基于深度学习的用户异常用电模式检测

针对电力用户的异常用电行为,提出一种基于深度学习的用户异常用电模式检测模型。利用Tensor Flow框架,构建了特征提取网络和多层特征匹配网络。基于长短期记忆(LSTM)的特征提取网络,从大量时间序列中提取出不同的序列特征。基于全连接网络(FCN)的多层特征匹配网络,利用提取出的特征数据,完成对异常用电数据的检测。实例分析表明,与非深度学习检测模型相比,所提模型可更加有效地完成异常用电模式检测。此外,与多层LSTM分类模型相比,所提模型具有更好的准确性和鲁棒性。

Anomaly detection for power consumption pattern based on deep learning

To deal with the anomalous power consumption behavior of users, a model of anomaly detection for power consumption pattern based on deep learning is proposed, in which the TensorFlow framework is employed to establish a feature extracting network and a multi-layer feature matching network. The feature extracting network based on LSTM(Long Short Term Memory) extracts different sequence features from large-scale time series data. The multi-layer feature matching network, which is based on FCN(Fully Connected Network),uses the extracted different sequence features to detect the anomalous electrical power consumptions. Case studies show that the proposed model can detect the anomalous power consumption mode with higher efficiency compared to non-deep learning models. Compared to the multi-layer LSTM classification model, the proposed model is more precise and robust.