A colleague has reached out to me with an interest in using LEAP to determine if, and when, renewable electricity generation capacity, and particularly solar PV capacity, has risen to a level where the grid in a given area cannot provide peaking power fast enough as the renewable generation decreases, for example, as the sun goes down during a summer peak. It seems to me that the load shape function--either expressed for the entire electricity load in an area or using individual load shapes on a demand-device by device basis--could be used, along with load shapes that describe the availability of solar PV power and other generators on the system. Has anyone else done or heard of anyone else doing this kind of analysis with LEAP? My second (related) question is, when modelling the growth of renewables (especially solar), what LEAP outputs would one used to be able to identify when peaking power resources become insufficient under particular generation expansion plans? Thinking out loud, I guess if one did modeled capacity and demand with yearly shapes (in an extreme example, on an hourly basis), one would be looking through the results of LEAP runs for periods of capacity shortfall, as indicated by negative reserve margins or generation shortfalls, and looking at the timing of same. Then the high-renewables scenarios would be modified to change the electricity resources available (for example, by providing peakers or storage) in order to bridge those gaps. Any feedback on this issue would be appreciated!
These are good questions, and ones that we're increasingly addressing in the LEAP modeling we're doing at SEI. With regard to the first, we have used timesliced availability curves for PV and other generators, and coupled them with load shapes to explore how other resources must backfill when the sun goes down. We've built models that respond to the intermittency of PV (and other variable renewable energy sources) with a combination of other generation (e.g., gas), storage (using a customized version of OSeMOSYS and some other enhancements), and imports. This modeling has generally used hourly timeslices, so it's been fairly finely resolved.