A comprehensively theoretical analysis on the broadband spectral energy distributions (SEDs) of large-scale jet knots in 3C 273 is presented for revealing their X-ray radiation mechanism. We show that these SEDs cannot be explained with a single electron population model when the Doppler boosting effect is either considered or not. By adding a more energetic electron (the leptonic model) or proton (the hadronic model) population, the SEDs of all knots are well represented. In the leptonic model, the electron population that contributes the X-ray emission is more energetic than the one responsible for the radio-optical emission by almost two orders of magnitude; the derived equipartition magnetic field strengths (B_eq) are ~0.1 mG. In the hadronic model, the protons with energy of ~20 PeV are required to interpret the observed X-rays; the B_eq values are several mG, larger than that in the leptonic model. Based on the fact that no resolved substructures are observed in these knots and the fast cooling-time of the high-energy electrons is difficult to explain the observed X-ray morphologies, we argue that two distinct electron populations accelerated in these knots are unreasonable and their X-ray emission would be attributed to the proton synchrotron radiation accelerated in these knots. In case of these knots have relativistic motion towards the observer, the super-Eddington issue of the hadronic model could be avoided. Multiwavelength polarimetry and the gamma-ray observations with high resolution may be helpful to discriminate these models.