Development of the series of probabilistic statistical models for electricity demand prediction in residential communities

This study developed a series of probabilistic statistical models for electricity demand prediction of residential communities. The series of probabilistic models were developed to reflect individual variations in the electricity demand depending on household characteristics and temporal variability in the pattern of hourly electricity use. We used the hourly electricity data, including plug-in and lighting energy use, from 23 households selected from the public data of the Korea Energy Agency. The prediction model consists of four models to capture variability in the electiricity demand at different indiviual and time scales. Models 1 and 2 are blinear regression models that predict the annual average electricity load depending on the household characteristics and variation in the daily electricity load, respectively. Models 3 and 4 are multivariate normal distribution probability density functions that generate average hourly electricity load profile and temporal variations from the average profile, respectively. The results demonstrarate that the series of probabilistic models sufficiently reflect actual individual and temporal variations.