113

10,800 千元

淨零排放科技-資源循環減碳技術計畫

2024/01/24

2024/11/30

王偉仲

循環署再利用推動組

鄧丕信

環資國際有限公司

廢棄物管理

資料科學、申報資料分析、申報流向分析

Data science, Application data analysis, Declaration flow analysis

王偉仲

02-66308000

web1798@eri.com.tw

本計畫致力於建構智慧化事業廢棄物流向分析與勾稽系統，強化環境監管，提升資源循環效率。為達成此目標，本計畫著重於資料庫擴充與精進、人工智慧技術應用、智慧化勾稽機制強化、緊急應變工具建置與資料公開、資訊安全與系統維運，以及持續推廣與精進等六大面向。 在資料方面，積極整合多元資料，強化資料廣度與深度。除了納入超過 290 萬筆稽查裁處資料、5,000 件環保相關司法判決書以及經濟部商工登記資料外，亦取得90 家清除業者、約 1,753 車次清運車輛的 GPS 軌跡資料。此外，評估後續納入更多環境保護相關資料集，包含清除處理機構、資源再利用、掩埋場容量等系統資訊，為深化分析應用奠定更堅實的基礎，並研擬司法院裁判書文字資料處理方法，規劃後續導入系統分析。 為提升資料分析效率，完成「資源循環流向架構圖」和「重點事業廢棄物」、「主題案例」、「橫向計畫主題」不同邏輯，以來源、特性、成分等面向分類廢棄物，提供使用者多元分析視角，並開發互動式篩選及向下鑽取功能。同時，優化系統介面，強化各主題戰情室資訊呈現，整合全國性概況圖表、數據排名和篩選功能，並利用循環流向圖呈現清運量，提升使用者體驗和數據分析效率。此外，也深入分析違規業者申報資料，建立預警模型，開發「申報量增減異常」及「去化異常之自動化示警」兩種異常樣態查詢功能，協助管理者快速掌握異常申報行為，強化對高風險廢棄物的管理。 為強化勾稽機制，本計畫建立更全面性的同業水準分析模式，考量事業單位身份別、製程、廢棄物種類等因素進行分群，並應用於勾稽預測，提升勾稽預測的準確性和效率。同時，利用 GPS 車跡資料識別潛在的車輛違規行為，並建議結合其他佐證資料交叉比對，持續優化分析方法。此外，更開發「再利用機構上下游關係圖」功能，以視覺化圖表呈現複雜的再利用產品供應鏈網絡，並整合知識資料庫中的異常樣態資訊，協助使用者辨識潛在風險。 為強化應變能力，開發了自動化快速報告服務功能，提供「固定式」和「緊急應變」兩種自動化報告，協助使用者快速掌握事業廢棄物管理資訊。開發「全國異常月份數統計報表」功能，提升使用者針對異常月份數資訊的掌握。針對資料開放議題，協助研擬以「逐步開放」策略，分階段擴展事業廢棄物原始資料執行原則。此外，也利用歷史申報數據進行趨勢預測和異常預警，導入 ARIMA 和 VAR 模型預測廢棄物產出、貯存及聯單清運數據，產出量能警示圖表和預警名單，協助主管機關掌握潛在的廢棄物去化問題，強化系統韌性。 本計畫高度重視資訊安全與系統維運，遵循嚴格的資訊安全規範，定期進行弱點偵測掃描、源碼掃描、安全更新，並制定完善的標準作業程序和應變機制，確保系統安全。同時，落實使用者帳號管理，定期盤點久未登入的帳號，以確保系統穩定運行，並有效保護敏感資料。 為促進系統的發展與應用，本計畫積極參與多項會議及活動，推廣計畫成果並促進跨單位交流，提升專案能見度。同時，亦邀請學者專家召開線上諮詢會議，探討模式模擬預測精進、強化數位治理增進決策支援等議題，以持續優化系統功能，為環境監管和決策提供可靠的資訊服務。

資料科學、申報資料分析、申報流向分析

@113結案定稿本.pdf

24MB

2025/2/1 上午 12:00:00

This project focuses on developing an intelligent analysis and reconciliation system for the flow of industrial waste to enhance environmental regulation and improve resource recycling efficiency. To achieve this goal, the project emphasizes six key aspects: database expansion and refinement, application of artificial intelligence technologies, strengthening of intelligent reconciliation mechanisms, development of emergency response tools and data transparency, information security and system maintenance, as well as continuous promotion and improvement efforts. In terms of data, the project integrates diverse datasets to enhance both breadth and depth. It includes over 2.9 million inspection and penalty records, 5,000 environmental court rulings, and business registration data from the Ministry of Economic Affairs. Additionally, GPS trajectory data from 90 disposal operators and approximately 1,753 transport vehicles have been collected. Plans are also in place to incorporate additional environmental protection datasets, such as information on disposal facilities, resource reuse, and landfill capacities, laying a solid foundation for deeper analytical applications. Methods for processing judicial rulings' textual data are also being developed for future system integration. To improve data analysis efficiency, the project has completed resource flow diagrams and created logic-based classifications for "key industrial waste," "thematic cases," and "cross-plan topics" based on source, characteristics, and composition. These enable multi-perspective analysis for users, supported by interactive filtering and drill-down capabilities. The system interface has been optimized to enhance thematic dashboards, integrating nationwide visualizations, data rankings, filtering functions, and resource flow diagrams for transport volume analysis, improving user experience and analytical efficiency. Furthermore, the project has analyzed abnormal reports from non-compliant operators, developing early warning models and two automated anomaly detection functions—"abnormal reporting volume changes" and "disposal anomalies." These tools help administrators identify irregularities and manage high-risk waste effectively. To enhance reconciliation mechanisms, a comprehensive industry benchmarking analysis model has been established, considering factors like business type, processes, and waste categories for clustering and predictive reconciliation. GPS trajectory data is used to identify potential vehicle violations, with recommendations to incorporate other corroborative datasets for cross-validation. Additionally, the "Upstream-Downstream Relationship Diagram for Recycling Organizations" feature has been developed to visualize complex supply chain networks of recycled products. It integrates anomaly patterns from the knowledge database, assisting users in identifying potential risks. To strengthen emergency response capabilities, the project developed an automated rapid reporting service, offering both "standard" and "emergency response" automated reports. This enables users to quickly access industrial waste management information. A "National Anomaly Month Statistics Report" function was also created to provide insights into anomalous months. On data transparency, a phased "gradual opening" strategy was proposed for releasing raw industrial waste data. Historical reporting data has been used for trend forecasting and anomaly prediction, incorporating ARIMA and VAR models to predict waste generation, storage, and transportation data. Capacity warning charts and early warning lists were developed to help authorities address potential waste disposal challenges and enhance system resilience. The project places great importance on information security and system maintenance, adhering to strict security standards through regular vulnerability scans, source code reviews, security updates, and well-established SOPs and response mechanisms. User account management is enforced, with periodic reviews of inactive accounts to ensure stable system operation and protection of sensitive data. To promote system development and application, the project actively participates in various meetings and events to share achievements and foster cross-departmental collaboration, increasing project visibility. Scholars and experts were invited to online consultations to discuss predictive model enhancements and strengthen digital governance for decision-making support. These efforts aim to continuously optimize system functionality and provide reliable information services for environmental regulation and decision-making.

Data science, Application data analysis, Declaration flow analysis