The photonic crystal structure containing surface defect with porous silicon is proposed, in which the defect cavity on the surface is established by introducing the porous silicon layer and the absorbing medium layer, and the sensing region of the sample detected is formed by the use of the efficient carrying mechanism of the porous silicon. Because of the absorption of ZnS, the light corresponding to the resonant wavelength is absorbed and the defect peak is obtained in the reflection spectrum. The back propagation neural network is adopted to optimize the thickness of porous silicon globally. The relationship model between the concentration of the sample detected and the defect peak wavelength is established according to the Goos-Hanchen shift and the sensing performance is analyzed. The simulation results show that the reflectivity of the defect peak decreases from 31.23% to 0.00129% and the Q value can attain to 1537.37 after the optimal design of structural parameter. The sensitivity of the sensor structure is about 2.5 nm at per 1% mass fraction, which can provide effective theory guidance for the detection of the concentration and composition of samples.