Fast Cloud Transients
One of the thorniest issues that seemed to be facing large, utility-scale PV systems was sudden changes in PV system output due to the movement of single clouds on an otherwise sunny day. Now it seems that this issue has been resolved favorably, in an economical manner.
Figure 1 shows a particularly excruciating example of solar variability.
This kind of sunlight graph put chills down the spine of utility engineers and caused a major setback for the deployment of large-scale PV systems. PV output responds instantaneously to such changes. Such variability also compares unfavorably to solar thermal electric (STE) systems, which have natural thermal inertia, which damps out the sudden changes. STE system installers gladly took note and publicized the PV problem to everyone making a choice between STE and PV.
Now we have good news from Clean Power Research from principals Tom Hoff and Richard Perez. There are simple, economic solutions to fast cloud transients, as implied in their report and presentation. The solutions are either
- Geographic separation for many small systems; or
- Very large fields, for large systems.
In the first case, the passage of individual clouds do not correlate, so the impact on the aggregated power is small. In the second case, the clouds are smaller than the array (which, for example, might be five by five miles, as in the recently announced agreement between First Solar and China for a 2 GW system). Being so small versus the array, the single-cloud impact is minimal (see our slide show). In fact, the more typical situation might be scattered clouds, with small parts of the array always affected by an average loss, depending on the local weather pattern. This is much more predictable than single clouds and so can be compensated for by conventional back-up power, e.g., from gas turbines.
This means that the issue of fast cloud transients can be solved by simple strategies that themselves do not add much or anything to the cost of PV. Only for isolated systems of a certain size (about the size of a system that can be covered by one fast moving cloud) is the variability a major issue. And even then, if that system is part of a regional network of similar systems, its variation will be a small percentage of the whole, and this average variation will be predictable.
Being predictable is key, since by including solar forecasting with these systems, it is then possible, with a minimum of losses, to back up the solar PV successfully and inexpensively.
I am sure Tom and Richard would caution me that these are preliminary results. Different cloud sizes, speeds, and assemblies are likely to drive the variability in different directions. Can this also be satisfactorily modeled or predicted? No doubt there is much to do. However, it appears that what was once seemingly intractable appears open to adequate solutions.
Clean Power Research has made a major contribution to our knowledge at a sensitive time. Deployment choices, policy, and government decision-making are being influenced by this question, which until now seemed like a possible show stopper for large PV systems.