Electricity Prediction under Edge Devices Based on Sparse Anomaly Perception

Abstract

Energy issues are closely related to human development. With the changes of the times and the rapid development of technology, power energy has become one of the indispensable energy in human social life, and is the most important part of energy field in modern society. Electricity prediction, as the basis for power system operation, optimization, and control, is facing new challenges in today's rapidly evolving energy system environment. A large number of machine learning technology and deep learning technology have been applied to electricity prediction and achieved good results. In the edge computing environment, anomalous data collection is characterized by sparsity and time window, and machine learning regression algorithms are often affected by anomalous data. Electricity prediction under edge devices based on sparse anomaly perception is proposed, which combines the drop out idea and subsample idea to alleviate the above problems to some extent. And it can achieve faster training and more accurate prediction.