Methanol gasoline because of its high octane number, low cost advantage to become the new fossil fuel alternatives, the methanol content of accurate detection is an important link in determine its quality, the quantitative analysis of methanol gasoline components is of great practical significance for alleviating the shortage of traditional petroleum resources but increasing demand in China. The conventional methods of methanol detection in methanol gasoline, such as alcohol analyzer determination, quick test box determination, etc., are complicated in operation and low in accuracy and quality , the conventional methods of methanol detection in methanol gasoline, such as alcohol analyzer determination, quick test box determination, etc., are complicated in operation and low in accuracy and quality. Near infrared analysis method is widely used in qualitative or quantitative analysis of components in many industries due to its detection speed and accuracy, Methanol gasoline near infrared spectrum are studied non-destructive detecting method, made up of 0.5%~30% methanol gasoline, nearinfrared spectrum acquisition system is designed and detect 60 components of methanol gasoline spectral data, Moving average smoothing method, S-G convolution smoothing and multiple scattering correction(MSC) were used to establish a prediction model after comparative analysis of spectral data, BP Artificial Neural Network(ANN) and Principal Component regression (PCR) were used to predict the determination coefficient and root mean square error of the mode, comparing the results and prediction effects of the two algorithms. The results show that the root mean square error of each model is less than 1%, and the fitting degree of SG smooth-principal component regression prediction model is the best, and the determination coefficient is 0.998 98, the model based on SG convolution smoothing algorithm and neural network algorithm has the smallest deviation between the predicted value and the true value, and its root mean square error (RMSEP) is 0.322 84%. This study shows that the performance of SG smooth-neural network prediction model in the application of near infrared spectroscopy detection and analysis technology to detect methanol content in methanol gasoline is good, and meets the application requirements, this study provides a theoretical basis for the practical detection and application of methanol gasoline components, and provides technical support for the effective development and utilization of methanol gasoline.