A data-driven intelligent optimization method for urban distribution networks

Under “Dual-Carbon” strategy in China, the high penetration of renewable energy introduces risks to distribution networks. The explosive growth of distributed photovoltaics, EV charging facilities, and energy storage systems has led to big-datascale, multi-source heterogeneous service data, causing transmission omissions, data errors, and heightened uncertainty that severely impact scheduling economics. This paper proposes a data-driven optimization algorithm for urban distribution networks: First, a scheduling model is formulated and transformed into a linear deterministic model via duality theory and Lagrangian Relaxation. Then, an innovative Data Trimming Algorithm (DTA) is designed to iteratively generate representative subsamples by removing outliers and extreme values from the raw dataset, mitigating their adverse effects on distributionally scheduling outcomes. Case studies demonstrate that the algorithm significantly improves dispatch economics while ensuring operational reliability.