Caltrack handles monthly data by apportioning monthly usage to daily values, then fitting regression models to the daily data. To reduce data processing and computational burden, we would like to consider a methodology that uses monthly data directly, with an adjustment for variable billing cycle lengths.
We propose to include a monthly model specification (hdd, cdd, intercept terms) that uses monthly data directly. Monthly usage is divided by the number of days in billing cycle before fitting the model, to yield average usage per day in each billing period.
hshaban
commented
6 years ago
Caltrack handles monthly data by apportioning monthly usage to daily values, then fitting regression models to the daily data. To reduce data processing and computational burden, we would like to consider a methodology that uses monthly data directly, with an adjustment for variable billing cycle lengths.
We propose to include a monthly model specification (hdd, cdd, intercept terms) that uses monthly data directly. Monthly usage is divided by the number of days in billing cycle before fitting the model, to yield average usage per day in each billing period.