Probabilistic evaluation of a power system's capability to accommodate uncertain wind power generation

With the rapidly growing integration of wind power generation (WPG), it is of great importance for an operator to grasp the ability of the power system to accommodate uncertain WPG. This study proposes two probabilistic methods to assess such capability of a power system based on the level of data availability. If the probability distribution type (PDT) of wind power prediction error (WPPE) is known, the total accommodation probability is calculated as the sum of a fully guaranteed probability and a partially guaranteed probability. The former one leads to a tri-level max-max-min optimisation problem which is solved via a dichotomy procedure, and the latter one can be obtained based on the geometrical analysis of the dispatchable region of WPG. If the PDT of WPPE is not exactly known, the authors tackle the problem via a data-driven uncertainty quantification method. More precisely, they consider a family of ambiguous probability distributions around the empirical distribution described by historical data in the sense of Wasserstein metric. The probability of failure in the worst-case distribution is calculated from a linear programming. The proposed method is tested on modified PJM-5 and IEEE-118 bus systems. Comparison with the traditional Monte Carlo Simulation method demonstrates its efficacy and efficiency.