babycow/TD-HER-Dataset

TD-HER数据集是一个用于电力系统频率极值和到达时间预测的预处理数据集，旨在支持论文Target-Dependent Heterogeneous Expert Routing for Power System Frequency Extremum and Arrival-Time Prediction的结果重现。数据集基于两个IEEE测试系统：IEEE 39-bus New England系统（10个发电机）和IEEE 300-bus系统（69个发电机），提供三种不同的数据表示：RepA（表格特征）、RepB（时空张量加静态特征）和RepC（图表示，包括邻接矩阵）。数据集涵盖多个早期窗口设置（如1毫秒、5毫秒、10毫秒、15毫秒、25毫秒），并包含训练、验证、测试以及跨条件和跨拓扑的分割。预测目标包括有符号的COI频率极值（单位：Hz）和从扰动触发到频率极值的非负时间（单位：秒）。数据通过PSS/E时间域仿真生成，模拟随机扰动和运行条件，并经过特征提取和表示构建流程处理。数据集总大小为18.0 GB，包含205个文件，许可证与TD-HER仓库相同。The TD-HER dataset is a pre-processed dataset for power system frequency extremum and arrival-time prediction, designed to reproduce results from the paper Target-Dependent Heterogeneous Expert Routing for Power System Frequency Extremum and Arrival-Time Prediction. It is based on two IEEE test systems: the IEEE 39-bus New England system (10 generators) and the IEEE 300-bus system (69 generators), offering three different representations: RepA (tabular features), RepB (spatiotemporal tensor plus static features), and RepC (graph representation with adjacency matrices). The dataset includes multiple early-window settings (e.g., 1 ms, 5 ms, 10 ms, 15 ms, 25 ms) and contains splits for training, validation, testing, as well as cross-condition and cross-topology scenarios. Prediction targets are the signed COI frequency extremum (in Hz) and the non-negative time from disturbance trigger to frequency extremum (in seconds). Data was generated via PSS/E time-domain simulations under randomized disturbances and operating conditions, processed through feature extraction and representation construction pipelines. The total dataset size is 18.0 GB with 205 files, released under the same license as the TD-HER repository.