With the advancements in materials and pattern recognition algorithms, electronic nose (E-nose) has been continuously developing and finding increasingly widespread applications. However, due to factors including but not limited to variations in temperature and humidity, as well as chemical and physical interactions among analytes and gases, E-nose is prone to drift during practical usage, leading to decreased accuracy of pattern recognition algorithms. But now, this challenge can be overcome with the help of meta-learning. We applied the model-agnostic meta-learning (MAML) architecture from meta-learning and designed a residual classifier with learnable per-parameter learning rates. Leveraging the ability of ‘learning to learn’, our method do not require a large amount of target domain data and there is no need to repeat multiple training iterations when transferring to any other target domain compared to other unsupervised algorithms, resulting it highly deployable. Additionally, we present a Gini value-based approach to enhance the certainty of model output labels in a multi-task setting. Through the evaluation of multiple experiments, we demonstrate the feasibility of our approach and achieve very high classification accuracy on a publicly available long-term drift dataset compared with other methods.