Measurement-Based Modeling of a Semitransparent CdTe Thin-Film PV Module Based on a Custom Neural Network

Semitransparent photovoltaic (STPV) can be employed in a wide application range to provide sunlight permeability for supplying solar electrical energy with some shading, which is preferable in hot areas. To predict the output power and formulate the performance of this type of photovoltaic (PV) system, the proposed approach analyzes a Thin-Film solar cadmium telluride-type module and develops a custom neural network (CNN) for modeling its generated power expressed by its mathematical formula. Experiments for single and multilayer installation topologies are conducted for performance analysis. The coefficients of the model equation are investigated based on a set of power-current curves. The developed model adopts three factors: a minimum number of hidden neurons, the use of all measured data to train the network weights, and a linear output activation function to reduce the complexity of solving the network equations. The results specify the limit at which this type of PV starts generating power from the experimental measurements and the comparison with its equivalent normal PV module. The CNN-based STPV module is verified by comparing with the experimental measurements results, which shows a reasonable R-square, while its performance is evaluated on the silicon-based PV by comparing its behavior with the two-diode model PV in the MATLAB-based simulation.