Deep learning model analysis for hybrid PV-TEG maximum power extraction and GMP estimation: A step toward cleaner energy generation

The hybrid of a photovoltaic-thermoelectric generating (PV-TEG) system offers great potential for solar energy harvesting, but its efficiency is limited by both sources components. This study proposes an evolutionary neural network (NN)-based maximum power point tracking (MPPT) techniques using an arithmetic optimization (AO) technique for enhancing the performance of the deep neural network (DNN) for real-time global maximum power (GMP) tracking. To further improve the control resilience, a Proportional-Integral (PI) regulator is employed, with gains optimized by the AO algorithm. Experimental results confirm the effectiveness of this approach, showing a median efficiency of 99.98%, minimal power oscillations, and a tracking time under 5 ms. Comparative analysis with other MPPT methods such as particle swarm optimizer (PSO), grasshopper optimization (GHO), and gray wolf optimization (GWO), validate the superior performance, stability, and real-time capabilities of the DNN-AO and AO-PI controllers for enhancing the PV-TEG system's energy generation.