A Simple Method of Non-Intrusive Load Monitoring Based on BP Neural Network

Abstract

Non-Intrusive Load Monitoring (NILM) is a method to breakdown the total power consumption into individual application, and it has great value for improving energy utilization. Deep-learning methods on high frequency data usually have a complex structure. Here, we use BP network with single hidden layer to achieve the task of NILM and get great results. The methods in this paper include event detection, feature extraction to get event-based differential current, data augmentation and application classification. BLUED dataset is used for this experiment. Our method has a good performance in BLUED dataset, with above 95% accuracy on phase A and B for application classification.