【摘要】 针对现有方法无法有效提取煤矿电缆短路故障深层特征而导致故障识别准确率和类型判定精度低的问题，提出了一种基于IDT-SAE-ELM的短路故障识别方法。首先采用IDT技术对传统SAE模型进行改进，以提高其高效捕获故障样本深层特征的能力；然后利用Adam算法优化IDT-SAE模型参数，实现了从原始电流信号自动获取短路故障特征量；最后利用ELM模型替代Softmax构造故障分类器，以提高SAE模型对特征差异性小的故障类型辨识能力，实现对煤矿电缆短路故障的识别与类型的智能判定。以煤矿电网实际参数进行短路故障仿真，分别利用Loss曲线与T-分布随机近邻嵌入算法可视化分析所提方法的抗过拟合能力与短路故障深层特征挖掘能力，采用准确率和精度对所提方法进行评价，结果表明：所提方法相较于传统SAE具有更好的故障特征提取能力和抗过拟合能力；所提方法对电缆短路故障的识别准确率稳定在99%左右，相较于RF、BPNN、ELM等人工智能方法，准确率分别提高了7.47%、5.82%、5.42%;在严重噪声干扰下，所提方法短路故障识别准确率始终保持在98.75%以上，有效提高了煤矿电缆短路故障识别准确率和类型判定精度，能够为越级跳闸原因判别、短路事故的分析与处理提供重要依据。更多还原

【Abstract】 A short-circuit fault recognition method based on IDT-SAE-LM was proposed to address the problem of low accuracy in fault identification and type determination due to the inability of existing methods to effectively extract deep features of coal mine cable short-circuit faults.Firstly, the traditional SAE model was improved by IDT technology to enhance its ability to efficiently capture the deep features of fault samples.Then, the Adam algorithm was used to optimize the IDT-SAE model parameters, and the short-circuit fault feature quantity was automatically obtained from the original current signal.Finally, the ELM model was used to replace Softmax to construct the fault classifier, so as to improve the ability of SAE model to identify the fault type with small feature difference, and realize the identification and type intelligent judgment of coal mine cable short-circuit fault.The short-circuit fault simulation was carried out with the actual parameters of the coal mine power grid.The Loss curve and the T-distributed random neighbor embedding algorithm were used to visually analyze the anti-overfitting ability and the deep feature mining ability of the short-circuit fault of the proposed method.The accuracy and precision were used to evaluate the proposed method.The results show that : Compared with the traditional SAE model, the proposed method has better fault feature extraction ability and anti-overfitting ability; Compared with artificial intelligence methods such as RF,BPNN,and ELM,the accuracy is improved by 7.47%,5.82%,and 5.42%,respectively.Compared with the traditional SAE method, the accuracy is improved by about 11%.Under severe noise interference, the accuracy of short-circuit fault identification in this method is always above 98.75%,which effectively improves the accuracy of short-circuit fault identification and type determination of coal mine cables, and can provide an important basis for the identification of override trip causes and the analysis and treatment of short-circuit accidents.更多还原