Box-Jenkins model parameter estimation algorithm based on singular value decomposition

parameter estimation;  singular value decomposition;  Progressive generalization of the least squares method;  Box-Jenkins model

Based on the principle of singular value decomposition and incremental generalization of the least squares principle, a recursive algorithm for parameter estimation of Box-Jenkins model is proposed. The conventional generalized augmented least squares algorithm is sensitive to rounding error, which will cause the covariance matrix to lose positive characterization and symmetry, resulting in the problem of pathological conditions and numerical instability. In order to improve the performance of parameter estimation, the Box-Jenkins model estimation algorithm is derived by using the singular value decomposition technique of covariance matrix. The algorithm has high recognition accuracy, fast convergence speed and good numerical stability. The simulation shows that with the increase of the noise-signal ratio. The new algorithm still has good performance.