摘要
为解决传统的电力系统调度检修方法无法满足电力系统安全可靠运行要求,提出基于深度学习的调度检修预案,以提高电力系统的可靠性与运行效率。通过对数据进行收集与预处理,消除原始数据中的异常值、缺失值、错误数据。采用长短期记忆网络对电力系统历史数据进行训练学习,预测电力负荷及设备状态。采用条件生成模型,通过对抗训练来自动生成优化的调度检修预案。将提出的调度检修预案生成技术应用于实际的电力系统中,得到了电力系统1周的调度检修预案。为电力系统运维人员决策提供了参考。
To solve the problem that traditional power system scheduling and maintenance methods cannot meet the requirements of safe and reliable operation of the power system,a scheduling and maintenance plan based on deep learning is proposed to improve the reliability and operational efficiency of the power system.By collecting and preprocessing data,outliers,missing values,and erroneous data in the original data were eliminated.Using long short-term memory networks to train and learn historical data of the power system,power load and equipment status were predicted.Using a conditional generation model,an optimized scheduling and maintenance plan was automatically generated through adversarial training.The proposed scheduling and maintenance plan generation technology was applied to actual power systems,resulting in a one week scheduling and maintenance plan for the power system.This provides a reference for decision-making of power system operation and maintenance personnel.
作者
蔡思烨
卢泉篠
胡鹏
杨恩龙
余玉良
顾小旭
CAI Siye;LU Quanxiao;HU Peng;YANG Enlong;YU Yuliang;GU Xiaoxu(State Grid Shanghai Jiading Power Supply Company,Shanghai 201800)
出处
《粘接》
CAS
2024年第3期153-156,共4页
Adhesion
关键词
深度学习
电力系统
长短期记忆网络
条件生成模型
预案生成技术
deep learning
power system
long short term memory networks
conditional generation model
plan generation technology
