基于无监督聚类算法的风电场高压电缆局放脉冲识别

针对风电场高压电缆局部放电在线监测中脉冲型干扰难以抑制的技术难题,分析了风电场高压电缆局部放电脉冲及各类干扰脉冲的分布特征,并在此基础上,提出了基于无监督聚类算法的风电场局部放电脉冲识别策略。首先对现场多个工频周期的原始脉冲群依次进行自适应、自学习聚类,然后根据聚类结果计算不同脉冲群的三维分布谱图,最后根据不同类脉冲的分布差异有效地识别出局部放电脉冲。该策略不需要脉冲样本知识,面向复杂的电磁干扰环境具有很强的适应性与实用性,某风电场工程应用实践证明了该策略能准确地实现脉冲识别。

Partial Discharge Pulse Recognition for High Voltage Cable in Wind Farm Based on Unsupervised Clustering Algorithm

Aiming at the technical problem that it is difficult to suppress the pulse interference during partial discharge online monitoring of high voltage cable in wind farm, the distribution characteristics of the partial discharge pulses and all kinds of interference pulses were analyzed. And then the partial discharge pulse recognition strategy was put forward based on unsupervised clustering algorithm. Firstly, the self-learning and adaptive clustering method was used to divide into all the pulses in the original signal into several pulses classes. Secondly, three-dimensional pattern chart of each pulse class was calculated. Thirdly, three-dimensional pattern chart of each pulse class was used to effectively identify PD pulses according to the difference among the distribution of different kinds of pulses. The strategy does not need sample knowledge of pulses and has strong practicability and adaptability in complicated electromagnetic interference environment. The application in a wind farm project proves that the proposed pulse recognition strategy has good practicability.