Portfolio item

Yield Prediction

Many companies have invested in green energy, mainly solar energy but also wind energy. This additional source of energy is clean, cheap but varies from one moment to the next. The figure below shows the mismatch between energy production and consumption reduces your profits. The left panel displays a typical shape of energy production due to solar panels, which peaks around 2pm, when the sun reaches is highest altitude. The middle panel shows an example of an energy-consumption profile in a company: there is always some consumption, but whenever heavy machines are turned on, an high amount of energy is required. The left panel shows that, due to the mismatch between both, sometimes excess energy has to be inserted into the grid (orange), and additional (more expensive) energy has to be bought at other moments. It would be beneficial to match the energy that is consumed in the company to the energy that is produced.

Energy production by solar panels for example, is mainly dependent on seasonal effects (the time of the year) and meteorological conditions as for example cloudiness. Seasonal effects can be estimated based on historical data, while meteorological conditions can be estimated by weather forecasts. Admittedly, the accuracy of these forecasts increases for forecasts that are in the near future, but still, very accurate predictions can be made a few days in advance and can help your company schedule its electricity consumption.

The figure below shows some simple insights: given a certain date in the year, an average yield can be estimated (black line.) This yield will be lower in the winter compared to the summer. However, given a date in the year, there is still quite some variability, which we can predict using weather forecasts. For example, including only cloudiness in the model, can narrow the yield-estimate considerably.