Prediction model for photovoltaic installations

The Luxembourgish researchers have been working on two levels: On the one hand, they calculate the performance a given installation can attain at a certain time and location. These predictions are based on weather, insolation and temperature forecasts. This data is combined with the researchers’ knowledge about the installation itself: direction, inclination, age, model…

On the other hand, the researchers got feedback from reference installations in order to compare the predictions with the real performance. Whereas the predictions are relatively easy and reliable when the weather conditions are good, the challenge is even bigger when the weather is changeable. The small surface of the country makes extrapolations difficult and increases the fluctuation risk. An unpredicted cloud plays a big role in Luxembourg and even more so for a calculation lap of 15 minutes. Such phenomena can be more easily extrapolated for bigger regions.

That’s the reason why the model aims to adapt the prediction in comparison with the real performance. This approach makes the project unique. Although the approach in this project is not yet fully developed, the results are promising. Site specific forecasts show good trends, compared to measurement values.

In the medium term, precise predictions will be important for energy suppliers and grid operators. With the growing part of renewable energies in the grid, suppliers will have to take them into consideration while completing their electricity needs on the intraday market. Grid operators have to guarantee the stability of the grid. Therefore it’s important to know in advance which kind of power source will be fed into the grid and where.