Quantifying the Impact of Daylight Saving Clock Changes on Energy Consumption

Methodologies currently used in load prediction are adapted to study the effect on load demand were the UK to move from GMT to BST (=GMT+1) during winter. In particular Support Vector Regression (SVR) is used to estimate energy demand on a half-hourly basis (given prevailing light and temperature). This resulted in predicted lower bounds on daily savings of 0.32% in November, 0.22% in December, 0.32% in February, and 0.32% in March. In terms of actual power involved this translates to approximate savings of 6.6GWh, 4.8GWh, 6.7GWh, and 6.2GWh on average over an entire day in those months respectively. Peak energy demand is investigated further and in comparison with existing methods for this analysis (linear and polynomial regression) we found that SVR with a Gaussian kernel significantly outperformed in terms of regression residuals. We concluded that peak savings could range from 0.5% in December to slightly more than 4.0% in March.