【摘要】 针对低压配电系统中,实际配电网负载端的负载串并联形式的多样化和多变性特点,研究干路点检测支路发生的故障电弧具有十分明显的实际意义,提出了一种基于深度长短时记忆(LSTM)网络的多支路串联故障电弧检测方法。首先,构建实验平台,采集支路发生不同串联故障情况下的干路电流信号共计72 000组;然后,将电流信号分为训练集和测试集;最后,通过Python平台优化深度LSTM网络模型结构以识别故障电弧,并输出检测结果。实验结果显示改进的LSTM网络对于每组实验单独分类检测准确率最低为96.8%,最高可达99.0%,多组实验统一进行检测准确率达94.88%。该方法能够有效识别多支路负载下的串联故障电弧,为低压串联故障电弧的准确检测提供了新的思路和有益探索。更多还原

【Abstract】 In low-voltage distribution system,due to the diversity and variability of load series and parallel connection at the load end in the actual distribution network,it is of great practical significance to study the fault arc of trunk road point detection branch. A multi-branch series fault arc detection method based on deep Long Short-Term Memory(LSTM)network was proposed to detect the fault arc. Firstly,the experimental platform was constructed to collect 72 000 sets of current signals of trunk road under different series faults of branches. Then,the current signals were divided into training set and test set. Finally,the deep LSTM network model structure was optimized by Python platform to identify fault arcs and output detection results. The experimental results show that the accuracy of the improved LSTM network is 96. 8% to 99. 0% when the experimental load is separately classified and detected. When multiple load current signals were detected together,the accuracy is 94. 88%. The method can effectively identify the series fault arcs under multi-branch load and provides a new idea and beneficial exploration for the accurate detection of low-voltage series fault arcs.更多还原