摘要
新能源电力系统在运行过程中会产生大量的相似数据,导致电力系统状态的自动化监测效果不佳。对此提出基于奇异值分解的含新能源电力系统状态自动化监测系统。采用PowerEdge R740xd系列服务器集群作为智能处理层的核心设备,接收来自设备参数辨识层的数据。结合特征向量聚类分析进行SVD分解,构建LSTM网络结构,对运行状态数据的长期依赖关系进行捕获,并结合交叉熵损失函数实现网络结构训练,对输入状态的异常情况进行判定。测试结果表明,该文系统具备较高的异常判定灵敏度及较为理想的监测效果。
The new energy power system generates a large amount of similar data during operation,which leads to poor automatic monitoring of the power system’s status.Propose a state automation monitoring system for new energy power systems based on singular value decomposition.The PowerEdge R740xd series server cluster is used as the core device of the intelligent processing layer to receive data from the device parameter identification layer.Perform SVD decomposition combined with feature vector clustering analysis to construct an LSTM network structure,capture long-term dependencies of operational state data,and train the network structure using cross entropy loss function to determine abnormal input states.The test results show that the proposed system has high sensitivity in anomaly detection and ideal monitoring performance.
作者
苏卫江
王鹏
姚凌峰
SU Weijiang;WANG Peng;YAO Lingfeng(China Green Development Investment Group,Beijing 100020,China)
出处
《电子设计工程》
2025年第16期173-177,共5页
Electronic Design Engineering
关键词
奇异值分解
含新能源电力系统
监测系统
LSTM
singular value decomposition
new energy-containing power system
monitoring system
LSTM
作者简介
苏卫江(1983-),男,山东济南人,硕士研究生,高级工程师。研究方向:建筑工程与绿色能源工程、新能源电力系统自动化。
