A Novel CMA-ES with an Eigen Coordinates Framework for Parameter Identification in Photovoltaic Models
Xiaofei Wang, Hui Zhao, Zhenglei Wei, Yajun Liang, Yintong Li, Tong Han · 2019
The parameter estimation of a photovoltaic (PV) model that measures the change of the current-voltage dataset is a key issue for PV system applications. It is regarded as a complex multimodal optimization problem, and there have been several attempts to solve the problem in this field of research. To further address this problem in a faster and more accurate manner, we demonstrate an improved covariance matrix adaptive evolutionary strategy (CMA-ES) with novel modifications in an eigen coordinates framework named EC-CMA-ES. The original CMA-ES suffers from premature convergence and has poor exploration performance. Thus, we propose an eigenvalue adjustment strategy on a covariance matrix in order to drive evolution towards the dominant domain by adjusting its eigenvalues. Moreover, we perform a local search strategy in a later stage by utilizing the neglected inferior solutions to enrich the population diversity. We apply EC-CMA-ES to address the parameter identification of three commonly used PV models. The statistical results and comparisons with other state-of-the-art algorithms demonstrate the competitive performance of our modified CMA-ES in terms of efficiency and accuracy.