Metering Anomaly Diagnosis(MAD)
收藏资源简介:
Metering Anomaly Diagnosis (MAD) 是一个用于智能电网变电站的计量异常诊断数据集,用于支持多元时间序列分类研究。数据完全来源于实际部署的高级计量基础设施(AMI)系统,对工程应用具有重要的实际意义。该数据集包含504个三相四线智能电表,其中382个正常运行,122个经过人工现场检查或专业标记为具有各种异常。采样间隔为30分钟。样本按每日周期划分,每个样本包含48个数据点。数据集预处理已完成,共生成7,733个可用样本,存储在MAD.npz文件中。训练集和测试集已经划分,分别包含2,319和5,414个样本。样本分为7类,其中0表示正常运行,类别1-6代表六种不同的异常类型。
Metering Anomaly Diagnosis (MAD) is a metering anomaly diagnosis dataset for smart grid substations, designed to support multivariate time series classification research. The data is entirely sourced from actually deployed Advanced Metering Infrastructure (AMI) systems, holding significant practical value for engineering applications. This dataset contains 504 three-phase four-wire smart meters, among which 382 operate normally, and 122 have been manually inspected on-site or professionally labeled as having various anomalies. The sampling interval is 30 minutes. Samples are divided on a daily cycle, with each sample containing 48 data points. Data preprocessing has been completed, with a total of 7,733 usable samples generated and stored in the MAD.npz file. The training and test sets have been split, containing 2,319 and 5,414 samples respectively. Samples are categorized into 7 classes, where class 0 indicates normal operation, and classes 1 to 6 represent six distinct anomaly types.
Metering Anomaly Diagnosis (MAD) 数据集概述
数据集简介
- 目的:支持智能电网变电站中智能电表的多元时间序列分类研究。
- 数据来源:实际部署的高级计量基础设施(AMI)系统。
- 电表数量:504台三相四线智能电表(正常382台,异常122台)。
- 采样间隔:30分钟。
- 样本划分:按每日周期划分,每个样本包含48个数据点。
数据集构成
- 总样本数:7,733个可用样本。
- 训练集:2,319个样本(x_train, y_train)。
- 测试集:5,414个样本(x_test, y_test)。
- 存储格式:MAD.npz文件。
样本类别分布
| 类别ID | 描述 | 样本数量 |
|---|---|---|
| 0 | 正常 | 6,021 |
| 1 | 异常-1 | 265 |
| 2 | 异常-2 | 124 |
| 3 | 异常-3 | 128 |
| 4 | 异常-4 | 747 |
| 5 | 异常-5 | 120 |
| 6 | 异常-6 | 328 |
样本特征
- 数据来源:三相四线智能电表。
- 变量数量:14个(3相电压、3相电流、总有功功率、3相有功功率、总功率因数、3相功率因数)。
- 预处理:所有变量已归一化。
引用格式
bibtex @misc{MAD, author = {Wang, F.}, title = {{Metering Anomaly Diagnosis}}, year = {2025}, note = {https://github.com/IISGLab/MeteringAnomalyDiagnosis} }



