A robust approach to Meta-EMS

This study was concerned with the meta-Energy Mediation System (meta-EMS) problem. We consider several companies from the horticulture industry, for which we have to match supply and demand of different commodities a few days ahead. We have introduced the problem and considered the relevant literature for the current subject. The problem is formulated as an MIP-model, and we have shown it to be NP-complete. The performance measure was minimal costs, and we have shown our cooperative model to improve on the current, non-cooperative situation. The computation time is exponential with the size of the instance. We have solved given test cases to optimality and improved on the heuristic that was made for these cases. Next, we have constructed our own, more complicated test cases. The model runs fast enough for models of smaller size, but when the size increases some heuristics might be required. We have proposed two heuristics, one based on aborting the solver and obtaining a very good bound, the other based on an LP-relaxation. A sensitivity analysis was performed and it appeared that that the model is insensitive with respect to changes in demand. However, the model is not robust with respect to changes in demand, which is shown by a Monte-Carlo analysis. We have given a small introduction to robustness and we have made the model robust against changes in demand by implementing the Affinely Adjustable Robust Counterpart (AARC). This improves on the worst case RC solution, but the size of the improvement is dependent on which portion of data is used to adjust the variables to the uncertain demand. Our research indicates that the most effective improvement is given when taking the current time step as the portion of data. This approach is suitable for small models. When the number of companies or time steps increases, the computation time becomes too large.