Data-Driven Analysis of Time of Day Pricing for Residential Consumers

Abstract

Time-of-day (or dynamic time-of-use, dToU) pricing is a mechanism by which system operators try to lower stress on the grid in times of high demand. The price for high demand periods is pre-set but the times of day they are applied is dynamic. Data on how residential consumers respond to the pricing scheme can inform more accurate models of consumption to maintain the integrity of the grid while lowering consumers' utility bills and optimizing renewable use. In this thesis, I analyze the data from a time-of-day pricing trial in London to see whether the treatment was effective in lowering consumption. I do this analysis using four different models and compare the accuracy of each and the results; an aggregated linear regression model, a multi linear regression model, an aggregated multi linear regression model, and a random forest regression time series model. I found that the time-of-day pricing during the trial was effective in lowering consumption and costs. A dependence on households' socio-economic status was observed.