基于Transformer的V2G功率容量预测

V2G Power Capacity Forecasting based on Transformer

中文关键词: V2G功率容量预测  Transformer  动态时间规整  局部特征提取  局部注意力

英文关键词: v2g power capacity forecasting  transformer  dynamic time warping  local feature extraction  local attention

中文摘要:

      随着电动汽车行业的迅猛发展,大规模电动汽车的接入对现有电力系统带来了前所未有的挑战和机遇。本文提出了一种基于Transformer的深度学习框架,针对大规模电动汽车的Vehicle-to-Grid (V2G) 功率容量预测进行研究。通过实时监测系统收集的大量真实充放电数据,本文提出的模型能够深入学习和理解电动汽车充放电行为的内在规律,实现精准的功率容量预测。预测模型采用了动态时间规整匹配(DTW)算法结合附加条件向量的Transformer模型,这不仅提高了预测精度,并且增强了模型的可解释性。实验结果表明,本文提出的模型在平均绝对误差(MAE)和均方根误差(RMSE)两个指标上均取得了优于现有Transformer模型和LSTM模型的性能,验证了本文所提方法的有效性和优越性。

英文摘要:

      With the rapid development of the electric vehicle (EV) industry, the integration of a large number of EVs into the existing power system presents unprecedented challenges and opportunities. This paper proposes a deep learning framework based on the Transformer for the Vehicle-to-Grid (V2G) power capacity forecasting of large-scale EVs. The model developed in this paper, leverages a vast amount of real charging and discharging data collected through real-time monitoring systems to deeply learn and comprehend the intrinsic patterns of EV charging and discharging behavior, achieving accurate power capacity forecasting. The forecasting model employs the Dynamic Time Warping (DTW) algorithm combined with the Transformer model augmented with additional conditional vectors, which not only enhances the forecasting accuracy but also strengthens the interpretability of the model. Experimental results demonstrate that the proposed model outperforms existing Transformer and LSTM models in terms of both Mean Absolute Error (MAE) and Root Mean Square Error (RMSE), validating the effectiveness and superiority of the proposed method.