Abstract: Huangshan Maofeng tea has become one of the most famous Chinese tea due to its amazing orchid fragrance and fresh, sweet taste. However, different quality grades of Huangshan Maofeng tea vary greatly in price. The quality evaluation of tea has posed a great challenge in the tea market. The quality grades of variant tea are also related to the different microelements and concentrations. Traditional sensory evaluation methods cannot achieve fast and accurate discrimination, particularly depending on the manual experience. Alternatively, the chemical analysis can serve as an essential method for the quality evaluation of tea. But the chemical analysis for all microelements was confined to its complexity and time-consuming in a large-scale production under gradually refined detection standards with the fast expansion of tea market. Previous studies reveal that the samples collected from the same production or origin places have the similar microelement compositions and concentrations, indicating that the variation of tea grades depends only on a few types of microelements. Therefore, it is reasonable to select the typical microelements for the distinguishing performance, thereby to optimize the traditional chemical analysis. In this work, a new method was proposed based on the feature extraction using the Elastic Net, in order to simplify the procedure of conventional chemical analysis, while to improve the grade evaluation. First, 96 samples of Huangshan Maofeng tea were collected from three original places (Fuxi, Yangcun, and Xintian village) with 6 quality grades (advance 1-3 grades, and 1-3 grades) using the traditional manual process. The chemical analysis was used to analyze the types and contents of 19 microelements. Second, a cross-validation method was used to determine the optimal parameters in the Elastic Net, and 9 feature microelements (Gallic Acid, Epicatechin Gallate, Catechin, Epicatechin, Gallocatechin Gallate, Epigallocatechin, Glutamate, Arginine and catechins bitterness index) were selected when the cost function was minimized. Third, the radar chart was used to visualize the selected 9 microelements, indicating the tea grade evaluation. To quantify the classification, a quality grade evaluation model of Huangshan Maofeng tea was established on the selected feature microelements using partial least squares regression. Monte-Carlo method with 100 times was chosen to evaluate the stability and robustness of the presented model. The proposed method can reduce the number of microelements from 19 to 9, and thereby to improve the identification accuracy of quality grade evaluation from 69.55% to 79.31%, compared with the traditional chemical analysis. A principal component analysis (PCA) was also taken for comparison. The recognition accuracies of PCA and the proposed method for validation set were 70.79% and 78.72% respectively in the Monte-Carlo experiment. The experimental results demonstrated that the selection of feature microelements was feasible to simply the traditional chemical analysis, and improve the prediction performance. The analysis model based on the typical microelements can simplify the current chemical process, and thereby provide a flexible selection to the quality identification of tea.