Efficient power system year-round hourly operation simulation based on multi-stage Stochastic Dual Dynamic Integer Programming

In the context of power systems increasingly reliant on renewable energy sources, the consideration of uncertainty becomes paramount for year-round hourly operational simulations aimed at assessing the efficacy of power grid development strategies. While multi-stage stochastic programming has been effective in capturing multi-scale power fluctuations, its adoption faces challenges related to computational complexity and convergence performance. To address these issues, this paper presents a novel fast multi-stage stochastic unit commitment method tailored for year-round hourly operational simulation. This method strategically incorporates the expectations of a limited number of future stages to expedite the iteration process, thereby mitigating computational burdens. The annual time-series data is adaptively segmented based on the fluctuation characteristics of power and load, ensuring a balanced sub-problem scale aligned with the number of stages. Results from rigorous testing across multiple standard cases demonstrate that the proposed method consistently achieves optimal lower bounds within 6-8 iterations, resulting in significant computational time savings of up to 50%. Furthermore, the efficacy of the proposed method is showcased through its application in the annual operational simulation of a real-world provincial high-voltage power grid in China. • Multi-scale uncertainty introduced to year-round hourly operation simulation. • Integrity of time-series maintained and number of stages reduced by AGS method. • SDDIP algorithm reformulated to satisfy the extra unequal-length subproblems. • Fast finite SDDIP method proposed to accelerate the convergence performance.