A linear regression data compression algorithm for an islanded DC microgrid

Ibrahim A. Bello,Malcolm D. McCulloch,Daniel J. Rogers,Ibrahim A. Bello,Malcolm D. McCulloch,Daniel J. Rogers

DOI: https://doi.org/10.1016/j.segan.2022.100901

2022-12-01

Sustainable Energy, Grids and Networks

Abstract:The exchange of data between energy stakeholders will play an important role in future smart energy systems. A key component of smart energy systems is the smart meter, which enables the utility provider to obtain energy consumption readings of customers at regular intervals. When smart meters are fully deployed, it is expected that several billions of data points will be generated per annum, which makes it necessary to compress the data to a suitable size, while preserving vital energy information. In this paper, we present a technique for data compression, which jointly compresses current and voltage time series data by substituting data points with linear regression coefficients. Given a set of input parameters, we demonstrate that the algorithm can also be invoked iteratively to autonomously improve the compression result. We apply the algorithm to load profiles obtained from an islanded DC microgrid laboratory experiment, and we demonstrate that our technique has a near-lossless compression performance and a high compression ratio of more than 50-to-1 for most of the datasets considered. As a proof of concept, we also apply the algorithm to energy data from an AC-grid-connected household and the results suggest that the developed algorithm is potentially applicable to more conventional energy systems. We compare our results with a similar linear-regression-based algorithm and our proposed technique demonstrates a better energy error performance on average for a given compression ratio and a 30-fold reduction in the data compression time.