An edge computing-based fast restoration for urban medium- and low-voltage distribution networks

The rapid and reliable restoration of urban medium- and low-voltage distribution networks is paramount for sustaining economic activities and social well-being. However, conventional centralised restoration methods are increasingly inadequate due to their inherent limitations in handling communication delays, heterogeneous real-time data integration, and the high computational complexity of stochastic optimisation, leading to prolonged outages and reduced service reliability. To address these challenges, this research proposes a fast power recovery method based on distributed edge computing for urban medium- and low-voltage distribution networks. The method enhances restoration efficiency through localised data processing, improved temporal performance, and the integration of heterogeneous data sources. Employing Box-Cox-combined Z-scale conversion for non-Gaussian temporal datasets and principal component-enhanced Dempster-Shafer deduction for information amalgamation, the approach transmutes multi-criteria recovery into singular-goal optimisation via decision matrices. Probabilistic voltage restrictions are reformulated as definitive quadratic mixed-integer constraints through sample average approximation, while second-degree conical relaxation manages non-linear current equations to establish a tractable mixed-integer quadratically constrained programming framework. Experimental outcomes demonstrate 2.89-minute recovery intervals, success probabilities exceeding 95%, and 0.82-0.91 load distribution equilibrium, exhibiting superior performance relative to comparative methodologies.