首页|期刊导航|家电科技|家用中央空调系统智能化技术从黑盒到灰盒的演进综述

家用中央空调系统智能化技术从黑盒到灰盒的演进综述OA

A review of the evolution of smart technology in household central air conditioning systems from black box to gray box

中文摘要英文摘要

家用中央空调的智能化升级对控制技术提出了更高要求,由数据驱动"黑盒"模型向物理-数据融合"灰盒"模型的演进成为当前研究热点.围绕"感知-建模-控制-保障-应用"的技术链条,以"黑盒向灰盒演进"为主线,首先从多源环境感知与三维建模角度,总结了负荷识别技术的发展现状;其次,对比分析了黑盒模型与灰盒模型的建模机制及适用特点,并综述了物理信息神经网络(PINN)与强化学习等典型方法;进一步从可解释性、跨场景泛化能力及系统稳定性三个维度,归纳了联邦学习与可解释人工智能等关键支撑技术;在此基础上,探讨了动态调控、多设备协同及健康干预等方向的工程实践前景.结果表明,融合物理约束的灰盒模型在能效、控制精度及可解释性方面优势显著,可为家用中央空调智能控制技术的研究与工程应用提供参考.

The intelligent upgrade of residential central air conditioning puts higher demands on control technology,and the evolution from data-driven"black-box"models to physics-data fusion"gray-box"models has become a current research hotspot.Focusing on the technical chain of"perception—modeling—control—assurance—application"and taking"the evolution from black-box to gray-box"as the main thread,First,summarize the development status of load identification technology from the perspective of multi-source environmental perception and three-dimensional modeling.Secondly,it comparatively analyzes the modeling mechanisms and applicable characteristics of black-box and gray-box models,and reviews typical methods such as physics-informed neural networks(PINN)and reinforcement learning.Furthermore,from the three dimensions of interpretability,cross-scenario generalization ability,and system stability,the key supporting technologies such as federated learning and explainable artificial intelligence are summarized.On this basis,the engineering practice prospects in directions such as dynamic regulation,multi-device coordination,and health intervention are discussed.The results show that gray-box models integrated with physical constraints have significant advantages in energy efficiency,control accuracy,and interpretability,and can provide references for research and engineering applications in intelligent control technology for residential central air conditioning.

杨梦瑶;吴庆壮;刘武祥;吴楠;廖敏

小米科技(武汉)有限公司 湖北 武汉 430074小米科技(武汉)有限公司 湖北 武汉 430074小米科技(武汉)有限公司 湖北 武汉 430074小米科技(武汉)有限公司 湖北 武汉 430074小米科技(武汉)有限公司 湖北 武汉 430074

信息技术与安全科学

家用中央空调灰盒模型物理信息神经网络强化学习可解释人工智能节能控制

Household central air conditioningGrey-box modelPhysics-informed neural networksReinforcement learningExplainable artificial intelligenceEnergy-saving control

《家电科技》 2026 (4)

18-26,9

10.19784/j.cnki.issn1672-0172.2026.04.002

评论