Stochastic peak shaving scenario generation for grid-friendly building energy system design

Leveraging the energy flexibility of demand-side building energy systems (BES) provides a viable solution to mitigate grid imbalance. The optimal design of BES's flexible resources is essential for enhancing its grid-friendly interaction capabilities. Previous BES designs have focused on regular time-of-use (TOU) pricing to improve grid-friendliness, without considering the grid's temporary peak-shaving needs, which are expected to increase in the future. This study proposes a grid-friendly BES planning method that integrates both temporary peak-shaving DR scenarios and long-term TOU pricing into design optimization. To address the challenge of limited historical DR data, a modified TimeGAN is introduced to generate stochastic peak-shaving scenarios that explicitly reflect the grid's regulation needs, including specific regulation time periods, required capacities, and offered incentives. A stochastic multi-scenario optimization model is proposed to integrate these scenarios and synergistically optimize BES's grid-friendly performance across both continuous and event-driven interactions, thereby determining the optimal design. Results show that, compared to the traditional method, the proposed method significantly enhances the grid-friendliness of BES without incurring extra system costs, offering on average 25 % higher peak-shaving capacities and 30 % lower net loads during DR periods. Furthermore, with an increased annual frequency of peak-shaving, the proposed method shows superior economic performance.