Background <p>Magnetic spring-based energy harvesting systems offer potential for converting low-frequency ambient vibrations into usable electrical energy. The dynamic behavior of such systems can be enhanced by utilizing nonlinear restoring forces and advanced prediction techniques.</p> Objective <p>This study aims to investigate the performance and energy-harvesting potential of a novel Single-Degree-of-Freedom (SDOF) magnetic spring oscillator-based linear electromagnetic generator.</p> Methods <p>A hybrid approach combining analytical modeling, numerical simulation, and experimental validation is employed. The system features a nonlinear magnetic oscillator with a floating magnet between two fixed magnets on a nonmagnetic shaft. Machine learning, specifically neural networks, is used to predict the magnetic restoring force. Both linear and nonlinear stiffness effects are analyzed, and a custom-built test rig is used for experimental validation.</p> Results <p>The proposed system demonstrates a broad operational frequency bandwidth and superior performance at low frequencies compared to conventional magnetic spring harvesters. The model predictions align closely with experimental results, confirming the efficacy of the nonlinear oscillator and the force prediction method.</p> Conclusion <p>The developed SDOF magnetic spring oscillator exhibits promising capabilities for low-frequency energy harvesting. Its broadband response, enhanced by nonlinear magnetic repulsion and machine learning-based modeling, makes it suitable for applications in environments where efficient energy conversion at low frequencies is essential.</p>