Analyzing the Computational Burden of Global-Linking Balance Equations in the Medium-Term Unit Commitment Problem

Nowadays, the challenging developments in modern power systems demand increasing the accuracy of energy models for supporting decision-making processes towards a clean energy transition. However, enhancing their details entails substantially enlarging the corresponding computational burden. Accordingly, simplifications are mandatory when real-size electricity markets are represented. For that reason, medium-term models frequently use aggregation techniques, integer programming relaxation, or not considering uncertainty. Nevertheless, market players require powerful tools to manage their generation assets skillfully. Hence, thermal unit itemization, hourly time spans, integer behavior, and uncertainty representation are desirable features. Consequently, some soft-linking methodologies that overcome the simplifications made in energy models by applying a post-processing phase have been recently proposed in the literature. This paper analyzes the computational performance of one of these approaches and boosts its scope by considering additional technical aspects for achieving more realistic, feasible thermal schedules. The methodology takes reliable results of a multi-area medium-term fundamental model, such as aggregated productions of the thermal units, hourly price forecasts, and the situation in the interconnection facilities. Later, these data are utilized as inputs in a unit commitment formulation with global-linking balance equations, which optimally distribute the generation according to profitability and strategic aspects.