Winner-leading competitive swarm optimizer with dynamic Gaussian mutation for parameter extraction of solar photovoltaic models

• Published in: Energy conversion and management Vol. 206; p. 112450
• Main Authors: Xiong, Guojiang, Zhang, Jing, Shi, Dongyuan, Zhu, Lin, Yuan, Xufeng, Tan, Zhukui
• Format: Journal Article
• Language: English
• Published: Oxford Elsevier Ltd 15.02.2020; Elsevier Science Ltd
• Subjects: Algorithms; Competitive swarm optimizer; Exploration; Fault diagnosis; Gaussian mutation; Heuristic methods; Mathematical models; Maximum power tracking; Mutation; Optimization; Parameter extraction; Parameters; Particle swarm optimization; Photovoltaic; Photovoltaic cells; Photovoltaics; Systems analysis; Photovoltaic; Competitive swarm optimizer; Gaussian mutation; Parameter extraction
• ISSN: 0196-8904; 1879-2227
• DOI: 10.1016/j.enconman.2019.112450

[Display omitted] •An improved competitive swarm optimizer named WLCSODGM is proposed.•Winner-leading search is designed to help losers locate more promising regions.•Dynamic Gaussian mutation is integrated to enhance the exploration.•WLCSODGM is applied to parameter extraction of four different PV models.•Twelve state-of-the-art algorithms are employed to verify WLCSODGM. Extracting accurate and reliable values for involved unknown parameters of solar photovoltaic (PV) cells/modules is considerably significant to the characteristic analysis, fault diagnosis, maximum power point tracking, and efficiency evaluation of PV systems. Solving this problem using metaheuristic algorithms has gained increasing attention recently owing to their versatile and promising applications in highly nonlinear multimodal optimization problems. In this paper, an efficient and effective variant of competitive swarm optimizer (CSO) named WLCSODGM is presented to solve the parameter extraction problem of PV models. CSO is an advanced variant of particle swarm optimization and performs well especially on unimodal optimization problems. However, it is easily trapped in local optima when solving complex multimodal optimization problems such as the one considered here due to its poor exploration. In WLCSODGM, two improved components are introduced to remedy the inadequacy of CSO. On the one hand, a winner-leading search strategy is proposed to favor the exploration and help losers locate more promising regions. On the other hand, a dynamic Gaussian mutation operator with stretchable mutation amplitude and adaptive mutation probability is integrated to further enhance the exploration to help individuals jump out of poor local optima. WLCSODGM is applied to four different PV models and verified using a total number of twelve state-of-the-art algorithms. In addition, the influences of improved components and relevant algorithmic parameters are also experimentally evaluated. Results demonstrate that WLCSODGM is significantly better or highly competitive compared with the other algorithms.