Pu-erh tea storage year detection has the problems of cumbersome operation and complicated evaluation process. On this basis, a voltammetric electronic tongue (VE-Tongue) came out for fast detection of different storage years of Pu-erh tea. Conventional pattern recognition method of VE-Tongue was mainly based on manual feature design combined with shallow machine learning algorithms. In this study, a deep learning algorithm was introduced into pattern recognition method of VE-Tongue. A hybrid pattern recognition method based on combination of one-dimension convolutional neural network (1-D CNN) and extreme machine learning (ELM) was proposed. The 1-D CNN-ELM model combined with VE-Tongue was utilized to distinguish Pu-erh tea with five different storage years. The result showed that compared with traditional models based on discrete wavelet transform (DWT) combining with support vector machine (SVM) or extreme machine learning, 1-D CNN-ELM model gained better classification performance, in which the accuracy, precision, recall and F1-Score were 98.32%, 98.0%, 98.0% and 0.98 respectively. This experiment illustrated that deep learning algorithm was suitable for pattern recognition dispose of VE-Tongue signal and could obtained superior classification accuracy and generalization ability.