MLS

  当信噪比较大时，采用广义最小二乘法可能会出现多个局部收敛点，解决这个问题的方法可用多级最小二乘法，一般来说多级最小二乘法包含三级辨识过程。利用输入输出数据，通过多级最小二乘法，可分别求的辅助模型，过程模型和噪声模型的参数估计值。在高噪声的情况下，多级最小二乘法明显优于广义最小二乘法，其收敛点唯一。(When the signal noise ratio (SNR), and by use of the generalized least squares method may appear more local convergence point, the way of solving this problem can be used multistage least squares method, and generally multistage least squares method includes three identification process. Using input/output data, through the multistage least squares, respectively for the auxiliary model, process model and noise model parameter estimation. In the high noise, multistage least squares method is obviously superior to the generalized least squares method, and its convergence point only.)