Solution & Technology
2015-07 / STEADYSAT / ISLAND ELECTRICAL SYSTEMS
7 Juli 2015
Satellite based solar forecasts and variability over time: the case of a 70kW power plant at the Martinique Island (Caribbean).
In non-interconnected areas such as islands, the impact of solar production on the electrical grid frequency stability is quite critical. A short term production forecast, typically over one hour, is a solution allowing a larger integration of this energy while keeping a sufficient security margin.
Our technology steadySat, which is based on satellite imagery, is part of the solution. However, one should keep in mind the limits of this solution, particularly when applied to one single solar plant.
The figure below shows the forecasted and measured solar productions on July 7, 2015 for a small 70 kW plant at the Martinique Island. The time horizon is 1 hour, and the time step of 1 minute.
Rapid production fluctuations are observed, which often outrange the confidence interval of 90% (upper yellow curve).
This situation is not so surprising since the pixels of the satellite images are covering an area of a few square kilometres each. As a result, each pixel represent a spatial averaging of the cloud situation in this area (over several square kilometres) while the power plant is affected by a little portion of this surface, in the order of 1000 square meters (that is to say 0,001 square kilometres).
Figure 1: Production (in Blue) vs Forecast (in Red/Orange, with the percentiles / confidence interval at a 10% step) – Time step: 1 minute
When one switch to a 30 minutes time step, the averaging of the expected production over the new time step, partially compensate for the above mentioned effect. The average production over a 30 minutes time step is, then, quite well anticipated, and remain in most of the cases within the uncertainty interval (the lower and higher curves representing respectively the 5% and 95% percentiles) as shown in figure 2 below.
Figure 2: Production (in Blue) vs Forecast (in Red/Orange, with the percentiles / confidence interval at a 10% step) – Time step: 30 minutes
The actual question is to define what level of timing accuracy did one really need to address the operational requirements? One can, then, define the adequate forecasting solution. For instance choose to operate with the satellite technology – steadySat – with a 30 minutes time step, or prefer an on-site ground based sky imager –steadyEye – with a one minute time step.