Sub-population improved grey wolf optimizer with Gaussian mutation and Lévy flight for parameters identification of photovoltaic models

Abstract

Accurate modeling of photovoltaic (PV) modules is key to design, simulate and control photovoltaic power systems. Due to insufficient data in the photovoltaic cell data sheets, PV models are represented using nonlinear current–voltage I-V characteristic curve behaviors with a large number of unknown parameters. Therefore, it is crucial to identify the parameters of the PV models accurately and reliably. Because of their nonlinear, multivariate and multimodal characteristics, most existing methods face the problem of falling into local optima and converging prematurely. To identify parameters of the PV models more accurately, a Sub-Population improved Grey Wolf Optimizer with Gaussian mutation and Lévy flight (SPGWO) is proposed in this work. In SPGWO, the population is equally divided into superior and inferior subpopulations based on the performance of individuals. These two subpopulations adopt different search strategies to update the position of individuals. The covariance matrix C of the superior subpopulation is used as a Gaussian mutation operator to guide the next generation by mutation perturbation of elite wolves α, β and δ, which can effectively improve the local exploration ability of SPGWO. To prevent premature convergence of the algorithm due to individual similarity, Grey Wolf Optimizer (GWO) is used to explore potential candidates in the inferior subpopulation to increase population diversity. Considering the local stagnation of the entire population, an improved Lévy flight (LF) strategy is introduced to facilitate the entire population stepping outside the local optimum. The proposed SPGWO is used to solve the parameter identification problem for five different PV models. Statistical results and analysis show that SPGWO has outstanding accuracy and reliability.