Online Load Data Compression and Reconstruction Based on Segmental Symbolic Aggregate Approximation

Abstract

To ensure the performance of load identification, existing non-intrusive load monitoring (NILM) solutions usually need to transmit large amount of continuous power consumption data for cloud analysis, which brings challenges to the communication, storage and computing of NILM systems. In this paper, we propose a segmental online compression and reconstruction method of load data to reduce the data volume for transmission while ensuring the application value of the data. Due to the fact that different segmentations of the load data contain different amounts of information, the time series load data is firstly divided into event segments and non-event segments (also called steady-state segments) by event detection. While the data is retained as it is for the event segment, the symbolic aggregate approximation (SAX) method with non-uniform partition of the time axis is adopted for data compression for the steady-state segment, so as to reduce data size for transmission as much as possible without losing important information. In the master station, the load data including the event and steady-state segments received are respectively reconstructed into original time series with fine resolution in different ways. Comparative experiment results on the private and public datasets show that compared with the existing methods, the proposed method has higher reconstruction accuracy and compression efficiency, which can provide support for the data analysis applications based on high resolution data.