Explainer
Every electricity supplier has to predict how much its customers will use, half hour by half hour. Get it wrong and the difference is settled at the imbalance price. Here is how demand forecasting works, what drives the errors and what market wide half hourly settlement changes.
By Christopher Fagan, Managing Partner, PEF Consulting Partners
Demand forecasting is predicting how much electricity a group of customers will use in each half hour settlement period. Suppliers use it to decide how much energy to buy in advance, to hedge, and to set their contracted position before each period closes. NESO, the National Energy System Operator, forecasts national demand to balance the system, but every supplier has to forecast its own portfolio.
In Great Britain each supplier’s contracted volume is compared with its metered volume for every settlement period. The difference is its imbalance, settled at the imbalance price, which can be far more expensive than energy bought in advance. Small percentage errors across a large portfolio add up quickly and fall straight to margin. Our guide to energy imbalance and settlement explains the mechanics.
Most suppliers combine a statistical or machine learning model with weather forecasts and their own historical consumption, then adjust for portfolio changes and known events. The model matters less than the data behind it: an accurate customer list, reliable meter data and a clear record of who joined and left, and when.
Under market wide half hourly settlement every meter point settles half hourly, using actual readings wherever smart and advanced meters provide them. That gives suppliers far better data to forecast with, but it removes a cushion: errors that used to be smoothed by profiling show up as real imbalance, and from 2 July 2027 they are settled on a four month timetable instead of fourteen.
Predicting how much electricity customers will use in each half hour, so a supplier can buy the right amount in advance and avoid imbalance charges.
Every supplier forecasts its own customers. NESO, the National Energy System Operator, forecasts national demand so it can balance the system.
The difference between what it contracted and what its customers used is settled at the imbalance price, which can be much more expensive than buying in advance.
It gives suppliers half hourly data for every customer, which should improve forecasts, but it removes the smoothing of profiles and shortens the settlement timetable, so errors cost money sooner.
Related reading: BM Units and the balancing mechanism, electricity settlement performance, MHHS readiness and compliance.
If imbalance costs are higher than they should be, we can trace them back to the forecasting, data and settlement errors driving them.