Identification of Hammerstein-Wiener time delay model based on approximate least absolute deviation

Nonlinearity, time delay and spike noises widely exist in industrial processes. Compared with the linear model, the typical nonlinear Hammerstein-Wiener (H-W) model can describe nonlinear characteristics of industry processes more accurately. In order to overcome the effect of spike noise on the identification results, we propose a stochastic gradient algorithm based on the least absolute deviation in this paper. To solve the non-differentiable problem of the least absolute deviation, an approximate least absolute deviation objective function is established by introducing a deterministic differentiable function to replace the absolute residual. Experiments show the proposed algorithm can suppress the influence of the spike noise on the identification results, and has high identification accuracy and strong robustness.