2. Design

2.1 Granularity

1. If we are to ignore the knowledge problem, then we need a very (some might say: impossibly) sophisticated model. A model based on aggregates will not do. The devil is in the detail. As Taleb’s aphorism goes: never cross a river that is on average 4ft deep. On average, the UK:
   • temperature is around 9°C,
   • wind speed is around 8.5 knots,
   • insolation is around 900 kWh/m²,
   • electricity consumption is around 33.3 GW and
   • energy consumption is around 183.2 GW.
2. But this is not very helpful for designing a system that will meet people’s needs as supply and demand vary. Never trust someone who is recommending energy policy on the basis of aggregate or average figures. They are either ignorant or concealing something.
3. The granularity matters. Annual figures may be useless, but monthly or daily figures are not much better. Diurnal variations of supply and demand are significant, and storage is a cost (and scarcely available to date).
4. On the other hand, very high frequency (intervals of seconds or a few minutes) is not practical (because the data are not available) and the marginal benefit over an intermediate frequency is small relative to the cost.
5. The only frequencies that provide reasonable granularity and alignment with data, without excessive complexity, are either hourly or half-hourly. Although a lot of electricity data are available on a half-hourly basis, most other energy and weather data are not. A great virtue of hourly figures is that power (MW) and energy (MWh) are aligned, minimising the risk of a common error. We decided to implement our model with an hourly granularity.