state error by 61% of the nominal value compared to the FCA-SCAR.