環境資源報告成果查詢系統

110年空污季高風險空氣污染事業智慧勾稽實務應用計畫

中文摘要 本計畫透過以下五大工項,輔助空污稽查之成效與提升決策品質: 1. 建立智慧勾稽模組辦理特殊製程申報資料智慧勾稽。 本工項目標是將稽查專家已有的通則與方向,進行量化分析處理。產出稽查目標工廠,達成實質稽查成效。針對中南部19縣市107~109年之事業空氣污染防制費申報資料、固定污染源許可原始資料與事業廢棄物申報資料之特定64項製程,進行資料處理、申報合理性分析判讀,處理共計202萬筆資料,產出兩大模組「智慧督察-空空勾稽」、「智慧督察-空廢勾稽」透過異質資料比對篩選疑似短漏報空污費業者,並建置可篩選之工廠地圖視覺化儀表板。此次透過篩選,具體提交3間高可疑工廠名單,作為後續進場稽查查核項目。 2. 微型感測器數據分析。 本工項目標是配合空污季,針對PM2.5, VOCs 進行跨年度12個月(包含108年10月~109年3月、109年10月~110年3月)微型感測器分析,以輔助實際稽查行動,產生實質稽查成效。針對中南部縣市,挑選5大工業區(雲林斗六工業區,高雄林園工業區、臨海工業區、大發工業區,屏東屏南工業區),透過感測器歷史數據,提供PM2.5、VOCs潛勢污染源(區域)分析評估,找出最異常的高潛勢熱區,並提交可疑工廠清單,以輔助空污季之稽查重點限縮。 3. 發展環境執法系統群數位治理應用機制提供決策分析。 本工項目標為依據內外部系統資料媒合情境整合議題,發展系統性之決策分析工具。計畫已完成環境執法數位治理之軟硬體配置及各項議題展示雛型規劃,並依據其中二項議題「非法棄置追蹤」、「情節重大案件」提出內外部三個系統資料作為議題規劃時的資料媒合及加值來源。 4. 規劃提升環保稽查處分管制作法及友善度。 本工項目標為藉由業務導向將各式功能以業務分類區隔,降低使用者自行找尋功能位置的模糊感。因應環保稽查處分系統上線多年,系統功能服務介面須加以整合,有鑑於此,本計畫提供三個系統操作介面調整版面,並設計規劃快捷功能設定,以提升系統操作友善度。 5. 配合相關行政作業,以提升計畫執行成效及品質。 協助本計畫相關行政工作、諮詢服務,相關研商、說明會之會議資料之彙整、紀錄及分析等,及協助辦理與本計畫相關之幕僚工作。如:協助6家中南區事業對象,進行智慧勾稽;鎖定1間高可疑工廠進行熱區分析;配合空污季稽查宣導報告,110年12月前往南區大隊開會進行高風險污染熱區說明。
中文關鍵字 異質系統資料智慧勾稽、資料視覺化儀表板、列管工廠、熱區分析、環境執法成果、執法關切議題

基本資訊

專案計畫編號 經費年度 110 計畫經費 4180 千元
專案開始日期 2021/08/31 專案結束日期 2021/11/30 專案主持人 楊雅晴
主辦單位 環境督察總隊 承辦人 洪文啟 執行單位 卡米爾股份有限公司

成果下載

類型 檔名 檔案大小 說明
期末報告 總隊_1210_成果報告定稿_有馬賽克_公開版.pdf 13MB 10年空污季高風險空氣污染事業智慧勾稽實務應用計畫-成果報告(公開版)

The 2021 Practical Audit Project for High-Risk Air Pollution Industries during Serious Air Pollution Period

英文摘要 In this plan, we support air pollution inspections to gain higher effectiveness and enable authorities to make decisions with higher quality through the following works: 1. Building AI audit modules to conduct special declaration of data. The goal is to quantitatively analyze and process the existing general rules and directions of the audit experts, and then list down audit target factories to achieve substantial audit results. For all the data declared from central and southern 19 cities in Taiwan between 2018 and 2020, including reports of industrial air pollution prevention and control expenses, raw data of stationary pollution source permits (in specific 64 kinds of processes only) and data of industrial waste reports, we conducted data processing and rationality analysis and interpretation of  declaration. In addition, we processed a total of 2.02 million of data and produced two main modules. The first module, in which we make use of AI supervision and audit to automatically check the amount of declared air pollution fee and permitted amount of different factories. The second module uses AI supervision and audit to check the consistency between the amount of declared air pollution fee and declared waste. With heterogeneous data comparison, we filtered out and pinpointed factories which are suspicious of under-reporting air pollution fees, and then we established a map-oriented data visualization dashboard which contains filtering functions for easier screening of factories. Through screening and with the help of AI, lists of three pinpointed highly suspicious factories were submitted as a follow-up inspection item. 2. Data analysis on microsensors. The goal is to coordinate with the “Serious Air Pollution Period” and conduct a 12-month microsensor data analysis across the year for PM2.5 and VOCs, including data from Oct. 2019 to Mar. 2020 and Oct. 2020 to Mar. 2021, hence to assist actual inspection actions and produce substantial inspection results. Selecting five major industrial areas located in central and southern cities of Taiwan: Douliu industrial park in Yunlin, Linyuan industrial park, Linhai industrial park and Dafa industrial park in Kaohsiung and Pingnan industrial park in Pingtung, we provided analysis and evaluation of potential pollution sources (areas) regarding PM2.5 and VOCs  through historical data from sensors, pinpointed hot-zones with the highest possibilities of air pollution, and listed down suspicious factories to assist air pollution inspectors to narrow down the audit highlights in serious air pollution period. 3. Develop digital governance applications for environmental enforcement systems to provide decision analysis. The goal is to develop a systematic decision-making analysis tool based on internal and external system data matching context integration issues. In this project, we completed the software and hardware configuration as well as the prototype planning for the digital governance of environmental law enforcement. According to two issues including illegal disposal and major cases tracking, we proposed three internal and external systems for data matching and value-added sources when planning related topics in this project. 4. Plan and improve the method and friendliness of environmental enforcement management.  The goal is to divide various functions into business classifications based on business orientation, reducing the ambiguity of users searching for locations of specific functions on their own. The EEMS has been operating for years, the system operation layouts should be integrated. This project provided three system layouts for adjustment. The quick access functions were designed and planned to improve user-friendliness. 5. This project has cooperated with related administrative tasks to improve the effectiveness and quality of project implementation. Cameo assisted this project by performing relative administrative works, consulting services, and organizing, recording and analyzing conference data of related business research and seminars. Last but not least, assisting execution and management of staff work related to this project. For instance, (1) assisting 6 agencies in central and southern cities of Taiwan conducting AI audits. (2) targeting a highly suspicious factory due to precise hot spot analysis. (3) cooperating with reports inspection teams, such as visiting the Southern District inspection team to discuss high-risk pollution hotspots in December.
英文關鍵字 Audit for heterogeneous data integration using artificial intelligence (AI), Dashboard of visualized data, Restricted and classified factories, Hot-zone analysis, Environmental law enforcement results, Enforcement topic of concern