Probabilistic Assessment of Power System Flexibility Based on Bayesian Neural Networks

As the share of renewable energy sources grows, it becomes essential to develop effective management strategies to handle the associated variability and uncertainty. This paper presents a novel approach to rolling optimization modeling aimed at enhancing data collection and developing flexibility indicators for power systems. These data and indicators are utilized as inputs and outputs for Bayesian Neural Networks (BNNs), enabling regression-based predictions of power system flexibility. By leveraging historical operational data and advanced predictive modeling techniques, the study facilitates rapid assessments crucial for effective management of renewable energy integration and overall system flexibility. The findings reveal significant fluctuations in flexibility driven by the variability of renewable energy sources, underscoring the model's high predictive accuracy as demonstrated through a case study of the IEEE-33 node system. Ultimately, by integrating rolling optimization with data-driven predictive models, this research contributes to the ongoing discourse on achieving a sustainable energy transition while ensuring the reliable and efficient operation of power systems.