Scholarlink

<p>In this paper, an improved parameter identification algorithm for new nonlinear models is investigated. The traditional parameter identification algorithm has low identification accuracy when dealing with systems containing hysteresis nonlinearities and unknown perturbations. An auxiliary model predictive gradient identification algorithm is proposed to determine coefficients of the model. By combining with the principle of hierarchical, a separable synchronous iterative algorithm based on the auxiliary model predicted gradient improvement is proposed to improve the accuracy and stability of parameter identification. The algorithm not only uses the predictive gradient descent algorithm to effectively jump out of the problem of local extremes but also ensures the convergence and convergence speed of the identification process by introducing the forgetting factor recursive least squares. Meanwhile, the paper also explores how to apply data preprocessing in machine learning to system identification. Finally, the effectiveness of the algorithm is verified by conducting experiments on real data from the electromagnetic scanning micro-mirror.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Separable synchronous auxiliary model hybrid predictive gradient identification for nonlinear models based on the data preprocessing

  • 期刊论文

  • 发表日期：

    2025年4月1日

查看全文
  • Ya Gu,
  • Lin Chen,
  • Chuanjiang Li,
  • Quanmin Zhu

摘要

In this paper, an improved parameter identification algorithm for new nonlinear models is investigated. The traditional parameter identification algorithm has low identification accuracy when dealing with systems containing hysteresis nonlinearities and unknown perturbations. An auxiliary model predictive gradient identification algorithm is proposed to determine coefficients of the model. By combining with the principle of hierarchical, a separable synchronous iterative algorithm based on the auxiliary model predicted gradient improvement is proposed to improve the accuracy and stability of parameter identification. The algorithm not only uses the predictive gradient descent algorithm to effectively jump out of the problem of local extremes but also ensures the convergence and convergence speed of the identification process by introducing the forgetting factor recursive least squares. Meanwhile, the paper also explores how to apply data preprocessing in machine learning to system identification. Finally, the effectiveness of the algorithm is verified by conducting experiments on real data from the electromagnetic scanning micro-mirror.