Abstract: Objective High-reflectivity metal components tend to suffer from penetration fluctuation, lack of fusion, and burn-through defects during laser welding. Therefore, online monitoring of weld penetration is a key approach to achieving stable welding quality. Methods This paper proposes an optical coherence tomography (OCT) based penetration monitoring method for power battery busbars. The effects of welding parameters on weld formation and OCT measurement are studied, and welding parameters that produce defect-free and well-formed welds are obtained. Results It is clarified that, with increasing welding speed, the OCT signal shows a measurement error that changes from underestimation to overestimation. A back propagation (BP) neural network model incorporating welding parameters is established to predict penetration depth under different welding parameters, and an average prediction error is within 0.1 mm. Conclusion The results show that combining low-frequency OCT signals with welding parameters that have practical physical significance enables high-accuracy online weld penetration monitoring.