On the Application of Parsimonious Periodic Autoregressive Models to Bursty Impulsive Noise in Low-Voltage PLC Networks

Abstract

This paper proposes Parsimonious Periodic Autoregressive (PPAR) models for modelling the bursty impulsive noise present in low-voltage power line communication (PLC) networks in the frequency range of 1 – 30 MHz. The acquired impulsive noise time series is seasonal and exhibit an autocorrelation structure that depends not only on the time lag between observations but also the season of the window length period of measurements. Assuming the seasons are grouped into groups of one or more seasons with similar autoregressive (AR) characteristics, individual AR models for various seasons are combined to obtain a single model for all seasons in a given group. Consequently after grouping, the parameters of the more PPAR models are estimated and diagnostically checked, validated through measurement data acquired from the University of KwaZulu-Natal and compared to other periodic time series models.