Substantial variability of household energy consumption is essential to recognize. In particular the variability occurs towards large magnitudes. This work focuses on modeling the distribution of household annual energy consumptions and promoting a practical modeling framework resulting to explicit models. Lognormal distributions describe well energy consumption of households that do not use electricity for heating. We model at city district level how consumption distributions depend on district housing characteristics. At individual household level, the joint distribution of the annual electricity consumption and key household characteristic (number of persons, floor area, number of bedrooms), is well approximated by multivariate lognormal distributions. This work utilizes numerous data sources: household annual electricity consumption data for city districts, city summary data on housing and people, data on sold apartments, and Irish Automatic Meter Reading (AMR) data with test survey information. We also demonstrate a process for adjusting the models when data relating individual household energy consumption with other characteristics are not available. (C) 2015 Elsevier B.V. All rights reserved.