Abstract: Abstract: In recent years, pesticide has been mass-producing and widely used. The problem of pesticide residues has attracted more and more attention. As the problem of food safety is becoming the focus of society, the pesticide residue detection has become a research hotspot. Among numerous methods of pesticide detection, surface-enhanced Raman spectroscopy (SERS) has become an area of intense research owing to a highly sensitive probe for the trace level detection of pesticide. The spectroscopic merits of SERS are the representation in the aspects of super sensitivity, high selection and water resistance, which make it one of the most popular detection techniques currently. In this paper, the organophosphorus pesticide phosmet and dimethoate were selected as the research objects. The blended pesticide residues of phosmet and dimethoate on navel orange were detected by the SERS combined with chemometrics algorithm. The silver nanowires were used as SERS substrate to detecte pesticide residue on navel orange. Firstly, the navel orange samples were fabricated with pesticide residues. Secondly, the silver nanowires SERS substrate was fabricated. Then the sample solution to be measured was dripped onto the dried SERS substrate. When the sample was dried, spectral data were collected. The spectral data were used to analyze pesticide residue qualitatively and quantitatively. It had a better enhancement effect on the qualitative analysis of mixing pesticides for silver nanowires substrate. Pesticide original spectral data were processed by the partial least square (PLS) modeling algorithm and the different pretreatment methods. The PLS regression combined with different data preprocessing methods was used to develop quantitative models of mixing pesticide residue. And the advantages and disadvantages of the models were compared. The results showed that the model built by the PLS combined with the second derivatives data preprocessing was ideal for mixing pesticides, whose correlation coefficient (Rp) for the prediction was 0.954, and root mean square error of prediction (RMSEP) was 4.822 mg/L. The model combined with the baseline was ideal for phosmet, whose Rp was 0.898 and RMSEP was 6.621 mg/L. The model combined with the multiplicative scattering correction (MSC) was ideal for dimethoate, whose Rp was 0.911 and RMSEP was 7.369 mg/L. Therefore, the study combines the SERS and chemometrics algorithm to detect pesticide residues qualitatively and quantitatively, which is feasible. Raman spectroscopy can be used as a fast and simple tool to detecte mixing pesticide residues. It provides a basis for the more insightful study on pesticide residues detection.