摘要
各级电力公司积累了大量的历史发电数据,但是一直以来都没有得到有效利用。针对采用手工方式进行电力调度存在效率较低的问题,提出了一种基于长短时记忆网络(LSTM)和蚁群算法的智能调度算法。基于LSTM深度学习算法,通过对电力公司历史数据进行建模,LSTM算法可以有效提取出有效特征,实现预测一定条件下机组的耗煤量。同时,利用蚁群算法(ACO)设计了一种智能电力调度算法,从而在满足在完成实时发电任务的情况下,尽可能地节能减排。试验表明,采用LSTM算法,相比较线性回归、随机森林等算法,预测耗煤量的均方误差更小;采用ACO算法,相比较等微增率法、动态规划法以及遗传算法,可以更加快速、有效地分配发电负荷。
The electric power companies at all levels have accumulated a large amount of historical power generation data,butthese data have not been effectively used. When conducting power dispatching, artificial manual operation is still used forscheduling and the efficiency is low.A new intelligent electric power dispatching algorithm based on long.term and short.termmemory network （LSTM） and ant colony optimization （ACO） is proposed.Based on the LSTM deep learning algorithm,bymodeling the historical data from the electric companies,the LSTM algorithm can effectively extract effective features and achievethe prediction of coal consumption of generation unit under certain conditions. In addition,by adopting ACO algorithm, anintelligent power dispatching algorithm is designed,so as to meet the requirements of real.time power generation tasks and saveenergy and reduce emissions as much as possible.Experimental results show that compared with linear regression,random forestand other algorithms,using the LSTM algorithm,the prediction of coal consumption has smaller mean.square error; compared withmicro.increment method,dynamic programming method and genetic algorithm,using ACO algorithm can dispatch electric powerload more quickly and effectively.
作者
梁肖
胡逸飞
黄太贵
李端超
高卫恒
孙仪
LIANG Xiao;HU Yifei;HUANG Taigui;LI Duanchao;GAO Weiheng;SUN Yi(Dispatching and Control Center,Anhui Electric Power Corporation,Hefei 230022,China;Department of Computer Science and Technology,University of Science and Technology of China,Hefei 230027,China)
出处
《自动化仪表》
CAS
2018年第5期98-102,共5页
Process Automation Instrumentation
基金
国家电网公司科技基金资助项目(521200160026)
关键词
智能电网
节能调度
机器学习
深度学习
长短时记忆网络
蚁群算法
Smart grid
Energy.saving dispatching
Machine learning
Deep learning
Long.term and short.term memorynetwork
Ant colony optimization
