Phenolic compounds are widely used in metallurgy, oil refining, machinery manufacturing, medicine, pesticide and paint industries, but they are toxic, if not treated properly they will pollute environment. Water is the source of life, and the detection of phenols in the water environment is particularly important. Three- dimensional fluorescence spectrometry has the characteristics of highly sensitivity, fast detection speed, convenient pretreatment and tracing detection. The second-order correction method can identify the interesting components in compounds. In this paper, three-dimensional fluorescence spectroscopy will be combined with second-order correction method to test phenols in the water. M-cresol and resorcinol were selected as the tested substances in this experiment, and they were divided into two kinds of samples: adding interference and without interference. The data of three-dimensional fluorescence spectra of eight corrected samples and eight predicted samples were measured by FLS920 steady-state fluorescence spectrometer, and the data above were preprocessed: scattering interference contained in the original spectrum was removed; corrected by the excitation/emission correction. Then, the spectral data were compressed by the wavelet packet generated by db3 wavelet function, and the redundant information in the spectral data was removed through this approach. The compression score achieved 91.67%, and the recovery score achieved 96.62%. Then two second-order calibration methods: parallel factor analysis (PARAFAC) and self-weighted alternating trilinear decomposition (SWATLD) were used to analyse the preprocessed data qualitatively and quantitatively separately. According to the results of consistency analysis combined with residual discriminant analysis, the component number of samples without interference was chose as 2, and the samples with interference was chosen as 3. Qualitative analysis showed that regardless of the existing or inexisting interference, these two second-order calibration methods all could identify m-cresol and resorcinol in samples accurately. The fluorescence peak position of m-cresol and resorcinol were located at λem=298 nm/λex=274 nm and λem=304 nm/λex=275 nm separately. The quantitative analysis results show that the average recovery rate of m-cresol and resorcinol reached 93.37%±4.92% and 95.19%±5.25% respectively by PARAFAC without adding interference and under interference, meanwhile the average recovery rate of m-cresol and resorcinol were 92.09%±2.64% and 97.08%±5.28%. Under the same conditions, when we chose SWATLD , the average recovery rate of m-cresol and resorcinol reached 93.11%±4.73% and 96.80%±5.04% respectively, meanwhile the average recovery rate of m-cresol and resorcinol were 97.30%±4.52% and 96.92%±5.61% respectively. The root mean square error of prediction (RMSEP) of the two methods are all less than 0.03 mg·L-1. The experimental results show that two second-order calibration algorithms: PARAFAC and SWATLD can all quickly and accurately test phenols in water when the fluorescence peaks are contiguous, and the spectra overlap seriously meanwhile there are interference in the compounds.