Fingerprint Discovery for Transformer Health Prognostics from Micro-Phasor Measurements
Fingerprint Discovery for Transformer Health Prognostics from Micro-Phasor Measurements
复制标题
通过微相量测量进行变压器健康预测的指纹发现
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
E. Stewart
中科院分区：
文献类型：
--
作者：
Jose Cadena;P. Ray;E. Stewart
电力从发电机传输到用户的关键部件是电力变压器。每个配电馈线可能有数百个设备分布在一个大的地理区域。变压器故障是电网弹性和可靠性的重要指标。从微相量测量单元（μPMUs）获得的大量高粒度数据的可用性为将机器学习（ML）技术应用于变压器健康诊断和预测提供了独特的机会。在这里，我们提出了一个贝叶斯非参数模型来揭示与变压器健康相关的“时间特征”，这可能用于开发风险分层系统。我们提供了来自加州河滨市的网格数据的结果，以证明我们提出的方法的有效性。
A key component in the transfer of electric power from the generator to the consumer is the power transformer. Each distribution feeder may have hundreds of devices spread throughout a large geographical area. Transformer failures are a key indicator of grid resiliency and reliability. The availability of large quantities of high-granularity data obtained from micro-phasor measurement units (μPMUs) provides a unique opportunity for applying machine learning (ML) techniques for transformer health diagnostics and prognostics. Here, we propose a Bayesian non-parametric model for uncovering ”temporal signatures” related to transformer health, which may be utilized for developing risk stratification systems. We provide results on grid data from Riverside, CA to demonstrate the efficacy of our proposed approach.