Search

Conventional model-driven methods are hard to handle large-scale power flow with multivariate uncertainty, variable topology, and massive real-time repetitive calculations. With the ability to deal with non-Euclidean graph-structured power system data, graph deep learning shows great potential in modern power flow calculation. However, general graph deep learning based power flow calculation has limited adaptability because of its sole mapping of node information and black-box attributes. In this paper, an edge graph attention network based power flow calculation (EGAT-PFC) model is proposed with improved adaptability for power flow analysis of complex system scenarios. First, the dual-model structure of the node model and edge model is constructed to realize a complete power flow mapping covering all information in power systems. Second, an improved learnable attention coefficient mechanism fusing node and edge features is proposed to ensure global information can be completely considered. Third, mechanisms of extended first-order neighborhood, dynamic normalization, and regularization-based loss function are designed to improve training performance. Finally, visualized interpretability is developed to show valuable information of vulnerable nodes and lines of power system operation. The numerical simulation verifies that EGAT-PFC has high accuracy, fast mapping, as well as excellent adaptability to variable topologies.

With the construction of new-type power system under the ”double carbon” target and the increasing diversification of the energy demand of the user side, the short-term load forecasting of power system is facing new challenges. In order to fully exploit the massive information contained in big data, this paper proposed a new short-term load forecasting method for new-type power system considering multiple factors, which based on Graph Convolutional Network (GCN) and Long Short-Term Memory network (LSTM). Spearman rank correlation coefficient was used to analyze the correlation between load and meteorological factors, and a quantitative model including meteorological factors, date factors and regional factors was established. Thus, GCN and LSTM were jointly used to extract the spatial and temporal characteristics of massive data respectively, and finally the short-term power load forecasting was achieved. The public data sets were used for performance verification compared with three comparison models, LSTM, CNN-LSTM and TCN-LSTM. The results show that the proposed method can make full use of the influence of multi-dimensional data, meanwhile improve the load prediction accuracy and training efficiency effectively.