Advances in Compression of Power Quality Signals

Abstract

The emerging technology of smart grids relies heavily on monitoring the distribution networks for disturbances using, for example, the devices installed for measuring power quality signals. Obviously, efficient compression of these electrical signals is of paramount importance to allow fast transmission, remote analysis and automation of response to disturbances, as well as archival storage. While prior approaches to compress electrical signal disturbances employed standard effective components such as wavelet transforms, non-uniform quantizers, and entropy coding, the overall system design was largely ad-hoc in the sense that it did not directly account for or adapt to data statistics. Instead, we propose to jointly design all system modules, including transforms, quantizers and entropy coders, within a genetic algorithm-based optimization framework, while accounting for the variation in statistics across different disturbances, within a two-step “classify then compress” procedure. Specifically, we jointly design the family of wavelets to be employed, the non-uniform quantizer structure, and probability tables for the entropy coder, to optimize the rate distortion trade-off. Experimental results for 8 classes of commonly occurring power quality disturbances, which were synthetically generated to ensure rich and comprehensive training and test sets, validate the effectiveness of the proposed approach with significant performance gains over prior techniques.