Using the German PEN 3.5 electronic nose, suitable sensor array combinations were screened, principal components were extracted by PCA analysis, and the mass concentration of reduced sulfides was determined by GC/MS. Response surface analysis was used to establish four mass concentration prediction models for reduced sulfides. On this basis, a odor concentration prediction model was established by applying the partial least squares method to the sensor response values of the electronic nose to different concentrations of mixed sulfides and the corresponding measured odor concentration values. The model was validated using actual samples and achieved good prediction results. Compared with other analytical techniques, electronic noses can more portable and quickly predict material mass concentration and odor concentration, and can meet real-time online monitoring requirements, providing technical support for odor pollution management and subsequent treatment system development.