LSTM混合算法在用户电量数据异常检测中的应用

Application of Hybrid Algorithm Based on LSTM in Anomaly Detection of User Electricity Data

提出一种基于LSTM混合算法的用户电量数据异常检测模型。首先,提出了一种用于电力数据特征提取的检测器,该检测器通过从样本中提取抽象特征重新构造输入,并将重构误差与阈值进行比较,实现电力用户高维特征表示。其次,提出了混合CNN LSTM的电力数据异常分析网络,利用CNN叠加特征表示,并基于双层LSTM捕获电力数据上下文关系,从而有效提高模型的分类和回归能力。实验阶段,以某电力公司提供的电力数据为例,对所提模型进行验证。实验结果表明,所提模型性能最优,准确率和召回率分别为89.3%和69%。

In the paper, an anomaly detection model of user power data based on LSTM hybrid algorithm is proposed based on the study of CNN and LSTM.Firstly, a detector for feature extraction of power data is proposed.The detector extracts abstract features from samples, reconstructs the input, and compares the reconstruction error with the threshold to realize the high-dimensional feature representation of power users.Secondly, this paper proposes a hybrid cnn-lstm power data anomaly analysis network, which uses CNN superposition feature representation, and captures the context of power data based on double-layer LSTM,so as to effectively improve the classification and regression ability of the model.In the experimental stage, taking the power data provided by a power company as an example, the proposed model is verified.The results show that the performance of the proposed model is the best, and the accuracy and recall are 89.3% and 69% respectively.