Accurate estimation of unknown parameters of complex photovoltaic models is crucial to whether photovoltaic generators can efficiently convert energy. When a photovoltaic model has multiple diode branches, its complexity increases geometrically. To address the problems of high complexity and difficulty in estimation, this paper proposes an effective improved algorithm based on success-history adaptation differential evolution with linear population size reduction (L-SHADE)—Bi-parameter coordinated updating L-SHADE with parameter decomposition method (CSpL-SHADED). First in CSpL-SHADED, dynamic crossover rate ranking technology is developed to bridge the relationship between individuals and crossover rates, thereby improving effective mutation capabilities. In addition, a dynamic sub-population mechanism is also proposed to divide the entire population into multiple sub-populations so that they are evenly distributed in the search space, so the search ability of individuals in local areas is improved. Secondly, the unknown parameters of solar photovoltaic models of different complexity are effectively decomposed into linear and nonlinear parameters by the decomposition method. The nonlinear parameters are accurately estimated by CSpL-SHADED, and the linear parameters are calculated based on the constructed matrix equation. Through experiments on four solar photovoltaic models of different complexity, CSpL-SHADED showed strong competitiveness to varying degrees compared to the comparative algorithms.
Read full abstract