Application of meta-heuristic algorithms to system identification problems

In recent years, there has been a significant increase in the number of meta-heuristic optimization algorithms used to solve various optimization problems. These algorithms are designed by taking inspiration from biological evolution, swarm behavior, plant growth processes, and other phenomena. While there are a wide variety of algorithms, genetic algorithms and particle swarm optimization algorithms are quite popular. Meta-heuristic algorithms are effectively used in a wide range of applications such as hyperparameter optimization in machine learning, controller design, finance, and more. Literature research has identified important shortcomings in the application of meta-heuristic algorithms to system identification problems. System identification methods aim to determine the mathematical model of a system using its input and output data. Obtaining the mathematical model of a system can often be a tedious and complex process. However, this process can be overcome by using system identification methods based on the analysis of input-output data. In this way, a model that can be used to understand and optimize the behavior of the system can be obtained. In this study, various meta heuristic algorithms were employed, including Artificial Ecosystem Optimization (AEO), Flower Pollination Algorithm (FPA), Ant Lion Optimizer (ALO), Moth Flame Optimization (MFO), Tug of War Optimization (TWO), Atom Search Optimization (ASO), Brain Storm Optimization (BSO), Water Cycle Algorithm (WCA), Coral Reefs Optimization (CRO), and Life Choice-Based Optimization Algorithm (LCO). These algorithms were applied to obtain the model of the system using the input/output data obtained from a hair dryer experiment. Factors such as time, maximum generation, early termination, and function computation limitations were considered, and the performance of the algorithms was examined. When comparing the performance of the algorithms, the AEO algorithm exhibited higher performance compared to other algorithms. The conducted analyses revealed that the proposed meta-heuristic algorithms can be easily and successfully applied to system identification problems.