以網路爬蟲技術探勘影響社區接受充電樁之關鍵要素

Abstract

為響應淨零碳排目標，電動車普及，充電樁之設置日益增加。然而，充電樁具用電安全與火災等潛在風險，且於社區安裝供電設備須符合相關規範，問題複雜，而社區居民往往缺乏對充電樁可能造成的風險認知及自我保護意識薄弱。唯探討影響社區接受充電樁之相關研究為極為有限，但此議題非常重要。為填補此研究缺口，本論文導入網路資料探勘、結構方程模型與多準則決策分析架構，導入保護行動決策模型（Protection Action Decision Model，PADM）與捷思式-系統性資訊處理模型（Heuristic-systematic Information Processing Model，HSM），推衍影響社區接受充電樁之關鍵要素。本研究首先探勘與社區充電樁風險相關之百度網頁，並以基於隱含狄利克雷分布（Latent Dirichlet Allocation，LDA）之主題分析模型（Topic Modeling），探勘網頁中蘊含之主題。其次，以階層式集群分析法（Hierarchical Cluster Analysis）將主題分群後，將各群之主題導入保護行動決策模型與捷思式-系統性資訊處理模型之構面，並以偏最小平方結構方程模型（Partial Least Squares Structural Equation Modeling，PLS-SEM）驗證其路徑之顯著與否。最後，本研究收集專家意見，以基於決策實驗室法之網路流程（DEMATEL based Analytic Network Process，DANP）推導初始影響矩陣，並得出每一主題之權重後，比較網路資料探勘與專家意見之差異，本研究以網站爬取之資料實證研究本分析架構之可行性，以PLS-SEM研究之結果顯示，「風險認知」與社區接受充電樁的關聯性最高，而依據DANP彙整專家意見推衍之結果，除「風險認知」之外，「行為意圖」亦為影響社區接受充電樁之關鍵要素。比較兩者結果顯示，「風險認知」及「行為意圖」構面中充電樁是否符合國家標準、公共充電樁存在風險、及充電樁設施之安全性，對於社區接受充電樁，有較高的影響程度。本研究結果除可作為瞭解社區安裝充電樁之風險外，也可提供未來充電樁安裝風險評估之參考。此外，經完整驗證之分析架構，亦可供企業及政府擬定因應充電樁風險策略之依據。

The growing number of electric vehicles (EVs) and the rapid deployment of charging infrastructure have gained prominence in response to the goal of reaching net-zero carbon emissions. However, installing charging stations raises concerns regarding electrical security and potential fire hazards. The issue is complicated further by the requirement to follow the relevant regulations while constructing electric vehicle supply equipment (EVSE) within communities.Unfortunately, community disaster readiness is lacking in infrastructure risk knowledge and self-protection. Despite its crucial necessity, research on the community acceptability of charging infrastructure is scarce. To address this research gap, this study presents comprehensive methods that combine techniques including web scraping, structural equation modeling, and a multi-criteria decision analysis framework. To determine community acceptance of charging infrastructure, the study uses the Protection Action Decision Model (PADM) and the Heuristic-Systematic Information Processing Model (HSM).This study begins by collecting charging station risks within communities using web scraping. Furthermore, Latent Dirichlet Allocation (LDA) is utilized to extract underlying themes from Baidu web-scraped data. Subsequently, the Hierarchical Cluster Analysis technique is employed to cluster the extracted topics. These clustered themes are then incorporated into the dimensions of the PADM and the HSM. Then, using Partial Least Squares Structural Equation Modeling (PLS-SEM), the path coefficients' significance is validated. Lastly, this study gathers expert opinions and employs the Decision-Making Trial and Evaluation Laboratory (DEMATEL) method. The DEMATEL-based Analytic Network Process (DANP) is employed to determine the weights for each topic and derive the initial relation matrix. After obtaining the weights for each topic, a comparison is made between the differences observed in web-scraped data and expert opinions. Based on the results obtained from web scraping and SEM, this study empirically validates the feasibility of the proposed analytical framework through data obtained from website scraping. The research findings reveal that “Risk Perception” demonstrates the highest correlation. Furthermore, after consolidating expert opinions through DANP, in addition to “Risk Perception”, “Behavioral Intentions” were also found to influence the acceptance of EV charging station installations.Examining the two sets of results indicates that factors such as charging infrastructure conforming to national standards, the risk of public charging infrastructure, and the safety of charging facility facilities have a greater influence on risk perception and behavioral intention. The findings of this study can serve not only to comprehend the risks associated with charging station installation within communities but also as a reference for future risk assessments of charging infrastructure installation. In addition, the thoroughly validated analytic framework established in this study may serve as a basis for businesses and government entities to formulate strategies.