Statistical consumer modelling based on smart meter measurement data

This paper presents an approach for modelling residential load profiles based on measurement data of 291 smart meters over the period of one year. It is shown that the generalised extreme value distribution describes the distribution of the occurring loads for every point in time very well. The parameters of the distribution function depend on the mean power value of all smart meters in each point in time. For modelling realistic load profiles the consumers are assigned to different states, which describe the load within one point in time in relation to the others. By introducing transition matrices for the individual consumers with subdivided time ranges, temporal state changes can be simulated. Finally the synthetic load profiles are generated using Markov chains, based on the transition matrices and the distribution functions.