基于移动数据窗的传递函数多新息随机梯度辨识方法

一些工业过程可以近似用一个传递函数描述,结合统计辨识方法和非线性优化策略提出传递函数参数辨识方法.该方法采用动态数据方案,使用系统观测数据获得系统更多的模态信息.基于动态观测数据,提出传递函数随机梯度参数辨识方法.为进一步提高辨识精度,利用动态窗数据将随机梯度参数辨识方法中的标量新息扩展为新息向量,提出传递函数多新息随机梯度参数估计方法.最后通过仿真例子对所提出的方法进行了性能分析和模型验证.

Moving data window based multi-innovation identification stochastic gradient parameter identification method for transfer functions

For industrial process transfer functions, this paper proposes a parameter estimation method combining the statistical identification method and the nonlinear optimization strategy. In order to obtain enough system modal information, a moving data window is used to deduce the identification method. Based on the dynamical measured data, a stochastic gradient parameter estimation algorithm is presented. In order to improve the estimation accuracy, a multi-innovation stochastic gradient parameter identification method is proposed for transfer functions by expanding the innovation scalar into the innovation vector. Finally, a simulation example is given to compare the proposed methods. Meanwhile the performance of the proposed algorithms is analyzed and the estimated models are verified by step response tests.