A Solar PV Model Parameter Estimation Based on the Enhanced Self-Organization Maps

Abstract

Solar photovoltaic (PV) is a commonly utilized renewable energy source, with PV cells often being modeled as electric circuits. The identification of suitable circuit model parameters for PV cells is vital for performance evaluation, efficiency calculations, and the implementation of maximum power point tracking in solar PV systems. However, modeling the solar PV system is a nonlinear problem that requires an efficient algorithm. In this paper, we employ the enhanced self-organization maps (EASOM) to efficiently reduce the search space for parameter estimation in solar PV models. Our algorithm trains the SOM network on a subset of solutions, identifies the top solution's neural unit, generates a population of potential solutions, and selects the best candidate using a cost function, which represents the best PV model parameters obtained. The performance of EASOM is verified by extracting the parameters of the single diode (SDM) and double diode (DDM) models for the STM6-40/36 PV module. EASOM outperformed state-of-the-art algorithms with the lowest RMSE and MSE values of 8.3 mA and 6.87e-05 and achieved the lowest maximum error values of 27.37 mA and 20.52 mA, as well as low power error of 66.04 mW and 62.8 mW for SDM and DDM models.