Parameter Optimization Method of Steam Generator Level Controller based on SVM and Improved PSO

The steam generator's (SG) water level response process is strongly nonlinear, and the cost of controller parameter tuning is high. To address the above challenge, a hybrid strategy based on traditional Model-based optimization (MBO) and Model-Free Optimization was proposed. An iterative model reconstruction method based on the support vector machine (SVM) and particle swarm optimization (PSO) was thus formulated based on the above idea. This method uses the adjacent iteration point data of the SG response to iteratively update the model and improve the initial population of the PSO so that the accuracy of the optimal solution of the PSO before each SG response could be improved and the iteration number of the iteration could be reduced. Moreover, a modified Simultaneous Perturbation Stochastic Approximation (SPSA) was tested in multiple batches compared with the method in this paper. Simulation results show that this method is more efficient than the SPSA-based model-free optimization method. Furthermore, the improved PSO was verified to have higher optimization efficiency than the traditional PSO.