基于时空卷积动态知识图谱的新能源消纳评估方法

Renewable Energy Accommodation Assessment Method Based on Spatio-temporal Convolution Dynamic Knowledge Graph

为构建新能源消纳知识图谱,首先,将电网积累的海量调度运行数据以动态四元组的形式,显式地表达调度运行数据的时空关联关系。通过滑动时间窗口快速搜索、提取局部时空图,构建子图数据集。然后,时空同步图卷积网络对局部时空图进行高维特征提取,充分挖掘历史数据的时空关联关系,利用新能源消纳知识图谱中存储的机理知识对模型进行引导,并通过多子图并行训练提升模型的学习效率。最后,基于中国西北某省级电网算例进行仿真和实验验证。结果表明,所提方法可以有效避免复杂的数学建模以及模型求解,相比于传统方法具有更高的评估精度与速度。

To construct a knowledge graph of renewable energy accommodation,firstly,the accumulated massive dispatching operation data of the power grid in the form of dynamic quaternions explicitly expresses the spatio-temporal correlations of dispatching operation data.The local spatio-temporal graph is quickly searched and extracted by sliding time windows to construct sub-graph data sets.Then,the spatio-temporal synchronous graph convolutional network extracts high-dimensional features from the local spatio-temporal graphs to fully excavate the spatio-temporal correlations of the historical data.The model is guided by the mechanism knowledge stored in the knowledge graph of renewable energy accommodation and trained in parallel by multiple subgraphs to improve the learning efficiency.Finally,simulation and experimental validation are conducted based on a provincial grid case in Northwest China.The results show that the proposed method can effectively avoid complicated mathematical modeling and solving,and has higher evaluation accuracy and speed compared with traditional methods.