Operational risk assessment and energy storage optimization for low-voltage distribution networks based on DeepAR-KAN

Introduction The operational environment of low-voltage distribution networks is becoming increasingly complex due to the high penetration of distributed energy resources and the growing diversity of end-user loads. Consequently, operational risks such as voltage deviations, three-phase unbalances, and active power overloads are becoming more prominent, posing serious challenges to power supply reliability and power quality. Methods To address these issues, this paper develops an integrated analytical and decision-making framework that unifies ultra-short-term probabilistic load forecasting, dynamic risk assessment, and energy storage optimization operations. First, an improved DeepAR-based probabilistic forecasting model is proposed, where the Kolmogorov–Arnold Network (KAN) is embedded to enhance the model’s capacity for capturing complex nonlinear and stochastic features, improving short-term load forecasting accuracy. Second, a dynamic multi-risk assessment framework is constructed, which simultaneously evaluates voltage deviations, phase unbalances, and active power overload risks, and introduces time-decay weighting factors to achieve multi-time-step risk aggregation. Finally, an energy storage optimization model was formulated with the objective of minimizing the comprehensive operational risk, enabling proactive and risk-aware control of distribution network states. Results Case studies conducted on the IEEE 13-bus distribution system and practical data from a low-voltage distribution network demonstrated that the proposed framework enhances the situational awareness, dynamic adaptability, and proactive control capability of distribution systems against multi-dimensional operational risks. Discussion The results provide both theoretical insights and engineering references for improving the secure and resilient operation of future low-voltage distribution networks.