This workflow builds an auto-regressive model to predict energy usage. The first week of the time series is used as a template for seasonality correction: the data are differenced by subtracting the values in the same hour in the previous week from the current values. Only past time series are used for prediction. No other external time series/data used. The regression model can be either a linear or a polynomial regression model.
This is a companion discussion topic for the original entry at https://kni.me/w/8ubom9akNs5Dsy8O