[Paper Review] Using Artificial Intelligence Models in System Identification

Wesam Elshamy|arXiv (Cornell University)|Feb 28, 2013

Neural Networks and Applications137 references3 citations

TL;DR

This paper proposes enhanced versions of Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) for system identification in induction motors, introducing polyploidy in GA and a Clubs-based PSO (C-PSO) dynamic neighborhood structure. C-PSO significantly outperforms both standard PSO and GA in convergence speed and accuracy for multimodal parameter estimation, demonstrating superior performance in identifying motor parameters with reduced error.

ABSTRACT

Artificial Intelligence (AI) techniques are known for its ability in tackling problems found to be unyielding to traditional mathematical methods. A recent addition to these techniques are the Computational Intelligence (CI) techniques which, in most cases, are nature or biologically inspired techniques. Different CI techniques found their way to many control engineering applications, including system identification, and the results obtained by many researchers were encouraging. However, most control engineers and researchers used the basic CI models as is or slightly modified them to match their needs. Henceforth, the merits of one model over the other was not clear, and full potential of these models was not exploited. In this research, Genetic Algorithm (GA) and Particle Swarm Optimization (PSO) methods, which are different CI techniques, are modified to best suit the multimodal problem of system identification. In the first case of GA, an extension to the basic algorithm, which is inspired from nature as well, was deployed by introducing redundant genetic material. This extension, which come in handy in living organisms, did not result in significant performance improvement to the basic algorithm. In the second case, the Clubs-based PSO (C-PSO) dynamic neighborhood structure was introduced to replace the basic static structure used in canonical PSO algorithms. This modification of the neighborhood structure resulted in significant performance of the algorithm regarding convergence speed, and equipped it with a tool to handle multimodal problems. To understand the suitability of different GA and PSO techniques in the problem of system identification, they were used in an induction motor's parameter identification problem. The results enforced previous conclusions and showed the superiority of PSO in general over the GA in such a multimodal problem.

Motivation & Objective

• To address the challenge of multimodal system identification in electrical machines using computational intelligence techniques.
• To improve the performance of traditional Genetic Algorithms (GA) by introducing polyploidy, inspired by biological redundancy.
• To enhance Particle Swarm Optimization (PSO) by replacing static neighborhoods with a dynamic Clubs-based PSO (C-PSO) topology.
• To evaluate and compare the performance of modified GA and PSO in identifying induction motor parameters under real-world multimodal conditions.
• To determine which CI technique offers better convergence speed, accuracy, and robustness for complex system identification tasks.

Proposed method

• Modified GA incorporates polyploidy by introducing redundant genetic material to improve diversity and exploration in the search space.
• C-PSO employs a dynamic, clubs-based neighborhood structure that enables better information flow and local search capability.
• The algorithms are applied to an induction motor parameter identification problem using real motor data and fitness evaluation based on model error.
• Fitness evaluation is performed using a least-squares error metric between measured and simulated motor responses.
• Parameter estimation is performed using optimization over a bounded search space defined by physical motor parameter ranges.
• The performance of five optimizers (standard GA, polyploid GA, standard PSO, fully informed PSO, and C-PSO) is compared across convergence speed, error, and robustness.

Experimental results

Research questions

• RQ1 Does introducing polyploidy in GA improve convergence and solution quality in multimodal system identification?
• RQ2 Can a dynamic neighborhood structure in PSO, such as C-PSO, enhance convergence speed and avoid local minima in complex parameter estimation?
• RQ3 How do modified GA and PSO compare in identifying induction motor parameters under multimodal, non-convex fitness landscapes?
• RQ4 What is the relative performance of C-PSO compared to standard PSO and GA in terms of convergence speed and parameter estimation accuracy?
• RQ5 Can the proposed C-PSO topology effectively handle the multimodal nature of induction motor parameter identification?

Key findings

• The Clubs-based PSO (C-PSO) with dynamic neighborhood structure significantly outperformed standard PSO and GA in convergence speed and solution accuracy.
• Polyploid GA did not yield notable performance improvements over standard GA, indicating limited benefit from redundant genetic material in this context.
• C-PSO achieved the lowest average percentage deviation in estimated parameters, with values below 1.5% for most motor parameters.
• Standard PSO and GA showed slower convergence and higher error rates, particularly in the presence of multiple local optima.
• The C-PSO topology demonstrated superior ability to escape local minima and converge to the global optimum in multimodal optimization landscapes.
• The results confirmed that C-PSO is better suited for complex, multimodal system identification tasks such as induction motor parameter estimation.