Open Access

ARTICLE

Training Neuro-Fuzzy by Using Meta-Heuristic Algorithms for MPPT

by Ceren Baştemur Kaya¹, Ebubekir Kaya^2,*, Göksel Gökkuş³

1 Department of Computer Technologies, Nevsehir Vocational College, Nevsehir Haci Bektas Veli University, Nevşehir, 50300, Turkey
2 Department of Computer Engineering, Faculty of Engineering and Architecture, Nevsehir Haci Bektas Veli University, Nevşehir, 50300, Turkey
3 Department of Electrical and Electronics Engineering, Faculty of Engineering and Architecture, Nevsehir Haci Bektas Veli University, Nevşehir, 50300, Turkey

* Corresponding Author: Ebubekir Kaya. Email:

Computer Systems Science and Engineering 2023, 45(1), 69-84. https://doi.org/10.32604/csse.2023.030598

Received 29 March 2022; Accepted 26 May 2022; Issue published 16 August 2022

Abstract

It is one of the topics that have been studied extensively on maximum power point tracking (MPPT) recently. Traditional or soft computing methods are used for MPPT. Since soft computing approaches are more effective than traditional approaches, studies on MPPT have shifted in this direction. This study aims comparison of performance of seven meta-heuristic training algorithms in the neuro-fuzzy training for MPPT. The meta-heuristic training algorithms used are particle swarm optimization (PSO), harmony search (HS), cuckoo search (CS), artificial bee colony (ABC) algorithm, bee algorithm (BA), differential evolution (DE) and flower pollination algorithm (FPA). The antecedent and conclusion parameters of neuro-fuzzy are determined by these algorithms. The data of a 250 W photovoltaic (PV) is used in the applications. For effective MPPT, different neuro-fuzzy structures, different membership functions and different control parameter values are evaluated in detail. Related training algorithms are compared in terms of solution quality and convergence speed. The strengths and weaknesses of these algorithms are revealed. It is seen that the type and number of membership function, colony size, number of generations affect the solution quality and convergence speed of the training algorithms. As a result, it has been observed that CS and ABC algorithm are more effective than other algorithms in terms of solution quality and convergence in solving the related problem.

---

Keywords

---