High-dimensional probability preserving scenario reduction method using adaptive selection with sampling for large-scale power system

Representative scenarios are critical in stochastic scheduling and reliability analysis of power systems. The generation of representative scenarios, particularly from extremely high-dimensional systems, faces challenges with the integration of renewable energy sources. Typically, representative scenarios are selected by optimization methods, and the complex computation process limits their application in extremely high-dimensional probability spaces, especially when the original scenario set exceeds the capacity of memory and cannot be traversed. In this paper, a novel scenario reduction scheme termed Adaptive Selection with Sampling (ASS) scheme is developed to obtain representative scenarios in extremely high-dimensional problems. The scheme is established by the selection of more representative scenarios through a compressed scenario set constructed from the sampling method, which can be applied irrespective of dimensional constraints. In order to cover the probability space, a novel sampling-based method considering low-probability scenarios is introduced. The ASS scheme incorporated with it achieves a higher level of representativeness compared to classic scenario reduction methods in the experiment. Furthermore, the reliability analysis is conducted on the IEEE test system. The calculation of expected wind curtailment and load shedding values within the ASS scheme requires only 1/100 of the samples to achieve convergence compared to existing sampling methods.