International Space University looks into Weather Forecasting and the Power Industry
This summer the International Space University brought its Summer Space Studies Program to the Netherlands. Partnering with Delft Technical University, Leiden University, and ESA ESTEC, hundreds of space exploration experts visited the Netherlands to exchange knowledge with participants from around the world. One of the their projects involved looking at the relationship of between weather forecasting and the power industry. International Space University students we able to have a discussion with meteorologists from MeteoGroup.
Weather forecasting and the power industry
Energy providers are very dependent on accurate weather forecasting. Mid-term trends can be estimated quite accurately using prediction models and allow good scheduling. However, major problems are the unpredicted peak loads. These unpredicted peak loads force energy providers to procure capacity from other sources whereby the cost has a quasi-exponential function with time. In other words, the earlier such peak loads can be known, the lower the cost of procuring energy from third sources to compensate for it. A typical example is the prediction of clouds around 17.00h when a high number of people return home from work.
At certain periods in time during the year, heavy clouds can lead to the rather simultaneous switching on of lights. This usually causes a considerable peak load. Power companies need to provide the right amount of electricity. Hourly and every day. This means they have to predict power consumption. Any misprediction means they will have to buy more electricity on the spot market at high prices or sell surplus electricity at low prices. The overall cost they incur here is called “variance charge”. It is therefore in the of interest of energy providers to have more accurate prediction models on weather conditions, with a very high granularity such as hourly updated predictions.
It is said that about 94% of relevant data for weather forecasting comes from satellites. One would expect that an increase in the number of satellites would immediately result in an improved quality of forecasts. Weather is a classic application where a large number of satellites could then have a sustained competitive advantage in describing the starting condition of the atmosphere more accurately than fewer satellites.
Consequently, the more that is known about the starting conditions, the better the forecast. In particular, GNSS-RO provides a highly promising modality of earth observation which - when created in large quantities - could dramatically increase the accuracy of weather forecasts.
Are more satellites better?
To check these assumptions regarding the effectiveness of satellites in real-life, the project team of the International Space University contacted MeteoGroup to discuss its assumptions and first findings with Wim van den Berg and Willy Zittersteijn. Both of whom are experienced meteorologists at MeteoGroup.
According to Wim van den Berg, the benefits of more satellites is only part of the story: “Of course, a lot of the data that weather models use to capture the starting state of the atmosphere comes from satellites. However, this data still requires additional surface data (land andocean) for calibration as well as a perfect data assimilation system to process all those types of data. In addition, all these data record the starting state of the atmosphere, after which the model has to start calculating and processing. The first results of all those calculations are only available as a forecast for at least several hours after the satellite observation time. In the meantime, the exact location of the clouds might have changed again." He adds: "It’s interesting to see what ECMWF says about it:
Take a look at this complex process of data assimilation in the picture below (provided by ECMWF).
“I’m not sure if adding additional satellite sensors gets us much further in the forecasting of individual clouds or storms, Personally, I think that we have to look towards faster processing of all this data. We should use more parallel processing and calculation power to assimilate (in near real-time) the starting state of all that data and satellite data in a weather mode. This is more important for the 15-minute forecast than having more (of the same) data without that computing power.”
His colleague Willy Zittersteijn agrees: “I think that getting a better picture of the situation as it is now, is more important. A higher resolution of good data (for example better data processing on the ground to create a more detailed picture of wind speeds) could be very interesting, e.g. in the energy market, for Dynamic Line Rating and Hotspot Analysis in energy grids. For that kind of application, the local wind speeds play a very important role.”
I’m not sure if adding additional satellites gets us much further in the forecasting of individual clouds or storms. Personally, I think that the key lies in the faster processing of all the data they provide.