Author(s)

Yuan Zhang^*, YanKun Li, Chao Si

Affiliation(s)

Electrical Department, China Institute of Atomic Energy, Beijing 102400, China.

Corresponding Author

Yuan Zhang

ABSTRACT

This study develops a unified dynamic modeling framework for nuclear-grade actuators, such as fans, pumps, and valves, critical for reactor cooling, feedwater regulation, and ventilation. These actuators exhibit nonlinearities, cross-coupling, and operational constraints, challenging conventional controllers to maintain performance and safety under fluctuating conditions. To address this, a neural-network-assisted model predictive control (NN-MPC) architecture is proposed. The framework incorporates a mask-based input encoding strategy to represent all actuators within a single model, and a fully differentiable neural network is trained to approximate actuator dynamics. A bias-compensation mechanism is introduced to address multi-step prediction drift caused by process variations and aging. The resulting nonlinear MPC integrates actuator constraints, output limitations, rate restrictions, and real-time optimization, ensuring millisecond-level performance required in nuclear applications. The implementation workflow includes real-time execution analysis, anomaly detection, safety fallback logic, and hardware-in-the-loop validation. Experimental results demonstrate that the NN-MPC framework provides accurate modeling, high-performance control, and compliance with nuclear safety requirements. This research offers a robust pathway for advancing intelligent, data-driven control systems in next-generation nuclear power plants.

KEYWORDS

Nuclear power plant actuators; Neural network prediction model; Model predictive control; Bias compensation; Real-time optimization

CITE THIS PAPER

Yuan Zhang, YanKun Li, Chao Si. Neural-network-assisted model predictive control for actuators in nuclear power plant drive systems. Eurasia Journal of Science and Technology. 2025, 7(6): 25-35. DOI: https://doi.org/10.61784/ejst3121.

REFERENCES

[1] IAEA. Instrumentation and Control Systems Important to Safety in Nuclear Power Plants. IAEA Safety Guide NS-G-1.3, International Atomic Energy Agency, 2021.

[2] Cho N Z. Dynamic characteristics of nuclear power plant systems. Nuclear Engineering and Design, 2019, 350: 72–84.

[3] Isermann R. Actuators for process control: modeling, design and applications. IFAC Annual Reviews in Control, 2018, 46: 1–12.

[4] Li Y, Wang X, Wang Z. Unified modeling of fluid actuators under complex boundary conditions. Journal of Process Control, 2022, 108: 19–33.

[5] Rawlings J B, Mayne D Q, Diehl M. Model Predictive Control: Theory, Computation, and Design. Nob Hill Publishing, 2017.

[6] Camacho E F, Bordons C. Model Predictive Control. Springer, 2013.

[7] Chen Y, Yin S, Gao H. Data-driven predictive control for nonlinear industrial processes: methods and applications. IEEE Transactions on Industrial Electronics, 2021, 68(7): 5881–5891.

[8] Parisini T, Zoppoli R. Neural approximations for nonlinear optimal control problems. Automatica, 1995, 31(10): 1443–1455.

[9] Raissi M, Perdikaris P, Karniadakis G. Physics-informed neural networks: A deep learning framework for solving nonlinear differential equations. Journal of Computational Physics, 2019, 378: 686–707.

[10] Diehl M, Ferreau H J, Houska B. Efficient numerical methods for nonlinear MPC and moving horizon estimation. In: Nonlinear Model Predictive Control. Springer, 2009.

[11] Lataire R, Butcher M, Nandola N. Real-time nonlinear MPC using automatic differentiation and neural surrogate models. Journal of Process Control, 2020, 92: 31–45.

[12] Zhang X, Jiang B, Shi Z. Fault-tolerant control of safety-critical systems: methods and challenges. Annual Reviews in Control, 2020, 49: 88–102.

[13] Wang W, Chen G, Hou Z. Safety-aware data-driven predictive control for critical energy systems. IEEE Transactions on Control Systems Technology, 2022, 30(6): 2312–2324.

[14] Park HJ, Kim D. Digital-twin-driven control optimization for nuclear power plant thermal-hydraulic systems. Annals of Nuclear Energy, 2023, 182: 109740.