Modeling and disaggregating hourly effects of weather on sectoral electricity demand

Abstract

Maintaining a reliable energy system requires accurate forecasts for energy demand in both the short and long term. The wealth of smart meter data stored by most modern utilities provides an important resource, however sub-metering of specific end-uses is still fairly limited. This makes it more difficult to assess the impact of new or different customer energy uses on aggregate demand curves. This paper presents a regression model to predict the impact of weather on aggregate sectoral electricity demand. The model enables disaggregation into three specific demand categories: base demand, heating demand and cooling demand. To improve the accuracy of predictions, the model uses temperature data at multiple temporal resolutions, varying changepoint temperatures where customers are assumed to switch between heating and cooling, and Probit analysis to model the use of portable air conditioning units. The models developed for the residential and commercial sectors showed good fits (coefficient of determination of 0.9710 and 0.9790 respectively), however had difficulty modeling the cooling demand. The disaggregation method showed promise when compared to data from another study but requires further validation. Once validated, these models could be applied to assess the impact of climate change and changing technologies on each sector.