Hierarchical linear and nonlinear adaptive learning model for system identification and prediction

摘要: In this paper, we propose a method to increase the model accuracy with linear and nonlinear sub-models. The sub-model applies least square error (LSE) algorithm uses neural networks (NN). two sub-models are updated hierarchically using Lyapunov function. proposed has advantages: 1) is multi-parametric model. Using model, weights of NN can be summarized into coefficients or parameters auto-regressive eXogenous/auto-regressive moving average (ARX/ARMA) structure, making it easier establish control laws, 2) learning rate ensure convergence errors at each training epoch. One improve whole system. We have demonstrated by experimental studies that technique gives better results when compared existing studies.