Multi-agent for analyzing power outage causes in distribution networks of new power system

This paper systematically reviews the research evolution and current status in the field of distribution network outage cause analysis. The review indicates that, alongside the high proportion of renewable energy integration and the digital transformation of distribution networks, current research faces challenges including prominent misclassification of “pseudo-outages”, fragmented analytical methodologies, and overreliance on single-model approaches such as random forests. While existing work has made progress in user profiling, multi-source data fusion, and deep learning applications, it generally suffers from issues such as ambiguous boundary conditions, methodological singularity, and insufficient system flexibility. To this end, this paper innovatively proposes a future direction for transitioning towards a “multi-agent collaborative” analytical paradigm: defining clear event discrimination boundaries, constructing a hierarchical architecture comprising specialised feature extraction agents, multi-disciplinary expert judgement agents, and collaborative decision-making agents, and extending its application to decision optimisation for enhancing distribution network reliability. This framework aims to integrate the strengths of physical knowledge, data-driven approaches, and agent-based tools, thereby enhancing system interpretability, analytical speed, and assessment accuracy in complex scenarios. It provides an innovative theoretical pathway and practical reference for intelligent operation and maintenance of distribution networks within the context of new power systems.