Supervised Machine Learning Techniques in the Improvement of Performance of Solar PV Systems due to Partial Shading

This paper aims to propose a method for using artificial neural networks to enhance the efficiency of solar PV systems under partially shaded conditions. It examines three system sizes ($4 \times 4,5 \times 5$, and $6 \times 4$) allows for flexibility in design, catering to varying energy needs and space constraints and various connection methods: These include Series-parallel (SP), Total-cross-tied (TCT), Bridge-link (BL), Honeycomb (HC), and SP-TCT, BL-TCT or HC-TCT by using MATLAB/SIMULINK. Analyze training state plots, performance plots, error histograms, and regression plots to assess these configurations. The findings revealed that integrating neural networks improves the performance of solar PV systems in shaded conditions, mainly through improved decision-making about how to complement energy derived from panels. Combinations with TCT and other traditional techniques yielded the highest results among all company sizes. From the training state plot, it is evident that the neural network model is stable, and converging very well from the performance plot showing good prediction. Working with the error histogram and observing only minor deviations in the predicted output; the regression plot shows a strong, positive correlation between the predicted and the actual values.