Subjects: Computer Science >> Integration Theory of Computer Science submitted time 2022-04-07 Cooperative journals: 《计算机应用研究》
Abstract： Traffic flow prediction plays a key role in the construction of intelligent transportation systems. However, existing prediction methods cannot mine the potential spatiotemporal correlation in the data accurately, and most of them use fully connected networks for single-step prediction. In order to further mine the spatial-temporal features of the data and improve the accuracy of long-term and short-term prediction tasks, this paper proposed a Gated Recurrent Graph Convolutional Network (GR-GCN) . Firstly, by using spectral graph convolution combined with gated recurrent unit (GRU) to construct spatial-temporal components (STC) to capture the spatial-temporal correlation of nodes in the network, fully extract the spatial-temporal features of the data. Secondly, use STC as an encoder unit and enter the time sequence data together with road network data into it. Finally, using gated recurrent unit as a decoder unit, combine the two into an encoder-decoder network structure (Encoder-Decoder) in chronological order, and decode the prediction results at each moment in turn. Experiments were carried out on the highway datasets PeMSD4 and PeMSD8 in the California Department of Transportation (Caltrans) performance evaluation system. The results show that model GR-GCN is better than most existing benchmark models in predicting the traffic flow in the future 15 minutes, 30 minutes, 45 minutes and 60 minutes, especially in long-term prediction.
Subjects: Computer Science >> Other Disciplines of Computer Science Subjects: Energy Science >> Energy Science （General） submitted time 2021-08-11
Abstract：负荷预测是电网系统中很多应用的关键部分，具有重要作用。然而，由于电网负荷的非线性、时变性和不确定性，使得准确预测负荷具有一定的挑战。充分挖掘负荷序列的潜在特征是提升预测准确率的关键。本文认为在特征提取时应该充分利用负荷序列的位置信息、趋势性、周期性和时间信息，同时还应构建更深层次的神经网络框架进行特征挖掘。因此，本文提出了基于特征嵌入和Transformer框架的负荷预测模型，该模型由特征嵌入层，Transformer层和预测层组成。在特征嵌入层，模型首先对历史负荷的位置信息、趋势性、周期性和时间信息进行特征嵌入，然后再与天气信息进行融合，得到特征向量。Transformer层则接受历史序列的特征向量并挖掘序列的非线性时序依赖关系。预测层通过全连接网络实现负荷预测。从实验结果来看，本文模型的预测性能优于对比模型，体现了该模型的可行性和有效性。
Peer Review Status:
Awaiting Review
Hits
Hits
Downloads
Comment