Impact of Wind Power Scenario Reduction Techniques on Stochastic Unit Commitment

Abstract

Stochastic unit commitment (SUC) is an effective method widely used to cope with the uncertainty of wind power. For the limitation of computation capability, only limited members of representative scenario can be considered in SUC. It thus rises the concern that whether the selected scenarios can fully represent the uncertainty nature of wind power. In this paper, the performance of reduced scenarios is quantified by both its statistical quality and its economic value on the optimality of SUC. Two metrics are proposed to quantify the distortion of the stochastic quality of wind power during the scenario reduction process: output uncertainty and ramp diversity. The economic value of reduced scenarios is evaluated as the difference between the optimal cost of the SUC model associated with limited scenarios and the expected "actual" operating costs when considering all the possible scenarios. Then, this paper reviews several typical wind power scenario techniques and categorizes them by both the scenario clustering approach and scenario reduction criterion. The quality of each method is tested using the real wind power data from NREL database and the modified IEEE RTS-79 system. Results show that the performance of SUC is more sensitive to the output uncertainty approximation rather than the ramp diversity approximation of reduced scenarios.