Recurrent neural network-based electrical modeling and simulation of solar photovoltaic cell

Abstract

Photovoltaics (PV) are an important renewable energy (RE) source and a key component of solar power plants. Accurate modeling of their electrical output characteristics is crucial for efficient system design, such as the current-voltage (I-V) and power-voltage (P-V) curves. These characteristic curves depend on the photovoltaic cell (PVC) parameters such as short circuit current (ISC), open circuit voltage (VOC), and maximum power (PMAX). Previous studies have explored the artificial neural network (ANN) in photovoltaics generation (PVG) systems to estimate the output power of the PVC. Nevertheless, most ANN-based approaches are limited to estimating output power only from the daily power generation data of the PVG system. In this paper, a PVC model was developed using a recurrent neural network (RNN) to increase the exactness estimation of electrical parameters. The proposed RNN-based PVC model employs solar irradiance (S), temperature (T), and series resistance (RS) as input variables, whereas the output variables are the power and current of PVC. The contributions of the PVC model is evaluated through simulations under changing physical and environmental conditions. The simulation yield of the RNN-based PVC model successfully reproduces I-V and P-V curves and also fundamental parameters consistent with the behavior of real PVC.