Identification Method Based on Improved Hierarchical Multi-innovation Stochastic Gradient Algorithm

An improved hierarchical multi-innovation stochastic gradient algorithm was proposed for the parameter identification problem of input nonlinear closed-loop control autoregressive systems. In the existing literature, a hierarchical multi-innovation stochastic gradient algorithm was used to identify model parameters, but there are some problems such as slow convergence speed and low identification accuracy. In order to deal with these issues, the paper proposed an algorithm that combines the hierarchical multi-innovation stochastic gradient algorithm with momentum methods. Among them, the momentum method can help the stochastic gradient accelerate and suppress the oscillation in the related direction. Simulation results showed that the algorithm is effective.