基于SGNN-TCN融合模型的电动自行车换电需求预测研究

Research on Electric Bicycle Battery Swapping Demand Prediction Based on the SGNN-TCN Fusion Model

摘要: 外卖配送行业的快速增长使得电动自行车的使用量显著增加，电动自行车在使用过程中需要频繁补充电能，而传统的充电模式面临着充电时间长、安全隐患大等问题。电动自行车换电柜作为一种高效、便捷的电池更换方式，逐渐成为解决充电难题的关键技术之一。而准确预测电动自行车的换电需求量，不仅能为换电柜的选址提供可靠的数据支持，也可以大大提高服务的效率和服务品质。因此，针对换电需求量的预测，本文提出了一种基于时空图神经网络(SGNN-TCN)的改进模型，利用时间依赖性、空间依赖性和天气依赖性进行建模，采用GNN对空间特征进行提取，并将提取后的空间特征输入TCN层进一步挖掘时间变化规律，该模型能够有效提取电动自行车使用数据中的空间和时间特征，并准确预测未来的换电需求量。

Abstract: The rapid growth of the food delivery industry has significantly increased the usage of electric bicycles. During operation, electric bicycles require frequent energy replenishment, while traditional charging methods face problems such as long charging times and high safety risks. As an efficient and convenient battery replacement method, electric bicycle battery swapping stations have gradually become one of the key technologies for addressing charging difficulties. Accurate prediction of electric bicycle battery swapping demand can not only provide reliable data support for the site selection of battery swapping stations but also greatly improve service efficiency and service quality. Therefore, this study proposes an improved prediction model based on a spatio-temporal graph neural network (SGNN-TCN) for electric bicycle battery swapping demand forecasting. The proposed model considers temporal dependency, spatial dependency, and weather dependency in the modeling process. The graph neural network (GNN) is employed to extract spatial features, and the extracted spatial features are then input into the temporal convolutional network (TCN) to further capture temporal variation patterns. The proposed model can effectively extract spatial and temporal features from electric bicycle usage data and accurately predict future battery swapping demand.