In this talk I'll discuss my group's recent efforts to understand the impact that distributed photovoltaic (PV) generation will have on intra-hour grid operations in future scenarios of very high renewables penetrations.  The analysis is based on a unique data set of PV production and a new machine learning approach to modeling PV variability.  We find that the amount of intra-hour balancing capacity required depends very strongly on the arrangement of PV systems across large geographic scales.  In the second part of the talk I'll discuss how we might adapt the grid to accommodate these increasingly large balancing requirements and the role of optimal control in the future of power system operations.