Application of multidomain data fusion in equipment-level load forecasting of distribution networks

With the rapid development of smart grid technology, equipment-level load forecasting in distribution networks has become increasingly important for power system scheduling and management. Traditional load forecasting methods mainly rely on single data sources, often failing to fully consider the interactions of various factors, leading to limited forecasting accuracy. This study proposes a multi-domain data fusion model based on Graph Neural Networks (GNNs), aiming to improve the accuracy and robustness of load forecasting by integrating load data, weather data, and equipment status data. By fusing data from different domains within a unified framework, the model can effectively capture the complex relationships between equipment status, weather changes, and load fluctuations. Experimental results show that the proposed GNN multi-domain data fusion model outperforms traditional models across several evaluation metrics, with particularly notable advantages in prediction accuracy and stability. This method provides more accurate load forecasting support for the scheduling and operation of power systems, better addressing the complex dynamics in distribution networks.