基于门控循环单元的非均衡数据驱动异常用电检测方法

Gated Recurrent Units based imbalanced data driven abnormal electricity consumption detection method

摘要:

异常用电检测能够及时发现异常用电行为，在减少能源浪费和经济损失的同时能够维持安全、稳定的电网运行环境。智能电表的普及使得用电数据获取十分容易，为数据驱动的异常用电检测方法提供了充足的数据支持。然而，在实际应用过程中，异常数据较少导致的数据非均衡问题严重影响了模型的训练效果。因此，针对上述问题提出了一种针对非均衡数据的门控循环单元异常用电检测方法。该方法利用边界合成少数类过采样技术实现了对少数类数据的有效扩充。为了更好的捕捉用电数据的时序特征，采用了门控循环单元实现对用电数据的分类。为了验证该方法的有效性，基于非均衡数据集进行了对比实验。实验结果表明，该方法能够更好的数据扩充效果以及更准确的异常用电检测效果。

Abstract:

Abnormal electricity consumption detection can detect abnormal electricity consumption behaviors in time and maintain a safe and stable power grid operating environment while reducing energy waste and economic losses. The popularity of smart meters makes it very easy to obtain electricity consumption data, which provides sufficient data support for data-driven abnormal electricity consumption detection methods. However, the problem of imbalanced data seriously affects the training effect of the model in practical application. Therefore, in this paper, a gated recurrent units based abnormal detection method for imbalanced electricity consumption data is proposed. The method adopts the borderline synthetic minority oversampling technique to realize the effective extension of the minority data. To better capture the time series characteristics of electricity consumption data, gated recurrent units are employed to classify electricity consumption data. To verify the effectiveness of this method, comparative experiments are done on imbalanced data. Experimental results show that this method has better data expansion effect and more accurate abnormal electricity consumption detection effect.