Article published Forecast
What sort of sunshine will we have tomorrow
The proportion of solar power on the electricity networks could increase thanks to a forecasting tool from researchers at LITEN.
By definition, photovoltaic energy is intermittent and random: when the weather is cloudy, production drops or possibly stops altogether. This is not a problem if the quantity of electricity produced this way is marginal. However, if it represents more than 30 % of energy consumed, which will increasingly be the case for the island networks on Réunion, Guadeloupe and Corsica, then management becomes difficult. As well as energy storage solutions which are still under development, forecasting could help to relieve this problem. “We have developed a tool which forecasts photovoltaic energy production 24 hours in advance, with a half-hour time-step”, explains Xavier Le Pivert, manager of the photovoltaic systems production at LITEN.
How can a phenomenon that appears random be predicted? By using mathematical learning methods. By comparing, actual power production over a period of several months with the forecasts made using data on local meteorology, panel orientation and system losses, the software "learns" and corrects itself as time goes on. The mean error recorded is no greater than 6 to 8 % on the two sites tested. For the time being, the data are processed at Ines*. However, in the long term the system will be completely automated. “We will be offering this service across the whole of France, starting this summer”, announced Xavier Le Pivert.
* National Solar Energy Institute
