Even if energy demand were linear to weather...

Let us suppose that heating energy demand to keep a building warm really was linearly dependent on external temperature below some specific condition.

Would this mean that energy consumption was linearly related to external temperature (eg double the degree-days - and you double the weather sensitive consumption) ?

Note this assumption already requires that we are ignoring a lot of reality :

  • that there is no thermal mass to gradually cool at the end of summer, 
  • that internal heat distribution is even and rock steady (so no reduction of temperature comfort targets overnight
  • that the building performance does not change from hour to hour (constant occupancy, constant activity - even at weekends)
The obvious answer is that this is at best a very weak approximation, but let's go a few steps further.

Do we assume finite plant capacity is relevant ?

If it keeps on getting colder, the demand for heat rises. However, in many real world situations parts of the building start to get cold (the attic), the west wing etc etc. as they start to get cold, comfort deteriorates, and consumption falls, because a cold building loses heat less fast than a hot one. If the plant ever gets to absolute maximum rated capacity, we can safely assume consumption will stop rising !

Do we assume there is diurnal (day/night) temperature variation ?

If we assume that at some time of day, maximum plant capacity is met, it does not mean that maximum plant capacity is met all day.  So actually consumption flattening with near maximum capacity, occurs well before a building reaches its design limit.

Do we assume there are time limiting controls ?

As soon as we allow for consumption to be limited to certain hours of day, the time distribution of demand is relevant.  Therefore Degree-days (which inherently smooth out timing by creating a single demand figure for a day) must fail to capture the information.

Do we assume that plant efficiency is constant ?

Most heating and cooling plant is most efficient at high load - at times of lower load it is less efficient, therefore the steepness of the real relationship between weather load and consumption varies as the efficiency to weather load varies.

These are just a handful more reasons why degree-day analysis should be handled with care.