A New Data-Driven Quasi-Monte Carlo for Probabilistic Optimal Power Flow

Abstract

Probabilistic optimal power flow (POPF) assists system operators in risk-based decision-making. Characteristics like non-intrusive nature and the ability to employ the whole deterministic model make sample-based techniques like Quasi-Monte Carlo (QMC) suitable for solving POPF. However, the downsides of QMC’s broad application include its uneven accuracy, fluctuating rate of convergence, and absence of a quality metric. To that aim, we present a data-driven nonparametric QMC framework for solving POPF with complex and correlated uncertainties accurately and efficiently. The proposed methodology employs the uniform experimental design (UD) as a QMC sampling approach. The suggested framework, based on the copula perspective, directly calculates the appropriate correlation matrix in Gaussian space, decreasing the computing cost. Furthermore, we suggest mixture discrepancy (MD) as a metric that can assist researchers in choosing the appropriate QMC sample set for POPF without the need for time-consuming simulation. Results from the case study on a modified 39-bus system reveal that the proposed UD-based QMC reduces computing effort while providing accurate POPF results compared to current QMC approaches.