DAS

This dataset, procured via Distributed Acoustic Sensing (DAS) technology, comprises field-collected data from power cables buried at depths ranging from 1.2 to 2.0 meters. Its primary objective is to furnish multi-class time-series data to facilitate research in the identification of intrusion events affecting power cables. The dataset encompasses five distinct intrusion event categoptical fiberories: hammering (labeled as 0), background noise (labeled as 1), vehicle vibration (labeled as 2), rainfall (labeled as 3), and fan interference (labeled as 4). Data for each event type were acquired using a SHHZSB-100 DAS system, employing a sampling rate of 5 kHz. The resultant data consists of single-channel time-series signals, each comprising 2500 consecutive sampling points, and is stored in CSV format. Each row of data contains 2500 columns representing floating-point signal values, normalized to the [0,1] interval, with the final column denoting the integer event label. Data preprocessing involved the application of a sliding window segmentation technique (window length: 500 ms, overlap ratio: 40%) and Min-Max normalization to ensure both the continuity and scale consistency of the event segments. The dataset was stratified by event category and partitioned into training (80%), validation (10%), and testing (10%) sets. The testing set additionally incorporates composite noise, inclusive of power grid harmonics and random impulses, to validate the generalizability of the models. This dataset is particularly suited for the development of intrusion detection algorithms based on multi-modal feature fusion (e.g., GASF/RP image transformation combined with BiLSTM time-series modeling) or attention mechanisms. It is especially recommended for validating the performance of end-to-end classifiers, such as GRT-Transformer, within complex buried environments. Researchers may extract time-frequency domain features (mean, variance, wavelet packet energy) or directly input the raw signals for model training. Data acquisition is subject to the CC BY-NC 4.0 license and requires application to the corresponding author to safeguard sensitive information regarding the location of power facilities.