Analysis of Weather, Time, and Mandatory Time-of-Use Pricing Effects on Aggregate Residential Electricity Demand
Catherine Rosenberg and Lukasz Golab (Management Sciences)
Residential electricity demand is affected by three types of external factors: weather, time, and the price of electricity. We implement a data-driven study using multiple linear regression to describe the effect size associated with these external variables. Our treatment of temporal variables such as month, working days, and hour-of-day are modelled in a manner consistent with prior work. However, previous studies vary widely in how the effects of temperature and weather are modelled. In temperate regions like Ontario, the relationship between residential electricity demand and temperature is notably non-linear across winter and summer seasons. We compare linear break point and natural spline forms for modelling non-linearity. We also study variable transformations that take into account the heat transfer time delay before a household's thermal controls react. We compare several forms of temperature's interaction with time such as lagged observations, rolling average, degree-hours, and exposure-lag-response. Finally, some studies choose to use dry-bulb temperature observations while others incorporate coincident weather variables such as humidity, wind chill, or solar irradiance. We compare the use of dry-bulb temperature, wind chill, heat index, and qualitative weather conditions. We report the explanatory power, out-of-sample prediction accuracy, and discuss impacts on model interpretability for these modelling choices using a data set of hourly smart meter readings in south west Ontario.
Additionally, Ontario has mandated time-of-use pricing for all residential customers in the province. We carry out a case study to quantify the effects of mandatory time-of-use pricing on residential electricity demand in south west Ontario using a well-performing, descriptive model chosen from those that were compared earlier. The time-of-use pricing policy is evaluated according to two of its stated objectives: energy conservation and shifting consumption out of peak demand periods. We show that during the summer rate season, time-of-use pricing is associated with electricity conservation across all price periods. The average demand change during on-peak and mid-peak hours is -4.3% and -2.4% respectively. Change during off-peak hours and non-working days is -0.6% and -0.3% but is not statistically significant. The uneven structure of electricity conservation shows that demand is more readily shifted away from peak periods. The peak-to-average ratio, a separate metric to measure shifted electricity demand, decreased 2% from 1.44 to 1.41 under time-of-use pricing.