HawkDATA

该数据集名为HawkDATA，是为了非侵入式家电负荷监测（NALM）而收集的。它旨在无需室内传感器，仅从总表就能识别单个家电的使用情况。该数据集平衡且多样化，以便于准确的事件和状态识别。在构建数据集时，采用了平衡的格雷码策略来生成事件，确保收集到的数据在多样性和平衡性方面均有所保障。在规模上，训练集包含2880个事件和2584个独特状态，验证集包含2112个事件和1962个独特状态。该数据集的任务是家电使用中的事件和状态识别。

This dataset, named HawkDATA, was collected for non-intrusive appliance load monitoring (NALM). It aims to identify the usage states of individual appliances solely using data from the main power meter without deploying indoor sensors. The dataset is balanced and diversified to facilitate accurate event and state recognition. During its construction, a balanced Gray code strategy was adopted to generate events, ensuring that the collected data guarantees both adequate diversity and balanced distribution. In terms of scale, the training set contains 2880 events and 2584 unique states, while the validation set includes 2112 events and 1962 unique states. The task of this dataset is event and state recognition in appliance usage.

搜集汇总

背景与挑战

背景概述

HawkDATA是一个用于非侵入式低功耗电器识别系统的评估数据集，大小约5.9GB，包含采样同步和算法验证数据。该数据集需结合特定项目代码使用，支持XGBoost模型进行电器状态识别，并与BLUED数据集进行性能对比。