Genetic algorithm-based adaptive weighted fuzzy logic control (awFLC) for traction power control
| Autor: | Dursun Ekmekci, Shahnaz N. Shahbazova |
|---|---|
| Rok vydání: | 2022 |
| Předmět: | |
| Zdroj: | Journal of Intelligent & Fuzzy Systems. 43:6909-6916 |
| ISSN: | 1875-8967 1064-1246 |
| DOI: | 10.3233/jifs-220753 |
| Popis: | One of the most important issues for FLC systems is the problem of finding the right balance between interpretability and accuracy. For this delicate balance, several methods which can be integrated into fuzzy logic, and tune the fuzzy logic parameters adaptively, have been proposed. One of these popular approaches is the heuristic optimization method. However, in terms of optimization, designing fuzzy logic control is a complex optimization problem that is discrete in terms of rule optimization and numerical in terms of optimization of membership degrees parameters. In this context, heuristic-based adaptive fuzzy control systems focus on either fuzzy rule optimization, weighting fuzzy rules, or parameter optimization. In this paper, unlike the others, an adaptive weighted fuzzy logic control (awFLC) method, which weights the inputs instead of the rules, is proposed. First, the membership degree of each input is calculated. Then, the resultant weight is determined by combining the weighted input membership degrees. For a crisp result, the average of the membership degrees of the resultant weight to the output membership functions is calculated. In awFLC, the interaction between membership functions is achieved by average membership degree, communication between inputs is achieved by the weighting of inputs, and mapping between inputs-outputs is achieved by the resultant weight value. Thus, the approach, which turns into a purely numerical optimization problem, provides convenience for heuristic search. In awFLC, optimal values for input weights and variable parameters are searched by the genetic algorithm. The performance of the method was tested on traction power control, and the results were compared with the ANFIS results. With awFLC, an 8.13% average error was obtained, while ANFIS produced solutions with an average error rate of 8.97%. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|