关键词: 偏最小二乘法, 互信息, 天牛须搜索算法, BP神经网络, 汽油, 辛烷值

Abstract: Aiming at the highly nonlinear and redundant characteristics of the characteristic variables in the data set, using an improved long-horned beetle antennae search algorithm (RSBAS) based on combined dimensionality reduction method of partial least square regression (PLS) and mutual information (MI), an optimized BP neural network model (PLS-MI-RSBASBP) was proposed, and applied to predict the octane number of gasoline in S Zorb desulfurization unit.Firstly, the characteristic variables related strongly to gasoline octane number are selected by partial least squares method and mutual information combination algorithm, and then the RSBASBP model is used to predict the gasoline octane number, and compared with the prediction results by BP, GABP, and BASBP network models. The results show that the MAE, MSE, and RMSE of the prediction results by PLS-MI-RSBASBP model are smaller than those by other models,indicating the higher prediction accuracy of the model.Moreover, the PLS-MI-RSBASBP model can be used to determine the characteristic variables that affect the gasoline octane number, which can be effectively controlled and optimized.

Key words: partial least squares, mutual information, beetle antennae search algorithm, BP neural network, gasoline, octane number