Abnormality detection of large difference electricity meter data under the prediction of change patterns
During the long-term operation of smart meters,there may be data loss and expansion of data range distribution,re-sulting in low detection accuracy in the face of large differences in meter data.For this reason,a method based on convolutional neu-ral networks(CNN)and edge computing is proposed to detect the abnormality of meter data.In the edge computing mode,collect the meter data.Fill in missing values based on the time continuity and periodicity of the electricity meter data to ensure data integrity.Standardize the filled electricity meter data to ensure that data with different features have similar ranges and distributions.Predict the variation pattern of electricity meter data,extract useful features from it,fuse vectors of different features,and form a comprehensive feature vector;Using an asymmetric convolutional kernel based convolutional neural network(AC-CNN)model,real-time process-ing is performed on the synthesized feature vectors and abnormal features are detected.According to the experiment,it can be conclu-ded that this method can reduce detection errors and effectively improve the accuracy and efficiency of anomaly detection for electricity meter data.
convolutional neural networkedge computingfilling in missing valuesfeature vectorasymmetric convolutional kernel
万方数据
维普
