基於時空過程模型預測電動汽車充電需求

Abstract

作為減緩氣候變遷的手段之一為讓電動汽車取代傳統燃油車，然而鑑於電動汽車公共充電站的覆蓋率不足導致消費者購買電動汽車的意願降低，因此唯有先了解電動汽車的充電需求才能在資源有限的情況下妥善地規劃充電站。本論文利用時空點過程的過程強度（intensity function）描述電動汽車充電的潛在需求，建構適合的時空點過程模型以及時空高斯過程模型，並找出影響電動汽車充電需求的因素。在案例研究中，我們配適了兩種解釋模型：時空點過程中的對數高斯 Cox 點過程（log Gaussian Cox process）以及時空高斯過程模型（spatio-temporal Gaussian process）來描述台北市電動汽車充電事件點的分布情形，並根據我們的研究結果顯示最適切於台北市電動汽車充電事件點模式之模型以時空高斯過程模型較佳，未來可運用該模型來預測台北市電動汽車充電需求的潛在時空分佈。

One of obstacles to replacing fossil fuel cars with electric vehicles is the lack of charging facilities. In order to set up charging facilities appropriately with limited resources, the charging demand needs to be analyzed and understood. This thesis uses the intensity function of the spatio-temporal point process to describe the potential electric vehicle charging demand. We construct a suitable spatio-temporal model and find out the factors that would affect the electric vehicle charging demand. In this study, we examine two explanatory models: log Gaussian Cox point process and spatio-temporal Gaussian process. In our case study, we find that the most suitable model for the electric vehicle charging event point model in Taipei is the spatio-temporal Gaussian process model. This model could predict the potential spatial and temporal distribution of the electric vehicle charging demand in Taipei in the future.