A Power Anomaly Detection Architecture Based on DNN
Wei Geng, Dongyu Liu, Xiu Cao · Proceedings of the 3rd International Conference on Computer Science and Application Engineering · 2019
Different from the traditional power data mining, it is proposed to use DNN (Deep Neural Networks) to deeply process the data, so as to discover power faults and abnormal behaviours in time. The result obtained ensures the normal operation of the power system; according to the characteristics of the user's electricity consumption data provided by the power company, the optimization methods such as oversampling are used to extract the maximum retained data information. At the same time, the accuracy of training is improved and the training complexity is reduced. A deep learning model is used to classify single point anomalies based on time series data. Agile AI (Artificial Intelligence) engineering architecture, offline batch data training and online model real-time detection are combined. Agile AI engineering is constantly improving accuracy. More importantly, the adaptability has been improved to ensure accuracy and efficiency in different scenarios.