Λεπτομέρειες βιβλιογραφικής εγγραφής
| Τίτλος: |
Iterative parameter identification with initial value optimization for colored noise Hammerstein systems under non–uniform sampling. |
| Συγγραφείς: |
Qin, Hongchen1 (AUTHOR), Ji, Yan1 (AUTHOR) yjichina@163.com |
| Πηγή: |
ISA Transactions. Aug2026, Vol. 175, p477-485. 9p. |
| Θεματικοί όροι: |
System identification, Mathematical optimization, Irregular sampling (Signal processing), Optimization algorithms, Signal filtering, Iterative methods (Mathematics), Noise, Nonlinear systems |
| Περίληψη: |
Given the presence of significant and possibly discontinuous nonlinearities on the input side, the widespread existence of colored noise in practical environments, and the prevalent use of non–uniform sampling strategies, the parameter identification problem for non-uniformly sampled piecewise nonlinear Hammerstein processes with colored noise becomes notably complex and challenging. This study employs a gradient iterative algorithm for parameter estimation. To suppress the interference of colored noise, a data filtering technique is introduced; to accelerate the convergence process, a momentum factor is incorporated into the gradient update. To further enhance the accuracy and efficiency of parameter identification, an equilibrium optimizer is embedded into the algorithmic framework to optimize the initial parameter values, thereby proposing the equilibrium–optimizer–based filtered momentum gradient iterative algorithm. Compared to the baseline algorithm without the optimizer, the proposed algorithm significantly improves the convergence speed and effectively avoids becoming trapped in local optima. Finally, numerical simulations and a wind power system case study are conducted to verify the effectiveness and superiority of the proposed method. • A novel identifiable modeling framework is established for non–uniformly sampled piecewise nonlinear Hammerstein systems under colored noise. By employing the key-term separation technique, the proposed model effectively decouples the nonlinear and linear subsystems, ensuring parameter identifiability in the presence of irregular sampling and correlated disturbances. • A filtered momentum gradient iterative (F–M–GI) algorithm is developed for parameter estimation. By integrating data filtering to suppress colored noise and a momentum mechanism to accelerate convergence, the proposed method significantly improves estimation accuracy and convergence speed compared with conventional gradient-based approaches. • An equilibrium optimizer (EO)-enhanced initialization strategy is proposed and seamlessly embedded into the iterative identification framework. By leveraging the strong global search capability of EO to determine high-quality initial values, the resulting EO–F–M–GI algorithm effectively alleviates the sensitivity to initialization, enhances global convergence performance, and avoids local optima. [ABSTRACT FROM AUTHOR] |
| Βάση Δεδομένων: |
Supplemental Index |