摘要
在向新型电力系统升级转型的过程中,新一代人工智能技术是其中的关键创新技术之一,可以与传统机理方法形成优势互补。然而随着电力人工智能技术应用的推广与深入,逐渐凸显出影响应用成效的3个关键问题:数据均衡问题、模型可信性问题、实时优化协同问题。针对以上3个问题,文章梳理总结其对应的核心技术分支、发展现状与典型应用,提出通过数据增强、迁移学习、仿真推演等数据增强推演技术解决小样本带来的模型过拟合与泛化性能下降问题,通过串行、引导、嵌入、反馈、并行等数据机理融合模式解决模型安全性、可解释性、鲁棒性等可信危机问题,通过群体智能、混合增强智能等智能优化决策技术解决源荷高度不确定下的大规模资源快速、实时、精准决策。最后,结合3个关键问题维度,对电力人工智能发展所需要重点突破的技术方向进行了展望。
In the process of upgrading and transforming to the new type power systems,the new generation of artificial intelligence(AI)technology is one of the key innovative technologies that can complement traditional mechanistic methods.However,as the application of AI technology in the power industry becomes more widespread and in-depth,three key issues that affect its application effectiveness have gradually emerged:the data imbalance problem,the model credibility problem and the real-time optimization and collaboration problem.In response to these issues,this paper summarized the corresponding core technical branches,development status and typical application scenarios,and proposed solutions to the problems of model over-fitting and generalized performance degradation caused by small sample size through data augmentation and simulation inference techniques such as data enhancement,transfer learning and simulation deduction;the issues of model security,interpretability and robustness are addressed through data mechanism fusion modes;and the challenges of making large-scale,real-time,and precise decisions in the face of high uncertainty in power system are addressed through intelligent optimization decision-making technologies such as swarm intelligence and hybrid enhancement intelligence.Finally,based on the three key issues,this paper outlines the key technological directions that need to be focused on in the development of AI application in power industry.
作者
蒲天骄
韩笑
PU Tianjiao;HAN Xiao(China Electric Power Research Institute Co.,Ltd.,Haidian District,Beijing 100192,China)
出处
《电力信息与通信技术》
2024年第1期1-13,共13页
Electric Power Information and Communication Technology
基金
国家自然科学基金项目资助“基于边云协同的区域能源互联网优化运行智能理论与关键技术”(U2066213)。
关键词
新型电力系统
人工智能
数据不均衡
可信智能
决策智能
new type power systems
artificial intelligence
data imbalance
trustworthy intelligence
decision intelligence
