基于VMD-IKHA-DBN的高压输电电缆局部放电模式识别方法

Partial discharge pattern recognition method for high voltage transmission cables based on VMD-IKHA-DBN

针对当前高压输电电缆局部放电模式识别准确性不高的问题,本文提出了一种基于VMD-IKHA-DBN的高压输电电缆局部放电模式识别方法。首先分析了高压输电电缆局部放电产生原因及类型,并搭建了高压输电电缆局部放电实验平台用于采集原始信号。然后采用变分模态分解算法完成局部放电的信号分解,并引入多尺度排列熵理论构建特征向量样本集合。最后提出一种基于Logistic混沌映射、动态反向学习和高斯变异的组合策略改进的磷虾群算法用于优化深度信念网络超参数,从而得到基于IKHA-DBN的高压输电电缆局部放电模式识别模型。实验结果证明本文所提出的方法识别准确率达到了98.3333%且识别效率较高,实现了高压输电电缆局部放电模式的高效准确识别,在电缆运检工作中可以充分发挥工程效能。

Aiming at the current problem of low accuracy of partial discharge pattern recognition in high-voltage transmission cables,this paper proposes a partial discharge pattern recognition method for high-voltage transmission cables based on VMD-IKHA-DBN.Firstly,the causes and types of partial discharges in high-voltage transmission cables are analyzed,and an experimental platform for partial discharges in high-voltage transmission cables is constructed to collect the original signals.Then the Variational Mode Decomposition(VMD)algorithm is used to complete the signal decomposition of partial discharges,and the multi-scale arrangement entropy theory is introduced to construct the sample set of feature vectors.Finally,an Improved Krill Herd Algorithm(IKHA)based on the combination of Logistic Chaos Mapping,Dynamic Reverse Learning and Gaussian Variation is proposed to optimize the hyper-parameters of Deep Brief Network(DBN),so as to obtain the IKHA-DBN based high voltage transmission cable partial discharge pattern recognition model based on IKHA-DBN.The experimental results prove that the recognition accuracy of the method proposed in this paper reaches 98.3333%and the recognition efficiency is high,which realizes the efficient and accurate recognition of partial discharge patterns of high-voltage transmission cables,and can give full play to the engineering efficiency in cable operation and inspection work.