<p>In this study, a buck-boost converter circuit based on machine learning was developed through real-time simulation, employing field measurements collected from traction center regions integrated with DC-fed rail systems operating at different voltage levels. By utilizing the proposed topology with Matern 5/2 Gaussian Process Regression (GPR), high performance was achieved in the power supply system. The best R-squared score was obtained under operating conditions in DC-powered traction centers. The performance metrics obtained with the proposed model are as follows: the R² value is 1, the RMSE is 0.00009, and the MSE is 9.0612e-09. These results demonstrate the effectiveness of artificial intelligence-based approaches in enhancing the efficiency of the buck-boost converter. The same dataset was applied to other machine learning methods known for producing successful outcomes, and their performance rates were compared with those of the proposed method. This circuit enables power conversion within real-time integration zones by employing highly efficient control signals that are generated through the application of machine learning to actual field data. At the conclusion of the study, the output voltage of the buck-boost converter, including a 750&#xa0;V DC supply, was presented with statistical values to demonstrate the successful operation of the proposed model.</p>