基于ATT-TCGNN的综合能源系统多元负荷短期预测

Multi-energy load short-term forecasting of integrated energy system based on ATT-TCGNN

综合能源系统多元负荷之间存在较强的复杂耦合关系,且多元负荷数据具有较强的波动性与随机性。针对上述特点,提出一种基于图神经网络、注意力机制、变分模态分解的多元负荷短期预测模型。首先,对多元负荷数据进行变分模态分解,削弱其波动性与随机性;然后,通过经注意力机制改进的图学习网络建立充分反映多元负荷耦合联系性、负荷与气象间关联性的图结构,并用图预测网络对图结构与多元负荷历史数据进行分析,实现多元负荷预测;最终,结合亚利桑那州立大学的实际数据对所提出模型与其他模型进行对比分析,结果表明,所提出模型具有更高的预测精度。

In an integrated energy system,there are complex and strong coupling relationships between the multi-energy loads,and multi-energy loads have strong volatility and randomness.In view of the above characteristics,a multi-energy load short-term forecasting model based on graph neural network,attention mechanism and variational mode decomposition is proposed.Firstly,the variational mode decomposition of multi-energy loads is carried out to weaken the volatility and randomness.Then through the graph learning network improved by the attention mechanism,a graph structure that fully reflects the coupling connection of multi-energy loads and the correlation between multi-energy loads and meteorology is established,and the graph prediction network is used to analyze the graph structure and the historical data of multi-energy loads to realize the prediction of multi-energy loads.Finally,the proposed model is compared with other models based on the actual data of Arizona State University.The results show that the proposed model has higher prediction accuracy.