A fast elitism Gaussian estimation of distribution algorithm and application for PID optimization.

Area of Science:

• Computational intelligence
• Optimization algorithms
• Probability statistics

Background:

• Estimation of Distribution Algorithm (EDA) is a probabilistic optimization method.
• Existing EDAs may face challenges with efficiency and convergence in complex problems.

Purpose of the Study:

• To propose a Fast Elitism Gaussian Estimation of Distribution Algorithm (FEGEDA).
• To enhance the efficiency and maintain convergent performance of EDA.
• To evaluate FEGEDA's capability in benchmark problems and PID controller optimization.

Main Methods:

• Utilizing a Gaussian probability model to represent solution distribution.
• Employing a fast learning rule for efficient parameter estimation.
• Incorporating an elitism strategy to preserve solution quality and convergence.
• Testing FEGEDA on 1D, 2D, and higher-dimensional benchmark functions.
• Applying FEGEDA to optimize Proportional-Integral-Derivative (PID) controllers for Permanent Magnet Synchronous Motors (PMSM).

Main Results:

• FEGEDA demonstrates superior performance compared to other algorithms and EDAs, particularly on higher-dimensional problems.
• Visualization of the optimization and probability model learning process is achieved using a 1D benchmark.
• FEGEDA shows competitive results when applied to PMSM PID controller optimization against classical-PID and Genetic Algorithm (GA).

Conclusions:

• FEGEDA is an effective and efficient intelligent optimization algorithm.
• The proposed fast learning rule and elitism strategy significantly improve performance.
• FEGEDA shows strong potential for real-world applications like advanced motor control.