As global climate change intensifies, the scale of water conservancy and hydropower projects continues to expand in arid, semi-arid, and plateau regions. However, these areas exhibit persistently low humidity and significant diurnal temperature variations, posing substantial challenges to the durability of hydraulic concrete structures. This study investigates the influence of pore structure on the freeze-thaw durability of hydraulic concrete by setting different curing humidity levels (20 %RH, 45 %RH, 70 %RH, 95 %RH) and establishes a quantitative correlation. Quantitative capture of changes in porosity and pore size distribution using nuclear magnetic resonance (NMR) technology; Combined with scanning electron microscopy (SEM), this study reveals the microstructural mechanisms underlying the effects of curing humidity and freeze-thaw (F-T) cycles on the freeze-thaw durability of hydraulic concrete. The gray relational entropy analysis (GREA) method was employed to identify the key pore size range that influences the relative dynamic modulus of elasticity (RDME), and a pore size-RDME regression model was established based on this analysis. The results indicate that reduced curing humidity impedes the hydration process, leading to an altered pore structure characterized by a decrease in the volume of micropores and mesopores, and an increase in the proportion of macropores and microcracks, which in turn elevates the population of harmful pores. The number of F-T cycles exhibits a significant positive correlation with porosity. As the number of F-T cycles increases, both the peak height and peak area of the T2 spectrum show a marked upward trend. The number of cracks within the concrete increases significantly, the mass loss rate rises, and the RDME decreases markedly. Among the four humidity conditions, the pore size distribution showing the strongest correlation with RDME is mesopores, followed by micropores. The correlation coefficient (R2) of the pore size-RDME model ranged from a maximum of 0.9886 to a minimum of 0.9327, with relative errors consistently below 5 %. The maximum residual was 0.01754, indicating a high-quality fit. This study quantifies the influence of pore structure on RDME, providing a reference for the design and optimization of concrete under different curing humidity conditions.