Please login first
PMA-MDO based Performance Optimization Strategy for Steam Generator Level Control System of Nuclear Power Plant
1 , * 1 , 1 , 1 , 2
1  Xiamen University of Technology,Xiamen,China
2  China Nuclear Power Engineering Co., Ltd.Shenzhen, China
Academic Editor: Juan Francisco García Martín (registering DOI)

Steam generator (SG) is the key equipment in the energy transfer process of nuclear power plant, and its level control is particularly important for the safe and stable operation of nuclear power plant. In the commissioning process of nuclear power plant, it is often necessary to adjust the control parameters of the steam generator level control system (SGLCS) to achieve performance optimization. Traditional solutions include model-based optimization (MBO) and model-free optimization (MFO), in which MBO depends on the accurate relationship model between control parameters and control performance. However, the level process of steam generator is time-varying and highly nonlinear, which makes it difficult to establish the model accurately. In addition, MFO is implemented without considering any prior information, and its optimization efficiency is also restricted to a certain extent. In order to make full use of prior data information and integrate the respective advantages of MBO and MFO, this paper proposes a multi-source hybrid data-driven optimization method based on the prior model accuracy (PMA-MDO) on the basis of the data-driven idea and stochastic approximation algorithm. Firstly, the method uses the prior data information to construct the initial optimization model. Then, the current iteration point is tested into the actual working condition and prior model to evaluate the accuracy of the local area of the current model. When the accuracy of the model meets the requirements, the model gradient estimation is used; otherwise, the online gradient estimation is used. Afterwards, a new iteration point is obtained by using step size calculation. Finally, the iteration termination criterion based on historical running data is taken as the judgment principle. If the new iteration point meets the iteration termination criterion, the optimal value will be output; otherwise, the current iteration data will be fused with the prior model for model reconstruction, and the iterative optimization process will be repeated until the system iteration process is optimal. In this paper, the PID parameter optimization tuning of three-impulse steam generator level control system is taken as an example. The simulation results show that this method has better optimization performance than the traditional SPSA, and can significantly improve the efficiency of steam generator level control performance optimization.

Keywords: Nuclear power plant; Steam generator; MBO; MFO; Multi-source hybrid data-driven; Prior model accuracy; PID parameter optimization tuning