For large scale energy consumers, like a university, the price of electrical energy may be negotiated. The rate depends on the peak power level, with higher rates for higher power levels.
The game runs as follows: the consumer specifies a peak power level for the consumption period to come. This determines the rate to be applied for that period. If during the period the power demand does not exceed this level, the power usage is charged at the agreed rate. If the power demand ever exceeds, a penalty is charged in addition.
To make the game more complicated, the specifies peak power may be increased from period to period, but in general it cannot be reduced without major re-negotiation.
The challenge is to give a useful prediction of the peak power level. Peak power level over time for the Heidelberg science campus, abbreviated INF, are given below.
Electrical energy used for climatization may be major part of the consumption. Available climatic data are temperatures and humidity indices.
One of the most revealing plots is power demand over temperature. We include a lowess smoother, and brush by time
An example of the kind of residuals which can be achieved is shown here.