偏差校正最小二乘法参数辩识的分析和仿真

本文旨在分析偏差校正法与其他参数辩识方法的关系。在前面部分，借助移位向量把二次准则表达成待辩识参数的显函数。这样广义最小二乘法，偏差校正法及伪线性法的离线算法便很容易地以这同一准则导出。后面部分从递推预报误差法导出以上三种方法的递推算法。通过以上分析并采用仿真手段清楚展现了各种算法间的相互关系。

Analysis and Simulation of BCLS for Parameter Identification

The purpose of this paper is to exhibit relationships between BCLS and other identification algorithms. In the first part a quadratic criterion is expressed as an explicit function of unknown parameters through the agency of a shifting yector. Off-line algorithms of GLS, BCLS and PLR can be derived easily from the same criterion. In the second part, recursive algorthms of the above three methods are derived from the RPEM by different approximations. Relations among these algorithms are distinctly exhibited by the above analyses and simu-lation.