I wanna deal with an uncertain hourly demand via Robust Optimization.

Since the demand is one-dimensional data, constructing box uncertainty sets are straightforward for me. For example, I just model the demand in a way that it could be perturbed safely within 10% of the nominal(forecasted) value.

I wonder, how can I construct an ellipsoidal uncertainty set, how can I even present this hourly demand via ellipsoidal sets?

Say, I wanna use historical data so to make that uncertainty set. ( I have data for demands of each hour for the past periods)

If my question is not clear enough, please let me know.