A three-scale system
PV production forecasting: a world first for Steadysun!
December 18th, 2015
ERDF (France’s main grid operator) and start-up STEADYSUN are innovating advanced technologies to facilitate the integration of PV solar into electricity grids. GreenLys, the first full-scale smart grid demonstration project in France, is currently underway in the greater Lyon, France area. The most novel approach currently being tested entails installing two “cloud cameras” that monitor the sky to provide citywide PV production forecasts for the next half hour.
The installed base of PV equipment on France’s national grid increased by a factor of 200 from 2008 to 2015, rising from 30 MW to 6,100 MW of total power. And this figure will increase further, driven by new energy-transition and sustainable-growth legislation, which, if passed into law, will set targets of 23% renewables in the final energy mix by 2020 and 32% by 2030. Renewables currently account for just 15% of the energy mix. The resulting rise in PV equipment on France’s national grid is, of course, a key challenge for grid operator ERDF.
The widespread adoption of solar energy will depend on factors like the ability to accurately forecast production. It is also crucial to ensure that PV energy can be integrated into the grid without compromising service. The GreenLys project has resulted in the development of a truly unique innovation: real-time local (citywide) PV production forecasting based on cloud observation.
Lyon, France, a living lab for PV forecasting
The experiment covers more than 300 PV plants in and around Lyon, France. Two fish-eye cameras provide non-stop 360-degree images of the sky. STEADYSUN’s software analyzes cloud movement and uses the information to predict fluctuations in PV production for the next half hour.
At the same time, Linky communicating electricity meters, also being tested under the GreenLys project, gather and transmit data from each PV installation. The data are aggregated and integrated into the forecasting model, substantially increasing the accuracy of the forecasts. In traditional systems, forecasts are based on the plant’s peak power and meteorological data. STEADYSUN’s advanced technology rolls in a self-learning algorithm, co-developed with CEA at INES (French National Institute for Solar Energy). The algorithm uses actual production data from the previous day to improve the day’s production forecasts, so that forecasts actually get better over time. And STEADYSUN offers several time horizons based on different forecasting methods.
Having such accurate information updated in real time—a technological revolution—will facilitate the integration of renewables into the grid. Grid operator ERDF will be better able to manage production peaks distributed across the grid to anticipate any issues before they arise. This information could also be used to match electric vehicle charging demand to PV electricity supply.
A three-scale system
Forecasts are generated at three scales:
– For each PV plant
– For each LV (low-voltage) network, with a group of several plants spread over an area of several square kilometers
– For the HV (high-voltage) network, with an aggregated sampling of several plants spread over several dozen square kilometers
This multi-scale approach is crucial to grid operator ERDF to optimize management of the grid at each scale (LV or HV) while integrating as much renewable energy production as possible.