Data-Driven Probabilistic Optimal Power Flow With Nonparametric Bayesian Modeling and Inference

In this paper, we propose a data-driven algorithm for probabilistic optimal power flow (POPF). In particular, we develop a nonparametric Bayesian framework based on the Dirichlet process mixture model (DPMM) and variational Bayesian inference (VBI) to establish a probabilistic model for capturing the uncertainties involved with wind generation and load power in power systems. In the proposed setup, the number of components in the mixture model can be automatically and analytically obtained from the consistently updated data. Moreover, we develop an efficient quasi-Monte Carlo sampling method to draw samples from the obtained DPMM, then propose the dynamic data-driven POPF algorithm. Performance of uncertainty modeling framework on publicly available datasets is examined by extensive numerical simulations. Furthermore, the proposed POPF algorithm is verified on multiple IEEE benchmark power systems. Numerical results show the feasibility and superiority of the proposed DPMM-based POPF algorithm for better informed decision-making in power systems with high level of uncertainties.