作者:\n\t高宝琪，马捍超，毕艳冰，张小军，吴苏州\n

Authors:\n\tGAO Baoqi,MA Hanchao,BI Yanbing,ZHANG Xiaojun,WU Suzhou\n
摘要:\n\t为对高压电缆运行状态实时监测、保证电力安全运输，研究基于深度信念网络的高压电缆局部放电信号捕捉方法。利用改进变分模态分解算法去除原始局部信号内部噪声；建立深度信念网络捕捉模型，利用对比散度算法预训练受限波尔兹曼机，获取捕捉模型预训练网络参数；利用自适应矩估计算法全局微调捕捉模型，在训练完成的捕捉模型内输入去噪后的信号，输出局部放电信号的捕捉类型。实验表明：所研究方法有效去除局放信号内部噪声；在不同干扰时，该方法的捕捉概率在0.9以上，具备较优的抗干扰性能，且精准捕捉绝缘缺陷引起的局部放电信号类型。\n

Abstract:\n\tIn order to monitor the running state of high-voltage cable in real time and ensure the safe transportation of power, a partial discharge signal capture method of high-voltage cable based on deep belief network is studied.The improved variational modal decomposition algorithm is used to remove the internal noise of the original local signal; the deep belief network capture model is established, and the limited Boltzmann machine is pre trained by contrast divergence algorithm to obtain the pre training network parameters of the capture model; using the global fine tuning capture model of adaptive moment estimation algorithm, the denoised signal is inputted into the trained capture model and the partial discharge signal is outputted.Experimental results show that the proposed method can effectively remove the internal noise from partial discharge signals.In the case of different interferences, the capture probability of this method is above 0.9, which has excellent anti-interference performance and accurately captures the partial discharge signal types caused by insulation defects.\n


PDF全文下载地址:

可免费Download/下载PDF全文