Intelligent Early Warning and Decision Support for Power Grids Based on the Hybrid Model of Isolation Forest and LightGBM

Le résumé fourni par la source

The monitoring data of modern power grids are characterized by being multi-dimensional, massive, and requiring high timeliness. Early warning methods based on thresholds or simple statistical models struggle to handle faint, early-stage anomalies under complex operating conditions, often leading to delayed or false alarms. This paper proposes an intelligent early warning and decision support model for power grids based on multi-dimensional data mining. The model utilizes the Isolation Forest (IF) algorithm for unsupervised anomaly detection on multi-dimensional operational monitoring indicators to capture early signs of deviation from normal operating states. Subsequently, it introduces the Light Gradient Boosting Machine (LightGBM) to classify the identified anomalies, diagnose potential alarm types, and provide decision support for operational personnel. Experimental results demonstrate that the proposed model shows significant advantages in key evaluation metrics, achieving a balance between precision and recall, and exhibiting excellent interpretability and robustness.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.