摘要
电力人工智能技术不断取得突破的同时,实际应用中也面临诸多挑战,在电力人工智能的假设分析与应用范式研究基础上,探索电力人工智能自主学习的创新应用模式。首先,针对电力人工智能研究中存在的可信伦理、数据分布与进化迁移等瓶颈,提出并详细阐述数据知识融合、平行互动、模型进化三大机制;进而,基于生物脑认知原理,提出适用于电力领域人工智能应用的意识引导的自主学习技术,通过构建电力领域机器意识引导算法进行模型构建、数据组织、训练调优等自主学习应用,解决规则复杂、数据价值低、场景泛化等情况下的电力人工智能模型训练优化难题;最后,在设备运维检修领域开展应用探索,通过意识引导知识、数据、任务的理解分析,构建面向复杂运维检修任务端到端算法生成的智能应用。
While electric power artificial intelligence technology continues to make breakthroughs,it also faces many challenges in practical applications.Based on the hypothetical analysis and application paradigm research of electric power artificial intelligence,this article explores the innovative application mode based on self-learning AI.First,aiming at the bottlenecks of credibility ethics,data distribution,and evolutionary migration that existed in the research of electric power artificial intelligence,three mechanisms of data knowledge fusion,parallel interaction,and model evolution are proposed and elaborated.Furthermore,based on the biological-brain-cognitive principle,a consciousness guided self-learning technology suitable for the application of artificial intelligence in the power field is proposed.The self-learning applications such as model construction,data organization,and training optimization are carried out by constructing the machine consciousness guidance algorithm to solve the problems of training and optimization of electric power artificial intelligence model under the conditions of complex rules,low data value,and scene generalization.Finally,application exploration is carried out in the field of equipment operation and maintenance.By guiding the understanding and analysis of knowledge,data and tasks through the consciousness mechanism,an intelligent application for the generation of end-to-end algorithms for complex operation and maintenance tasks can be built.
作者
蒲天骄
张中浩
谈元鹏
莫文昊
郭剑波
PU Tianjiao;ZHANG Zhonghao;TAN Yuanpeng;MO Wenhao;GUO Jianbo(China Electric Power Research Institute,Haidian District,Beijing 100192,China)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2023年第10期3705-3717,共13页
Proceedings of the CSEE
基金
国家电网公司科技项目(5100-201912483A-0-0-00)。
关键词
电力人工智能
意识引导
类脑智能
自主学习
驱动机制
electric power artificial intelligence
consciousness guidance
brain-like intelligence
self-directed learning
driving mechanism
作者简介
蒲天骄(1970),男,教授级高级工程师,博士生导师,长期从事电力系统优化运行、电力人工智能等方面的研究工作,tjpu@epri.sgcc.com.cn;通信作者:张中浩(1991),男,博士,工程师,研究方向为电力人工智能等,zhangzhonghao@epri.sgcc.com.cn;谈元鹏(1987),男,博士,高级工程师,主要从事电力人工智能等方面的研究工作,tanyuanpeng@epri.sgcc.com.cn;莫文昊(1996),男,硕士,工程师,主要从事电力人工智能等方面的研究工作,mowenhao@epri.sgcc.com.cn;郭剑波(1960),男,教授级高级工程师,博士生导师,长期从事电力系统规划、运行分析和电网可靠性等方面的研究工作。
