oemof/oemof-solph

Minimum uptime or downtime of the NonConvex class are great options to avoid on-off cycling of e.g. Converters. The current implementation enforces these constraints within the assessment period, which makes sense for one-off simulations. However, this can lead to a "bypass" of the constraints, if the simulation period is supposed to be a representative one, which "repeats" itself (e.g. when calculating LCOH from annuities and repeating constant operating cost from the rep. simulation period). This can be marginal in case of a year, but if you're simulating, e.g., days or 72h periods (see also TSA), the effect can be quite consequential. In any case, allowing for "wrapping/cyclic" minimum uptime or downtime should only be optional. Maybe, a not too laborious approach could be working with modular division. See this example for minimum uptime (mu), for each timestep (t) in the assessment period (T): $$(status[t] - status[(t-1) \mod T]) \cdot mu \leq \sum_{n=0}^{mu-1} status[(t+n) \mod