Robust Full Model Identification Method For Input Nonlinear Dynamic Systems With Noise Model

Industrial processes generally have inherent complex problems such as uncertainty,nonlinearity,parameter distribution and time-varying,but the current modeling ideas and identification theories for complex dynamic processes are imperfect.In addition,the noise in the traditional identification theory is generally white or colored Gaussian noise.For non-Gaussian noise,especially heavy tail noise or impulse noise,the relevant identification theory and identification research are insufficient.Based on the above situation,this paper develops a robust identification method under the influence of colored non-Gaussian noise for dynamic process with input nonlinearity.First,in terms of modeling,the paper models the input nonlinear system as a Hammerstein model,and the colored non-Gaussian noise is generated by a non-Gaussian white noise source through the Box-Jenkins colored noise model.Then,based on the Hammerstein Box-Jenkins model,the paper uses the least absolute criteria and proposes a two-step iterative full model identification algorithm.Specifically,the uncertain and unknown items in the information vector are effectively approximated by the hierarchical iterative identification theory.Besides,the performance index of least absolute criteria can effectively suppress the impact and impact of impulse noise and heavy tail noise on the identification results by summing the absolute value of errors,so as to further realize the robustness of model parameter estimation.Finally,under the influence of colored non-Gaussian noise,the digital simulation experiment uses the proposed iterative identification algorithm to achieve robust estimation of all parameters of the process model and the noise model,and verifies the effectiveness and efficiency of the identification algorithm from multiple perspectives.