An adaptive dynamic scenario clustering method for power system optimization control under spatiotemporal correlated uncertainties

Abstract In the stochastic optimization control of power systems, renewable energy has strong uncertainty and underlying correlations at spatial and temporal scales, which need to produce a large number of discrete scenarios to characterize. However, it will bring a huge solving burden to the stochastic optimization, thus requiring necessary scenario reduction operation. Moreover, the inability to accurately measure and mitigate correlation distortion poses a major challenge for uncertainty quantification, potentially compromising the fidelity of stochastic optimization. In this paper, a novel adaptive dynamic scenario clustering algorithm (ADSCA) is proposed for the problem that the spatiotemporal correlations cannot be well preserved. The concept of spatiotemporal correlation loss is firstly presented to measure the deviation degree of linear and tail dependence after reduction. Based on the constructed clustering model aiming to minimize spatiotemporal correlation loss, the optimal number of preserved scenarios does not need to prespecified, and can be dynamically adjusted through iterative splitting and merging operations, ultimately obtaining the most representative scenarios while improving reduction efficiency. Numerical results show that ADSCA has outstanding advantages in preserving the statistical and correlation features, while ensuring the representativeness of scenarios. Application studies to probabilistic optimal power flow are also conducted to verify the effectiveness of ADSCA in improving the power system reliability and economic benefits.