Identification of Hammerstein FIR Systems Based on Attenuated Maximum Stacking Length Multi‐Error Algorithm

ABSTRACT A modified multi‐error information gradient parameter estimation algorithm is suggested to identify a Hammerstein nonlinear system with FIR linearity. While traditional information gradient algorithms are less computationally intensive, they exhibit slow convergence rates. The new method employs stacked gradients instead of a single gradient, integrating the multi‐error concept to enhance speed. However, determining the appropriate stacking length for multi‐error algorithms presents new challenges. To resolve this, the strong Wolfe criterion is used. This criterion's selection of stacking length can guarantee that the cost function declines at a reasonable pace. The preset maximum stacking length is substituted with a stacking length sequence that attenuates by the inverse cotangent function to produce precise estimates. The suggested algorithm was validated by real case study and numerical experimentation.