P8
By Magnus Dahl, industrial PhD fellow at AffaldVarme Aarhus and Aarhus University
I have applied this technique from meteorology to heat demand forecasting. I have obtained an ensemble forecast of the weather in Aarhus: 25 slightly different weather forecasts. This is a standard product sold by the Danish Meteorological Institute DMI as well as many other weather services around the world. The 25 forecasts are then fed through a model for predicting the heat demand based on the weather. The result is 25 slightly different heat demand forecasts that diverge and converge according to how uncertain we are on the weather forecast. Figure 1 illustrates this. Each grey curve represents the heat demand forecast in a different possible weather situation. The blue curve is the best prediction of the heat demand, based on the average weather forecast. With the ensemble heat demand forecasting tool in hand, we can begin to distinguish situations in which we are very certain of the heat demand and situations where our predictions could be far off.
It is possible to ensure security of supply while reducing economic risk. District heating production planners who are aware of the changing uncertainty in their forecasts make better decisions. In the following, I present a way to assess weather- related risks based on methods from weather forecasting. Any district heating production planner or system operator knows that heat demand forecasts are far from perfect. This is especially true for forecasts that reach more than a few hours into the future. When trading on the day-ahead electricity markets, production planners need forecasts for the following day. Forecasts on this time horizon often differ significantly from the realized heat demand. Therefore, creating realistic production plans and executing them is always a challenge. Both researchers and private companies are making efforts to improve heat demand forecasts. However, there is a limit as to how accurately it is possible to predict the heating demand of a city a few days into the future. This is why we at AffaldVarme Aarhus work actively with researchers from Aarhus University. Together, we assess the level of uncertainty that we are facing in the production planning and daily operation of the Aarhus district heating system. Since our knowledge of the future is never completely certain, we continuously estimate how uncertain our predictions are. When we are very certain of our predictions, it allows us to put more at stake in the decisions we make. When we are very uncertain, we need to exercise caution. Better knowledge of the operational risk we are facing leads to better decision making. Using more weather data The weather is hard to predict, and as a consequence tomorrow’s heat demand is hard to predict. Errors in weather forecast can carry over and produce erroneous heat demand forecasts. Therefore, I have developed a way to estimate the weather-based uncertainty in a heat demand forecast. I use a technique, adopted from weather forecasting, called ensemble forecasting. Meteorological institutions have used ensemble forecasting to estimate the uncertainty of weather forecasts since the early 1990s. The concept of ensemble forecasting is quite simple. Instead of creating a single weather forecast, meteorologists generate many slightly different forecasts. They generate these forecasts in such a way that their spread reflects the level of uncertainty. This means that when the forecasts are all very similar, we can be very sure of the predicted weather. Conversely, if the forecasts are far apart we are uncertain of the weather.
Proof of concept – lowering supply temperatures In a first proof of concept application, I have demonstrated how ensemble heat demand forecasting can be used to control the supply temperature from an area substation (figure 2). Ensemble heat demand forecasting makes it possible to adjust the temperature in a smart way depending on the weather- based uncertainty of the forecast. When the heat demand forecast is very uncertain, we regulate the temperature more conservatively. When our confidence in the forecast is high, we can lower the supply temperature without gambling with the security of supply. Reducing supply temperatures in a district Figure 1: Heat demand forecast predicting the total heat production in Aarhus. The spread of the 25 forecasts (in grey) indicates how certain the forecast is. In high-risk situations, production planners and system operators have to be careful in their decision-making. In low-risk situations, it is possible to maintain security of supply while making more economically favorable decisions.
E N E R G Y A N D E N V I R O N M E N T
Made with FlippingBook Ebook Creator