Multivariable control of large-scale pressurized water reactor based on LPV-MPC algorithm over a wide operating range

For energy conservation, large-scale Pressurized Water Reactors (PWRs) are required to participate in grid peak shaving. Compared to traditional PID control, Model Predictive Control (MPC) algorithm offers advantages in handling multivariable control problems. However, PWR systems exhibit strong nonlinearity under wide-range operation, limiting conventional MPC performance. Therefore, Linear Parameter Varying (LPV) models, valued for their nonlinear approximation capability and low computational cost, are employed in nonlinear MPC. This paper proposes an LPV model-based MPC algorithm to enhance the control of a nonlinear HPR1000 model. An iterative optimization strategy is adopted for LPV model identification: first, initial local models are identified using the asymptotic method; then, parameters of the weighting functions are estimated via least squares; finally, the LPV model is optimized iteratively based on the total data. The local models demonstrate good performance in both time-domain and frequency-domain validation. The optimized LPV model shows high similarity to the plant’s step responses at both high and low loads, confirming its suitability for MPC controller. The LPV-MPC is compared with both the original control system and a fixed-model MPC (F-MPC) through test scenarios at 100 % and 50 % FP. The results indicate that the LPV-MPC controller offers comparable performance at 100 % FP, but exhibits faster regulation speed and stronger capability in suppressing SG water level fluctuations at 50 % FP. The comparison with F-MPC demonstrates that the LPV-MPC possesses stronger adaptability and robustness. Meanwhile, a large-scale transition scenario was implemented. The comparative analysis confirms that the LPV-MPC controller possesses superior load-following performance.