This study concentrates on the parameter estimation of multivariate pseudo-linear autoregressive moving average systems by means of the multi-innovation identification theory and data filtering technique. A multi-inno...
详细信息
This study concentrates on the parameter estimation of multivariate pseudo-linear autoregressive moving average systems by means of the multi-innovation identification theory and data filtering technique. A multi-innovation stochasticgradientalgorithm is derived by introducing the innovation length in the stochasticgradientalgorithm. Then, the original system is transformed into two subsystems by using a filter. A filtering-based multi-innovation stochasticgradientalgorithm is presented, whose parameter estimation accuracy is higher than the multi-innovation stochasticgradientalgorithm. The simulation results confirm that these two algorithms are effective.
In this study, the authors consider the parameter estimation problem of the response signal from a highly non-linear dynamical system. The step response experiment is taken for generating the measured data. Considerin...
详细信息
In this study, the authors consider the parameter estimation problem of the response signal from a highly non-linear dynamical system. The step response experiment is taken for generating the measured data. Considering the stochastic disturbance in the industrial process and using the gradient search, a multi-innovation stochasticgradientalgorithm is proposed through expanding the scalar innovation into an innovation vector in order to obtain more accurate parameter estimates. Furthermore, a hierarchical identification algorithm is derived by means of the decomposition technique and interaction estimation theory. Regarding to the coupled parameter problem between subsystems, the authors put forward the scheme of replacing the unknown parameters with their previous parameter estimates to realise the parameter estimation algorithm. Finally, several examples are provided to access and compare the behaviour of the proposed identification techniques.
暂无评论