水資源監控平台之建立與應用

近年來面對氣候變遷，跨區或跨標的水資源調度、海淡水與再生水開發及合理運用地下水源，成為穩定供水的重要方式，物聯網（Internet of Things, IoT）為現今資訊科技發展重要趨勢之一，其應用提供一種以預測性維護概念，減少設備異常或無法運作發生的機率，以同步改善、建構優質的水生活環境包括改善水資源管理，減輕水災水患，並確保更便利且更可持續享有衛生設施。本研究主要探討「水資源資訊管理系統工作平台」在三種不同場域(海綿城市、污水維護預測及工業園區廢污水監測系統)應用結果，將分散於各地污水處理廠設備維護資訊、即時水質監測資訊，利用遠端網路監控式系統平台技術，將蒐集資料整理分析或建立水質水量預測模式。「海綿城市」以兩種多孔舖面進行15個月時間測試以每10分鐘之間測頻率進行長時間的降雨及晴天時間量測，對地下水體儲槽進行監測，並以雨水管理模型模擬多孔舖面對於減少逕流之影響；「污水維護預測」於金門榮湖污水處理廠裝設WSN系統，實際監控各單元現況3個月，結果顯示該系統能夠即時呈現水位及設備運轉資料，瞭解及進一步分析該廠操作現況，達到預期目標及性能要求；「工業園區廢污水監測系統」於工業區內設置監測點，建立污水加壓站控制管理系統、排水防洪監測系統等，進行驗證量測值與準確值之差異性分析，R2值分別大於0.99，監測值具良好準確度，監測數據延伸應用於水質水量預測模式，以暴雨逕流管理模式模擬淹水及相關潛勢區域，其模擬情境為五年重現期設計暴雨，可模擬暴雨開始後淹水狀況發生的時間點，另水質預測模式則以類神經網路進行模式訓練分析比較，利用溫度、pH值、SS、COD四個輸入變數及到輸出預測目標 DO濃度，結果以XGBoost模型為最佳水質預測模式，其分類準確率達89%。綜合以上應用結果顯示利用即時監測設備全時監測不同項目，將測量數據及影像即時透過物聯網回傳，以使人員可於遠端監看即時測值，並整合各項目之監控平台，確保數據品質與準確度狀況下，不僅有助於設施設備監控預警，亦能檢討評估處理操作成效及預測模式分析，以達成預測性維護及災害預防，提供現場運作及建置改善的參考建議。

關鍵字

水資源即時監測；物聯網；無線感測網路

並列摘要

In recent years, the management of cross-regional or cross-standard water resources, development of desalination and reclaimed water technologies, and rational utilization of groundwater sources are vital principles to stabilize water supply under the effect of severe climate change. To achieve this purpose, the Internet of Things (IoT) is essential for bonding information technology and water management. IoT application provides a predictive maintenance concept to reduce the probability of abnormal or inoperable equipment and simultaneously improve and construct a high-quality water living environment, including water resources management advancement, flood prevention and reduction, and a convenience and steady facilities. This research mainly discusses the application results of the "Water Resources Monitoring Platform" in three different fields (sponge city, wastewater maintenance prediction, and industrial park waste and sewage monitoring system), which will collect equipment maintenance information and real-time water quality monitoring data from dispersed wastewater treatment plants. In this manner, water resources monitoring platform can analyze and establish a water quality and quantity prediction model. The water resources monitoring platform system demonstrated its advantages in a three-months period under the case study in Long Lake Wastewater Treatment Plant in Kinmen County. The prediction model shows a good accuracy when compared to the real observation (R2 value are greater than 0.99). The monitoring data is also extended to predict the water quality and quantity under the five-years deluge simulation scenario, which analyze temperature, pH value, SS, and COD to predict the output DO concentration. As a result, the XGBoost model reaches an 89% accuracy rate. The water resources monitoring platform not only provide help in monitoring and warning, but also examine, analyze and evaluate the effectiveness of processing operations and predictive mode analysis.