An Efficient Data-Driven Conditional Joint Wind Power Scenario Generation for Day-Ahead Power System Operations Planning

Advanced stochastic programming-based operations planning (OP) requires wind power forecasts in the form of scenarios. The quality of the decisions made under uncertainty is sensitive to the quality of the infeed scenarios. Consequently, there is a need for accurate and efficient methods of scenario generation (SG). This article proposes a conditional joint wind power SG method for multiple wind farms. The proposed framework models the dependence among wind farms using vine copula and presents a novel analytical conditional sampling algorithm (CSA). Unlike the existing work, this CSA is inherently accurate, avoids the tedious manual conditional sample selection, and enables control over the number of conditional scenarios generated. Also, a uniform design-based variance reduction is introduced and integrated into the proposed CSA, which benefits the downstream OP applications with improved convergence and accuracy of the solutions. A detailed two-stage scenario evaluation procedure is carried out on a real-world dataset: univariate and multivariate quality metrics-based statistical evaluation in the first stage and an application-based evaluation in the second stage. Results suggest that the proposed SG method significantly improves the overall quality of the forecasted wind power scenarios and provides a better cost-risk balance in the application compared to the benchmarks.