A Modified Artificial Protozoa Optimizer for Robust Parameter Identification in Nonlinear Dynamic Systems

Area of Science:

• Computational intelligence
• Optimization algorithms
• Nonlinear dynamics

Background:

• Accurate parameter identification in nonlinear and chaotic systems is crucial for modeling and control.
• Existing optimization algorithms struggle to balance global exploration and local refinement in complex, multimodal landscapes.
• Challenges include high dimensionality and the need for robust solutions in dynamic environments.

Purpose of the Study:

• To develop a modified artificial protozoa optimizer (mAPO) for enhanced parameter identification.
• To improve global search capability and local refinement in optimization.
• To validate the performance of mAPO on benchmark functions and in real-world nonlinear system identification tasks.

Main Methods:

• Developed mAPO by integrating a probabilistic random learning strategy and a Nelder-Mead simplex-based local refinement stage into the original artificial protozoa optimizer (APO).
• Evaluated mAPO using the CEC2017 benchmark suite, including shifted/rotated, hybrid, and composition functions.
• Applied mAPO to parameter identification of the Rössler chaotic system and a permanent magnet synchronous motor (PMSM) under static and dynamic conditions.

Main Results:

• mAPO demonstrated improved mean performance and reduced variability on the CEC2017 benchmark suite, indicating enhanced robustness.
• In nonlinear system identification, mAPO achieved smaller objective function values, more accurate parameter estimates, and superior statistical stability compared to APO and other state-of-the-art optimizers.
• Exact parameter reconstruction with zero error was achieved for the PMSM, with rapid and smooth convergence observed.

Conclusions:

• The proposed mAPO effectively balances global exploration and local exploitation for accurate parameter identification in complex nonlinear and chaotic systems.
• mAPO offers enhanced robustness and scalability, outperforming existing methods on benchmark functions and practical dynamic system identification.
• The algorithm shows significant potential for applications requiring precise parameter estimation in dynamic and time-varying environments.