105 samples of soybean oil were prepared by adding phosphatidic acid to first-grade soybean oil， and the content of phosphatidic acid was determined. The terahertz spectrum of the sample was collected. Compared with the reference signal， the Terahertz （THz） wave has different degrees of delay in time after passing through the sample， and the terahertz wave amplitude has decreased to varying degrees. The presence of phosphatidic acid increases the absorption of terahertz waves in soybean oil samples. The higher the content of phosphatidic acid in the sample， the lower the frequency-domain spectral amplitude. There is a more obvious characteristic absorption peak at 1.58 THz. Qualitative analysis of phosphatidic acid in samples can be performed using terahertz spectroscopy. The first and second derivatives were used to pre-process the absorption spectra of the 91 samples remaining after excluding abnormal samples. As input variables， a mathematical model was established using Partial Least Squares （PLS）. The prediction coefficient（R2） of the prediction set is 0.9980， the root mean square error of prediction （RMSEP） of the prediction set is 0.0813， and the relative standard deviation （RSD） is 8.16%， which can accurately predict the content of phosphatidic acid in soybean oil.