UncleEbenezer wrote:modellingman wrote:(who spent some of the early part of his career modelling the relationship at UK regional level between daily gas consumption in different market sectors and daily weather).
Interesting. Were you pursuing science as such, or incidentally?
Did you ever look at questions like placebo? People who wouldn't dream of turning the heating on for a cool July day but who do turn it on for the same (or even higher) outside temperatures in this season?
The models were concerned with aggregate demand for gas rather than daily demand at the level of individual consumers.
At the time, around 40 years ago, I was employed by one of the 12 regions of the then British Gas Corporation. The role was to develop and use statistical models (basically fancy multiple linear regression models) relating regional daily gas demand to variables such as weather (broadly represented by daily average temperature and wind speeds), day of the week and season. The models were, naturally, based on historical data and part of the role was to project the parameters of the models to future years. Like all forecasting, this was part science and part art.
In simplified terms, future year models were used to forecast future peak day and average day demands, nationally and regionally, in turn these fed into supply and storage planning - the development of the Rough field as a storage facility was one of the outcomes of the work into which the regional modelling activity provided inputs.
40 years ago, relatively little load research work was done (or practical) at individual household level. A colleague drove through some pioneering work which attempted to measure consumption of domestic consumers on a daily basis. The technology was primitive - a slightly modified index meter with an optical trigger (dot of paint) allowing revolutions of the rotating meter index to be counted. This was linked to a fairly simple electronics board which then wrote on a daily basis the count value to a Walkman type audio cassette recording device. Customers in the panel of around 120 customers (and panel selection was an art and science in itself) returned their cassettes on a monthly basis. It was a favorite activity (not) of new graduate recruits in the team to receive and process the monthly batch of tapes using a specially built tape reading machine. The individual customer units were powered by a pack of 6 D-cell batteries which needed swapping out every 6 months.
The challenge in the 1980s was making sense and best use of limited amounts of data. Technology has moved on considerably since then and, potentially, smart meters provide massive amounts of data. However, the challenges of making sense of these data still remain, though are very different from when I was a practitioner.
It was the behaviour of customers at an aggregate rather than individual level which formed the focus of the work. The effect that you mention was captured, effectively, by differences between models applied to different seasonal periods (typically October-March and aaApril-September). Otherwise, they would be absorbed in the mathematical statistician's get-out clause, the residual.
modellingman