Application of Artificial Neural Networks to Predict and Optimize the Performance of a Segmented Photovoltaic-Thermoelectric System

Power generation from the Sun is dominated by photovoltaics (PVs) whose efficiencies are severely limited by overheating due to the underutilization of the solar spectrum. Thermoelectric generators (TEGs) are used to reduce the overheating of PVs by forming hybrid PV-TEs capable of utilizing the broad solar spectrum. Nevertheless, the PV-TE is a developing technology, hence further enhancements in the device efficiency are urgently needed to make them as competitive as their non-renewable energy counterparts. We propose a segmented TEG (STEG) to replace the traditional TEG in a PV-TE to provide better cooling in a PV. Furthermore, an artificial neural network (ANN) is introduced as a replacement to the time consuming traditional finite element methods (FEMs) used to optimize the PV-TE performance. Finally, the pitfalls of the pocket papers on PV-STEGs are addressed. Results show that the PV-STEG is 97% more efficient than the traditional PV-TEG and that the ANN is 90.44% faster than the FEM in optimizing the PV-STEG performance