基于改进非支配排序遗传算法的智能变电站状态检测方法  

Intelligent substation state detection method based on improved non-dominated sorting genetic algorithm

作　　者：崔宸昱 CUI Chenyu(State Grid Shanxi Electric Power Company Ultra High Voltage Substation Branch,Taiyuan 030000,China)

机构地区：[1]国网山西省电力公司超高压变电分公司,山西太原030000

出　　处：《技术与市场》2024年第11期7-10,16,共5页Technology and Market

摘　　要：为提高变电站设备运行状态检测结果的可靠性,规范其运行,基于改进非支配排序遗传算法的应用,以某智能变电站为例,开展其状态检测方法的设计研究。布置传感器进行电力设备运行数据的采集,引进变分模态分解方法,进行电力设备运行数据去噪与降维处理;引进遗传算法,通过在知识领域与多个目标之间的迭代,找到最优平衡点,筛选出最有效的特征组合,实现对变电站中电力设备运行数据的知识集合生成与特征提取;引进深度迁移学习,构建并训练自组织映射(SOM)网络,此网络含多个神经元节点,自适应聚类输入特征,实现智能变电站设备的在线管理与异常检测。对比试验结果表明:该方法可以精准识别智能变电站在运行中的电力设备异常状态,可通过此种方式实现对变电站的智能管理。In order to improve the reliability of the operation status detection for substation equipment and standardize its operation,based on the application of improved non-dominated sorting genetic algorithm,this paper takes a certain intelligent substation as an example to carry out the design and research on its status detection method.Sensors are arranged to collect operational data from power equipment,and a variational mode decomposition method is introduced to denoise and reduce dimensionality of this data.Introducing genetic algorithms,by iterating between knowledge domains and multiple objectives,finding the optimal balance point,selecting the most effective feature combination,and achieving knowledge set generation and feature extraction from power equipment operation data in substations.Introducing deep transfer learning,constructing and training a self-organizing map(SOM)network consisting of multiple neural nodes,adaptively clustering input features,and achieving online management and anomaly detection of intelligent substation equipment.Comparative experimental results show that this method can accurately identify abnormal states in power equipment during operation,thereby enabling intelligent management of substations.

关 键 词：改进非支配排序遗传算法 智能变电站 在线管理 特征提取 检测方法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TM63[自动化与计算机技术—控制科学与工程]