Irreducible water saturation is a key parameter for reservoir fluid property identification and reservoir evaluation. Using geophysical logging data to evaluate reservoir irreducible water saturation is a commonly used method. Nuclear magnetic resonance logging generally calculates the irreducible water saturation of the reservoir by determining the cutoff value of the transverse relaxation time. In the nuclear magnetic resonance experimental study of the M layer group in H oilfield in the Middle East, it is found that some experimental samples have the phenomenon of ' diffusion coupling ', which makes it difficult to accurately calculate the irreducible water saturation of the reservoir through cutoff. By extracting the characteristic parameters reflecting the pore structure and fluid occurrence state information in the nuclear magnetic resonance spectrum, and combining with the data of casting thin sections, the correlation between these parameters and irreducible water saturation is discussed, and the cut-off is avoided to avoid the influence of ' diffusion coupling ' on the model. The experimental results show that there is a correlation between irreducible water saturation and nuclear magnetic porosity, interval pore component and geometric mean. Based on the above parameters, a prediction model of irreducible water saturation is established. The prediction model is verified by actual wells in the M layer group of H oilfield in the Middle East, and good application results are obtained, which lays a solid foundation for the subsequent identification of reservoir fluid properties and water flooded layer identification.