摘要
电力电子设备在新能源、电动汽车、储能等众多领域发挥着重要作用,保障了电能的高效、精准、灵活利用。基于数字控制的电力电子设备拥有采样、计算、存储、通信环节,具备了实现智能功能的基本要素。然而当前电力电子设备的数字处理环节以实时控制为主,未考虑电力电子设备的智能化需求,难以承载监测、优化及其他非线性复杂智能任务。该文计及电力电子设备数量多、聚集度高、成本控制严苛的应用特点,将其视为智能边缘端,分析实现边缘智能的潜力,探讨实现边缘智能的路径,展望智能化应用方式。电力电子的边缘智能与其他领域的边缘智能既有相似性也有差异性,电力电子边缘智能的发展成果可为其他领域的智能化发展提供借鉴参考。
With the rapid development of artificial intelligence,embedded computing,and Internet of Things technologies,the intelligence of power electronics systems is gradually moving from conceptual exploration toward practical implementations.This paper systematically examines the potential,implementation pathways,and future application prospects of edge intelligence in the domain of power electronics,aiming to provide both a theoretical foundation and a practical framework for the development of new generation power electronics systems.This paper first examines the potential for edge intelligence in power electronics from the perspectives of the available data resources sampled naturally by the embedded sensors and the computational resources installed inside the equipment.On one hand,power electronics systems operate at high-frequency sampling,generating massive amounts of operational data that have yet to be fully analyzed or utilized.On the other hand,the digital signal processors(DSP)embedded in these equipment continuously upgrade in the past few decades,whose computational capability is far beyond the control consumption,in consequence,a large portion of computational resources remains underutilized while only real-time control tasks are executed.Therefore,the abundant data,coupled with redundant computational resources,provides the foundation for implementing edge intelligence in power electronics systems.Building on this foundation,the paper proposes three typical implementation pathways for edge intelligence of power electronics:(i)a lightweight pathway based on conventional processors,which leverages the residual computational and storage resources of DSPs to perform light tasks such as intelligent power conversion control and status monitoring;(ii)a computationally enhanced pathway based on external AI chips,embedding high-performance AI processors in power electronicsequipment to enable complex intelligent reasoning and decision-making while ensuring the real-time performance of high-frequency control;and(iii)a multi-equipment collaborative pathway based on the Internet of Things,which promotes the transition of edge intelligence from individual equipment to coordinated groups through equipment interconnection and resource sharing.Subsequently,three case studies—intelligent inverter control,DC arc detection,and intelligent completion of missing data—are presented to simply demonstrate the engineering feasibility of these pathways under different computational conditions and application scenarios.Furthermore,the paper presents an application-oriented perspective on edge intelligence across three levels.At the equipment level,it highlights the evolution from conventional control units to intelligent nodes endowed with autonomous sensing,collaborative decision-making,and security protection capabilities,thereby enabling intelligent functions of power electronics equipment.At the system level,the focus is on multi-equipment coordination and intelligent optimization to achieve system-wide adaptive control,efficient resource allocation,and enhanced overall operational performance,which could serve the intelligent operation of renewable generation station,smart factory,and etc.At the smart city level,it emphasizes the deployment of edge intelligence for real-time energy consumption monitoring,carbon emission management,and rapid emergency response,ultimately driving urban infrastructure toward greater intelligence,sustainability,resilience,and safety.Overall,this paper provides the theoretical framework,implementation pathways,and application value of edge intelligence in power electronics systems,offering significant potential for practical engineering applications.Edge intelligence in power electronics shares both similarities and distinctions with edge intelligence in other domains.Advances in this field could provide valuable insights for intelligent systems in other areas.
作者
高峰
Gao Feng(School of Control Science and Engineering Shandong University,Jinan 250061 China)
出处
《电工技术学报》
北大核心
2026年第3期725-737,共13页
Transactions of China Electrotechnical Society
基金
国家自然科学基金重点项目(52537008)
国家自然科学基金杰出青年科学基金项目(52225705)资助。
关键词
电力电子
边缘智能
智能终端
物联网
轻量化模型
Power electronics
edge intelligence
intelligent terminal
internet of things
lightweight model
