This paper proposes a model for short-term electricity load with differentiated temperature and daylight effects by time of day, which are determined by variations in intraday economic activity. The relationship between electricity load and economic activity implies that the electricity demand response to changes in exogenous variables like temperature is non-linear as well as non-homogeneous along the day. The proposed framework, a smooth transition regression model with double threshold (LSTR2), models the observed intraday patterns in load curves to explicitly capture the effect of the circadian rest-activity cycle on the distinct responses of electricity demand to temperature and daylight variations throughout the day. The model shows that the sensitivity of demand to low temperatures is significantly larger in the “active” compared to the “rest” state. If temperatures decrease from 10 °C to 0 °C, electricity demand in the “active” state increases by 960.5 MW h per 1 °C decrease, but by only 26.6 MW h per 1 °C decrease in the “rest” state. When temperatures are higher, in the “rest state” demand decreases by 602.9 MW h per 1 °C if temperature falls from 26 °C to 21 °C, while in the “active” state demand only decreases by 323.6 MW h per 1 °C variation