碩士 / 國立臺灣海洋大學 / 應用地球科學研究所 / 104 / The pleasant weather and abundant water resource at Pingtung plain, southern Taiwan are very suitable for growing agriculture and aquaculture. In 1970’s, the aquaculture was rapidly developed and the groundwater was over-pumped. These activities caused the environment hazards such as seawater intrusion and subsidence. The seawater intrusion resulted in serious problems and harm to human’s life. The aquifers were contaminated and very difficult to be recovered, and the water resources cannot be used for a long time. In order to remedy the contaminated groundwater and find the best strategies of groundwater usage, a management model which coupled with numerical model and optimal algorithm is established. A three-dimensional density-dependent flow and transport simulation model, called SEAWAT and developed by U.S. Geological Survey, is selected to simulate the phenomenon of seawater intrusion at the coastal area of Pingtung plain. The optimal algorithm of differential evolution (DE) is applied to solve the management problem. When solving the management problem, the numerical model should be run for many times which will take much time. A surrogate model called artificial neural network (ANN) is used to reduce the computational time. The complete-trained ANN model will be used to replace the numerical model and then coupled with DE algorithm. In this study, the objective function of the management model is to minimize the total amount of injection water and a set of concentration constraints is applied to ensure the targets are met. Two different scenarios are set to remedy the seawater intrusion. First scenario is to set the injection wells along the coastal line forming as a barrier and expect to decrease the driving force of seawater intrusion. Second scenario is to set the injection wells in each aquifer based on the areas with higher concentrations. These areas are located at inland and the purpose is to improve the water quality at these areas. The result shows that the complete-trained ANNs can substitute the numerical simulation model and produce the concentration data accurately and efficiently. Under appropriate setting of objective function and constraints, DE can find the optimal injection rates at predefined barriers and the concentrations at the target locations can meet the requirements within the planning horizon of 20 years. Through proper setting of management model and suitable selecting of optimal algorithm, the management model can provide the reliable information and suggestion to remedy the seawater intrusion. The optimal theory could also be applied to other issues, such as water resource management and remediation of contaminated plume.
Keyword: seawater intrusion, groundwater management model, SEAWAT, artificial neural network, differential evolution
| Identifer | oai:union.ndltd.org:TW/104NTOU5135007 |
| Date | January 2016 |
| Creators | Huang, Po-Syun, 黃柏勳 |
| Contributors | Chiu, Yung-Chia, 邱永嘉 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 112 |
Page generated in 0.0025 seconds