Abstract: Based on the data of 100 groups of gasoline samples analyzed by near infrared spectroscopy, the prediction model of gasoline octane number (SSA-BP) was established by pre-processing with different methods and optimizing BP neural network model with the sparrow search algorithm (SSA). The predicted value of the model was fitted with the measured value of gasoline research octane number (RON) and compared with those results obtained by partial least squares model and BP neural network model. The results showed that the SSA-BP model had the highest prediction accuracy, with a validation set determination coefficient (R²) of 0.976 0 and a prediction standard deviation (RMSEP) of 0.326 after the standard normal variate transformation(SNV) data pre-processing. Among the three gasoline octane number prediction models, SSA-BP neural network model has the best prediction accuracy and good applicability and stability. It shows that the SSA-BP neural network model, which preprocesses the spectral data with SNV method, can be better used for the prediction of gasoline octane number. The prediction results have good accuracy, it can realize the rapid non-destructive detection of gasoline octane number.

Key words: gasoline, octane number, sparrow search algorithm, BP neural network, near infrared spectroscopy, partial least squares regression