Hybrid Offline/Online Methods for Optimization Under Uncertainty

This work considers multi-stage optimization problems under uncertainty. In this context, at each stage some uncertainty is revealed and some decision must be made: the need to account for multiple future developments makes stochastic optimization incredibly challenging. Due to such a complexity, the most popular approaches depend on the temporal granularity of the decisions to be made. These approaches are, in general, sampling-based methods and heuristics. Long-term strategic decisions (which are often very impactful) are typically solved via expensive, but more accurate, sampling-based approaches. Short-term operational decisions often need to be made over multiple steps, within a short time frame: they are commonly addressed via polynomial-time heuristics, while more advanced sampling-based methods are applicable only if their computational cost is carefully managed. We will refer to the first class of problems (and solution approaches) as offline and to the second as online. These phases are typically solved in isolation, despite being strongly interconnected. Starting from the idea of providing multiple options to balance the solution quality/time trade-off in optimization problem featuring offline and online phases, we propose different methods that have broad applicability. These methods have been firstly motivated by applications in real-word energy problems that involve distinct offline and online phases: for example, in Distributed Energy Management Systems we may need to define (offline) a daily production schedule for an industrial plant, and then manage (online) its power supply on a hour by hour basis. Then we show that our methods can be applied to a variety of practical application scenarios in very different domains with both discrete and numeric decision variables.