What Is a Heat Demand Forecast?

A heat demand forecast predicts how much heat a district heating network or a group of buildings will need in the coming hours and days. It is usually expressed as heat load in MW, in hourly or 15-minute steps.

How a heat demand forecast is created

Heat demand depends above all on the outdoor temperature, but not only on the current value. Buildings and the network itself store heat, so demand reacts with a delay; good forecasts therefore use smoothed or lagged temperatures. Other factors are:

• Wind and solar irradiance, which change heat losses and passive solar gains of buildings.
• Daily and weekly patterns from hot water use and heating schedules, plus holidays.
• Network operation, for example the flow temperature, which affects losses in the pipes.

Classical approaches use heating degree days and regression on temperature. Today, statistical and machine learning models are trained on the measured network load, so they learn the individual behaviour of each network, including the transition seasons in spring and autumn when simple temperature models often struggle. As with all weather-dependent forecasts, the quality of the input weather forecast matters, and several weather models together are more robust than one.

What heat demand forecasts are used for

Utilities use heat demand forecasts to schedule CHP plants, boilers, heat pumps and storage, to procure fuel and to nominate the electricity schedules of their CHP units. A heat forecast error therefore often becomes a gas or electricity deviation that has to be balanced.

In a benchmark for a district heating network in the Rhine-Main area, our forecast error (MAE) was about 21 % lower than that of the existing system of an established EDM vendor. You can read more on our load forecast page, which covers electricity, gas and heat.