Identification Of Linear-in-parameters Systems Based On The Filtering

System identification and parameter estimation are essential for model-based control problems. As the control systems upgrade and transform, the mathematical model of systems becomes more complex and the computation of the identification processing is also growing. To deal with the problem of linear-in-parameters systems identification in the presence of colored noise and reduce the amount of the algorithm calculation,this dissertation entitles “ Identification of Linear-in-Parameters Systems Based on the Filtering”, and it has important theoretical significance and practical value. The main contribution of this paper is showed as follows.1. For a class of linear parameter systems with autoregressive moving average noise,a linear filter is developed to simplify the identification process by obtaining white noise interference systems. Due to the unknown intermediate variables and the noise terms in parameter vectors, using the idea of auxiliary model identification to replace the unestimated. A stochastic gradient(SG) algorithm and a recursive least squares(RLS) algorithm are proposed using the idea of auxiliary model and the data filtering technique. To improve the convergence speed, a mlti-innovation SG algorithm is derived by extending the single innovation vector and extracting useful information.Also, the paper discusses a least squares based iterative and a gradient based iterative,resulting in a high utilization of the measured data.2. Using the hierarchical identification principle, the linear-in-parameters systems are decomposed into several subsystems, which are identified by the interactive estimation method, respectively. This paper presentes a multi–innovation SG algorithm, an RLS algorithm, a gradient and a least squares based iterative algorithms by using the filtering and decomposition technique, and provides the comparison for several key algorithms computational amount. The results show that decomposition technique is capable of reducing the dimension of the parameter vectors, lowering the complexity of the algorithm and then reducing the computational burden. At last, some simulation examples are given to compare the parameter estimation accuracy between the proposed algorithms.In summary, this paper derives the recursive algorithms and the iterative algorithms for a class of linear parameter systems with colored noise interference. The numerical simulation results indicate the proposed algorithms are effective. Finally,this paper introductes the unresolved problems encountered in the course of the study and makes a brief outlook on the development of system identification direction.