跳到主要內容 :::

年度	113	專案性質	實驗性質
專案類別	模場試驗	研究主題	整治
申請機構	國立臺灣大學	申請系所	環境工程學研究所
專案主持人	駱尚廉	職等/職稱	特聘教授
專案中文名稱	以人工智慧結合遠端監控優化場址之現地整治成效
中文關鍵字	人工智慧,整治成效優化,遠端監控,機器學習
專案英文名稱	Optimization of remediation performance with artificial Intelligence and remote sensing Technology
英文關鍵字	Artificial Intelligence, machine learning, remote sensing, remediation performance optimization
執行金額	2,460,000元
執行期間	2024/12/15 至 2025/11/29
計畫中文摘要	本研究結合人工智慧（Artificial Intelligence, AI）與物聯網遠端監控（IoT）技術，針對石油類污染場址整治系統，建立即時預測與操作優化模型，以提升整治效率、能源利用率與操作智慧化水準。研究以多場址整治歷程數據為基礎，透過 Kernel Ridge Regression (KRR) 與 Random Forest (RF) 等模型，分別建立污染潛勢預測與抽氣效能模擬模型。進一步發展之 RF-QC 兩階段模型可預測下一小時之污染物移除效率，並依效率門檻自動提供開關機與運轉時長建議。研究同時導入生命週期評估（Life Cycle Assessment, LCA）方法，評估能源消耗與碳排放，建立碳效率（CE, g/kg CO₂）及環效比（EER）指標，以平衡整治成效與永續表現。模擬結果顯示，最佳化操作策略可在節能 38% 的條件下維持約 87% 的污染移除量，單位能耗密度下降 41%，碳排放降低 40%。此外，AIoT 遠端監控系統能即時分析抽氣頻率、真空度與 PID 變化率，並自動產生操作建議，顯著降低人力與交通成本。本研究建立一套整合 CSM 模型、機器學習預測、LCA 能碳評估與遠端監控回饋的智慧整治決策架構，展現整治作業智慧化與低碳化之實質效益，符合國家 Green & Sustainable Remediation (GSR) 之推動方向。
計畫英文摘要	This study integrates Artificial Intelligence (AI) with Internet-of-Things (IoT)–based remote monitoring to optimize the operational performance of petroleum-contaminated site remediation systems. Based on multi-site operational datasets, Kernel Ridge Regression (KRR) and Random Forest (RF) models were developed for remediation potential assessment and dynamic extraction efficiency prediction. A two-stage RF-QC model was further established to forecast the next-hour removal efficiency and to provide automatic on/off and duration recommendations based on defined efficiency thresholds. Life Cycle Assessment (LCA) was incorporated to evaluate energy consumption and carbon emissions, introducing the Carbon Efficiency (CE, g/kg CO₂) and Environmental Efficiency Ratio (EER) indices to balance remediation effectiveness and sustainability. Simulation results indicated that the optimized operational strategy reduced energy use by 38% while maintaining 87% of the removal efficiency, lowering energy intensity by 41% and total carbon emissions by 40%. The AIoT monitoring system also enabled real-time analysis of vacuum pressure, flow rate, and PID variation, offering on-site operational guidance and reducing manpower and transportation costs. Overall, this research establishes an integrated decision-making framework combining Conceptual Site Models (CSM), machine learning prediction, LCA-based energy-carbon evaluation, and remote monitoring feedback, demonstrating significant advances toward intelligent, low-carbon, and sustainable remediation practices in alignment with national GSR goals.