近年來流行的人工智慧正逐漸影響著我們的世界。透過人工智慧相關技術如機器學習及深度學習，可以對大量數位資訊進行分析，進而對於未知領域進行闡述或預測。傳統地下水參數推估數值模式可估算含水層參數與空間差異。基於機器學習之隨機森林技術，我們開發一個自動辨識模型，可以反推估含水層水力傳導係數(K)。首先利用合成測試案例建立訓練與驗證資料，完成所開發的自動辨識模型的效能與準確性。本研究於桃園觀音工業區進行抽水試驗，利用現場地下水水位觀測數據，應用傳統反推估模式與隨機森林模型完成模式驗證與比較。本研究規劃3階段工作項目：(1) 將引進創新科技，透過新興機器學習技術，建立隨機森林自動辨識模型，開發具有自動推估水力傳導係數空間分佈模型。(2)開發LoRaWAN低功率遠距離傳輸系統。 (3) 應用現地水為觀測數據測試與驗證開發模型。 ;AI (Artificial Intelligence) has increased its influences on the world nowadays. AI techniques such as machine learning and deep learning can analyze information from huge digital data and predict the trend in the unknown area. In this study, we develop an automatic identification model that combines random forest technique and properties estimation method of the traditional aquifer to analyze the distribution of hydraulic conductivity (K). In this study, we implement the training and verification of the model based on the real data obtained from pumping experiment in Guanyin industry site to improve the accuracy and efficacy of this model. It aims to finish this work, we arrange our work schedule into 3 steps, including: 1. Develop a new estimating model that can analyze the spatial distribution of aquifer properties with random forest technique. 2. Exploit LoRaWAN data transfer system with the ability of low-power consuming and long distance transporting. 3. Verify the accuracy and efficacy of this model with in-situ pumping data.
