Scaling issues in day-Ahead formulations of stochastic unit commitment

Mitigation of uncertainty has always been a priority to power system operators. As systems around the world expand to include more intermittent resources, i.e. wind and solar power generators, complications from uncertainty could grow beyond existing practices to ensure reliability. Stochastic optimization techniques have been investigated for many years as a method of prepositioning systems of critical events; often these techniques are dismissed in part due to the challenges to obtain quality solutions within reasonable timeframes. Further, scaling issues persist, where using more scenarios within stochastic models can dramatically expand the computational time necessary to arrive at solutions, regardless of their quality. However, modeling decisions can be refined in a way that limits the potential for models to become too large to effectively solve. This paper analyzes the computational performance of different formulations of the day-Ahead unit commitment problem on the RTS-96 test case when expanded to include elements of stochastic optimization. Effective modeling frameworks could be harnessed in real-world settings without compromising the tight time requirements necessary for incorporation within a power system operational paradigm is demonstrated. A review of existing practices is also presented to provide a comprehensive review of the challenges to be faced to implement stochastic programming.