家电科技 ›› 2025, Vol. 0 ›› Issue (1): 43-47.doi: 10.19784/j.cnki.issn1672-0172.2025.01.006

• 论文 • 上一篇    下一篇

基于DeepSeek推理大模型的智能家电智能体技术路线及应用

焦利敏1,2,3, 李红伟1,2,3, 曲宗峰1, 孔令磊4, 温燕斌5, 霍伟明6, 胡亚欣2,3   

  1. 1.中国家用电器研究院 北京 100176;
    2.中家院(北京)检测认证有限公司 北京 100176;
    3.国家智能家居质量检测检验中心 北京 100176;
    4.青岛海尔电冰箱有限公司 山东青岛 266510;
    5.海信空调有限公司 山东青岛 266000;
    6.广东美的制冷设备有限公司 广东佛山 528300
  • 出版日期:2025-02-01 发布日期:2025-05-13
  • 作者简介:焦利敏,学士学位。研究方向:智能家居技术、标准、检测、认证、检测设备等。地址:北京市大兴区亦庄经济技术开发区博兴八路3号中国家用电器研究院。E-mail:jiaolm@cheari.com。

The technical route and application of intelligent home appliance agents based on the DeepSeek reasoning large model

JIAO Limin1,2,3, LI Hongwei1,2,3, QU Zongfeng1, KONG Linglei4, WEN Yanbin5, HUO Weiming6, HU Yaxin2,3   

  1. 1. China Household Electric Appliance Research Institute Beijing 100176;
    2. CHEARI (Beijing) Certification & Testing Co., Ltd. Beijing 100176;
    3. National Smart Home Quality Supervision & Inspection Center Beijing 100176;
    4. QingDao Haier Refrigerator Co., Ltd. Qingdao 266510;
    5. Hisense Air Conditioning Co., Ltd. Qingdao 266000;
    6. Guangdong Midea Air-Conditioning Equipment Co., Ltd. Foshan 528300
  • Online:2025-02-01 Published:2025-05-13

摘要: DeepSeek自2025年1月发布以来,凭借其较强的推理能力、混合专家模型MoE(Mixture of Experts)、门控机制等创新技术,以及算力要求和训练成本的降低,吸引了亚马逊、微软、百度、腾讯等国内外科技企业和海尔、美的、海信、TCL等头部家电企业纷纷接入DeepSeek大模型,使得普通百姓享受到了DeepSeek给生活和工作带来的便利,具备了从用户拉动基于DeepSeek推理大模型的智能家电智能体的基础,为解决智能家电用户体验不佳、市场销售不及预期的问题提供了新的方案。因此,研究DeepSeek的运行机理和智能家电智能体的架构,从问题边界、知识库构建、人机交互界面等方面研究基于DeepSeek推理大模型的智能家电智能体使用技术路线及应用,期望加速DeepSeek在智能家电行业的应用,提升用户使用体验的满意度,为家电行业的智能化转型提供一种新的思路。

关键词: DeepSeek, 推理大模型, 智能家电, 智能体

Abstract: Since the release of DeepSeek in January 2025, it has attracted the attention of domestic and foreign technology companies such as Amazon, Microsoft, Baidu, and Tencent, as well as leading home appliance companies like Haier, Midea, Hisense, and TCL, due to its strong reasoning ability, innovative technologies such as the Mixture of Experts (MoE) and gating mechanisms, and the reduction in computing power requirements and training costs. This has enabled ordinary people to enjoy the convenience brought by DeepSeek to their lives and work, laying the foundation for users to drive the intelligent home appliance agents based on the DeepSeek reasoning model, and providing a new solution to the problems of poor user experience and unsatisfactory market sales of smart home appliances. Therefore, studies the operation mechanism of DeepSeek and the architecture of intelligent home appliance agents, and explores the technical routes and applications of intelligent home appliance agents based on the DeepSeek reasoning model from aspects such as problem boundaries, knowledge base construction, and human-computer interaction interfaces. It is expected to accelerate the application of DeepSeek in the smart home appliance industry, enhance the satisfation of user experience, and provide a new idea for the intelligent transformation of the home appliance industry.

Key words: DeepSeek, Reasoning model, Intelligent home appliance, Agents

中图分类号: