Hybrid grey model and machine learning approach for outage forecasting in medium voltage power distribution networks

In modern power systems, the increasing complexity of medium voltage (MV) distribution networks and rising environmental risks necessitate accurate and timely outage forecasting. This paper proposes a hybrid data driven framework that combines grey prediction modelling and support vector machine (SVM) classification for multi scale outage prediction. Annual and monthly outage counts are estimated using an optimized Grey Model GM (1,1) enhanced by Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). For day-ahead prediction, SVM is employed to classify outage risks using historical weather and fault data. The proposed method is validated using real operational data from the Italian distribution grid between 2008 and 2017. Experimental results demonstrate that the hybrid approach significantly improves prediction accuracy and supports proactive maintenance and resource allocation in MV networks.