Evolutionary multi-task optimization for parameters extraction of photovoltaic models

• Published in: Energy conversion and management 2020-03, Vol.207, p.112509, Article 112509
• Main Authors: Liang, Jing, Qiao, Kangjia, Yuan, Minghua, Yu, Kunjie, Qu, Boyang, Ge, Shilei, Li, Yaxin, Chen, Guanlin
• Format: Article
• Language: English
• Subjects: Algorithms; Crossovers; Differential evolution; Evolutionary algorithms; Evolutionary multi-task optimization; Human performance; Mathematical models; Optimization; Parameter extraction; Parameter identification; Performance enhancement; Photovoltaic cells; Photovoltaic models; Photovoltaics; Solar energy
• ISSN: 0196-8904; 1879-2227
• DOI: 10.1016/j.enconman.2020.112509

•Multitask optimization is developed to extract Photovoltaic models parameters.•Online local similarity is measured to guide the knowledge transfer.•The helpful knowledge is used to enhance the algorithm performance.•Results indicate the superior performance of the proposed method. As the demand for solar energy increases dramatically, the optimization and control of photovoltaic systems become increasingly important, accurate and reliable parameter identification of photovoltaic models is always required, which proposes an urgent need for accurate and robust algorithms. To this end, many heuristic algorithms have been proposed to extract the parameters of different photovoltaic models. However, they only extract the parameters of one model in a single run, which is inconsistent with the human ability to solve multiple tasks simultaneously and ignores the useful information derived from different models. Therefore, in this paper an evolutionary multi-task optimization algorithm is proposed to extract the parameters of multiple different photovoltaic models simultaneously. To be specific, the helpful information found by the population is transferred through the cross-task crossover to improve the performance in terms of solution quality and convergence rate of the population. The proposed algorithm is evaluated by extracting the parameters of three different models simultaneously, i.e., single diode, double diode, and photovoltaic module model. Comprehensive results demonstrate that the proposed algorithm has better performance with respect to the accuracy and robustness in comparison with other state-of-the-art algorithms.