An information-theoretic analysis of electricity consumption data for an AMR system

The current trend in the electric utility industry is toward time-varying pricing models where the price varies with demand. This requires automated meter reading (AMR) systems that can obtain accurate data about the consumption of electricity over relatively short intervals, which requires that AMR systems have increasingly higher bandwidth and make more efficient use of available bandwidth. Efficiency of existing AMR technologies can be improved by the use of data compression. Based on data from a large number of electric meters, we estimate probability distributions and correlations of electricity consumption values, and show that the probability density of electricity consumption is accurately modeled by an exponential distribution, and that correlation effects are relatively minor. The information entropy in the data is about 3 bits per value less than the number of bits needed to transmit the values uncompressed. With correctly parameterized Golomb codes, the data can be compressed to a size relatively close to the entropy bound.