Abstract: To address the problems of poor physical interpretability and huge sample size requirement when using neural networks to fit nonlinear control system models for state prediction, this paper ...