Predicting energy poverty with combinations of remote-sensing and socioeconomic survey data and spatial analysis