Resumen
Unit commitment (UC) models used in real-life island power systems (IPSs) require complex modelling details and low execution times since they have to be launched repeated times for different assumptions and changing conditions. For this reason, system operators and power generation companies discard for daily analysis sophisticated models such as stochastic, robust or frequency-constrained UC. Instead, they prioritize the use of deterministic standard UC (STUC), where units are scheduled imposing a static reserve criterion such that the amount of available spinning reserve is sufficient to cover expected disturbances. This paper outlines why in IPSs the spinning reserve requirement increments dramatically the computational burden of a STUC, proposing practical solutions to contain runtimes. Proposed solutions include incrementing convergence tolerance, dividing a weekly optimization in seven sequential daily problems and formulating alternative equations for the reserve constraints. Results of real-life small and big size Spanish IPS will demonstrate how execution times are drastically diminished with acceptable loss of accuracy.
