Frontiers in artificial intelligence · 2026
DynEC: dynamic evolutionary clustering for power user load profiling using multi-view graph neural networks.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Zhao Lei, Zhao Hong, Li Mengjie, Wang Jia, Ke Xingsi, Yao Yumin
Аннотация
INTRODUCTION: With the deep integration of generation-transmission-load-storage systems, the power demand side has become highly dynamic and stochastic, challenging the traditional assumption that user behavior remains stationary over time. Static clustering models therefore suffer from sensitivity to daily noise and false user identity switching.
METHODS: This study proposes Dynamic Evolutionary Clustering (DynEC), a multi-view graph neural network framework for power user load profiling. DynEC constructs a sparse multi-view dynamic graph that captures geometric proximity, temporal alignment through constrained dynamic time warping, and statistical dependencies. A gated spatiotemporal graph neural network then optimizes a dual-objective loss to learn latent representations while balancing current snapshot quality and historical temporal smoothness.
RESULTS: Experiments on real-world datasets show that DynEC outperforms existing baseline methods. The proposed framework identifies genuine concept drift more accurately while reducing erroneous cluster switching.
DISCUSSION: DynEC provides a stable and reliable profiling tool for modern power grid management by modeling load profiling as a continuous evolutionary process rather than a set of independent static clustering tasks.
METHODS: This study proposes Dynamic Evolutionary Clustering (DynEC), a multi-view graph neural network framework for power user load profiling. DynEC constructs a sparse multi-view dynamic graph that captures geometric proximity, temporal alignment through constrained dynamic time warping, and statistical dependencies. A gated spatiotemporal graph neural network then optimizes a dual-objective loss to learn latent representations while balancing current snapshot quality and historical temporal smoothness.
RESULTS: Experiments on real-world datasets show that DynEC outperforms existing baseline methods. The proposed framework identifies genuine concept drift more accurately while reducing erroneous cluster switching.
DISCUSSION: DynEC provides a stable and reliable profiling tool for modern power grid management by modeling load profiling as a continuous evolutionary process rather than a set of independent static clustering tasks.