"Emotional Intelligence (EI) of Large Language Models (LLMs) consists of emotion perception, cognition, and expression, playing a crucial role in perceiving user states, making appropriate responses, and enhancing user experience. Previous studies focus on improving emotion perception or emotion expression, but often ignore the exploration of a comprehensive improvement in EI of model. Moreover, as these studies do not take into account the conflicts between different emotional tasks, resulting in limited improvements on specific emotional tasks, we construct an emotion enhancement framework, Emo-MoE, to address the aforementioned issues. This framework consists of an inter-task routing expert and an intra-task routing expert. The inter-task routing expert enables effective cross-task knowledge integration, while the intra-task routing expert refines decision-making for internal task features. Through the collaboration of different experts, Emo-MoE comprehensively and accurately models the emotional semantics of various tasks, thereby enhancing the EI of the model. To alleviate the problem of catastrophic forgetting in models, we develop a replay module to generate high-quality replay data. Extensive experiments demonstrate that our emotion enhancement framework improves the EI of the model."