摘要
海底观测网电力系统供电海缆易发生绝缘故障,由于其在海底极端环境下运行,实现高精度的电缆故障定位,对于降低其维修成本至关重要。基于海底观测网电能监控系统实时采集的电气测量参数,本文提出了基于主成分分析(Principal Components Analysis,PCA)和长短期记忆(Long Short-Term Memory,LSTM)网络的供电海缆高阻故障定位方法,先使用PCA进行数据降维,再将处理后的数据输入LSTM网络进行训练,捕获多源数据的时间特征,挖掘电力系统故障特征与电气参数之间的对应关系,实现海缆故障精确定位。通过在海底观测网原型系统上的实验证明,该方法经K折交叉验证方法(K-fold Cross Validation,K-CV)验证,准确率可达91%左右,故障定位误差约为0.4 km。
The submarine cables of seafloor observatory networks are prone to insulation failure.Since the cables operate in extreme subsea environments,achieving high-precision cable fault location is critical to reducing the maintenance costs.Based on the real-time electrical measurement parameters collected by the power monitoring system of submarine observatory networks,the Principal Components Analysis(PCA)and Long Short-Term Memory(LSTM)network is used in the submarine cable high-impedance fault location method proposed in this paper.The PCA is used for data dimensionality reduction,and then the processed data is put into the LSTM network for training.The LSTM network is applied to capture the time characteristics of multi-source data,mining the correspondence between cable fault characteristics and electrical system parameters to achieve accurate location of submarine cable faults.On the prototype system of the East China Sea seafloor observatories,the proposed approach is verified by K-fold Cross Validation(K-CV),and the experiments prove that the accuracy rate is about 91%and the fault location error is about 0.4 km.
作者
耿坤
吕枫
GENG Kun;LYU Feng(Tongji University,School of Ocean and Earth Science,Shanghai,200092,China)
出处
《海洋技术学报》
2020年第3期22-29,共8页
Journal of Ocean Technology
基金
上海市科技创新行动计划资助项目(16DZ1205000)
国家自然科学基金资助项目(61936014)。
关键词
海底观测网
电力系统
故障定位
长短期记忆网络
主成分分析
seafloor observatory networks
power system
fault location
long short-term memory
principal component analysis
