郭振泰臺灣大學：土木工程學研究所謝斌暉Hsieh, Pin-HuiPin-HuiHsieh2007-11-252018-07-092007-11-252018-07-092007http://ntur.lib.ntu.edu.tw//handle/246246/50530本研究的目的在以集水區為尺度建立一個最佳管理作業的最佳配置模式。模式包括三個互動的單元：集水區暴雨模式（本研究採用VAST 及BASINS 模式）、零維水庫水質模式（非穩態、時變性的Vollenweider 模式）及一個最佳化模式（遺傳演算法，GAs）。 本研究三個主要任務為：1.暴雨期間集水區水文及水質的影響；2.發展一個可信賴、有效率的最佳化模式整合最佳化工具與水庫水質模式，來尋找最佳管理策略系統的最佳配置方式；3.考量總懸浮固體物及總磷濃度，且有三種最佳管理作業納入模式中。 本研究發展一個後處理程式，包括：最佳管理作業模擬、河道演算、水庫水質模式及遺傳演算法等。集水區暴雨模式的結果，如VAST 及BASINS 的流量與污染物濃度，被當成一個反應矩陣。程式測試的結果發現單一小的暴雨並沒有對水庫水質產生嚴重的影響，換句話說在此狀況下不需要任何最佳管理作業。程式可以依據不同的染色體數量在被接受的時間內找到最佳解。經由調整不同的污染物濃度限制式，本研究分析三種不同的水質情況。總磷濃度必須滿足貧養、中養與優養的水質；同時，總懸浮固體物濃度必須控制在25 ppm 以下。 真實降雨事件發生在03/28/2003-04/14/2003 及 2002 至2003 年間分別被用來檢定與驗證VAST 與BASINS 模式。模擬的結果顯示VAST 與BASINS 模式在翡翠水庫集水區是好用的暴雨分析工具。縱使非穩態、時變性的Vollenweider 模式僅有一個污染物沈降速率，我們採用2002 年台北翡翠水庫管理局的總懸浮固體物與總磷濃度的變化與模式模擬結果做比對，其結果亦相當良好。 本研究共分析三種時間尺度的情況，包括：短期、長期及單一設計暴雨等。其中，以設計暴雨的結果每年需要投入最多的經費，因為2 年/2 小時、5 年/4 小時是謝等人（2003 年）找到的最強的暴雨能夠對水體水質產生最大的衝擊。在1988 到2004 年間，2002 年是最乾旱的一年，分析所得的所需年經費亦相對較小。 本研究所搜尋到的接近最佳解也與翡翠水庫集水區最佳管理作業配置的相關研究做比較，並採用他們的配置最為搜尋前的初始染色體。本研究採用的尖峰濃度限制較前人因為簡化的需要所用的平均濃度要大；但由於非穩態的優勢能顯現出真實暴雨污染物的傳輸現象，本研究能找到比前人研究更經濟省錢的最佳管理作業配置方式。決策者或設計者可以採用本研究所得的最佳管理作業配置方式，並應充分瞭解最佳管理作業的一些限制，如果有地形或用地取得不易等因素，許多如范與張（2006 年）建議的非結構性最佳管理作業也可以用來當作另一個非點源污染控制的選擇。The purpose of this study is to establish an optimization model for the optimalplacement of structural Best Management Practices (BMPs) at the watershed scale. It consists of three interacting components, i.e. a watershed stormwater simulation model(VAST/BASINS), a zero-dimensional reservoir water quality model (unsteady, time variable Vollenweider model), and an optimization module (the genetic algorithms, GAs). Three major tasks are completed in this work: 1) the investigation of how the watershed hydrology and water quality is affected during the storm period; 2) develop a reliable and effective optimization framework coupling the optimization tool and the reservoir water quality model to explore the optimal placement strategies of BMPs system; and 3) Both total suspended solids (TSS) and total phosphorus (TP) are taking into account and three kinds of BMPs are considered. A post treatment program is developed which combines BMPs simulation, channel routing, reservoir water quality models and GAs. The results analyzed by watershed stormwater models, such as VAST and BASINS models, are taken as a response matrix. The program test results show that a single small storm does not cause significant effect on the reservoir water quality, i.e. no BMPs is required under such situation. The program also finds the optimal solutions in an acceptable time depending on the chromosome population. By changing the pollution concentrations in the constraints, three scenarios are analyzed in this study. The TP concentrations should satisfy oligotrophic, mesotrophic and eutrophic conditions while the TSS concentrations should be kept below 25 mg/l at the same time. Real storm events during 03/28/2003-04/14/2003 and 2002-2003 are adopted to calibrate and validate VAST and BASINS models respectively. The modeling results demonstrate thatVAST and BASINS are good stormwater analyzing tools in Fei-Tsui Reservoir watershed. Although there is a pollutant settling rate in the unsteady, time variable Vollenweider model, the TSS and TP concentration variations in 2002 are also compared with the field data observed by Taipei Fei-Tsui Reservoir Bureau. There are three different time scale scenarios analyzed in this study, they are short term, long term and single design storm. The results analyzed by design storms cost much more money than the other two. Because the 2-year/ 2-hour and 5-year/ 4-hour storms are the critical storms found by Hsieh et al. (2003) that cause the serious damages on the water body. During 1988 to 2004, the 2002 is the driest year so that the annual costs analyzed by the program are relatively smaller. The near optimal solutions found in this study are compared with former research in Fei-Tsui Reservoir watershed. This study also uses their placement strategies as one of the initial chromosome. The peak pollutant concentrations are larger than the averaged values they used in former studies. Due to the advantage of unsteady state model, more economical placement strategies are found and suggested in this study. The decision makers or designers can adapt the BMPs placement strategies found in this study and should be good at the limitations of BMPs. Several non-structural BMPs proposed by Fan and Chang (2006) in Fei-Tsui Reservoir watershed can be good alternatives for nonpoint source pollution control when there are topographic and land ownership problems in the future.誌謝 I 摘要 ii Abstract iv Table of Contents vi List of Tables viii List of Figures ix 1. Introduction 1-1 1.1 Background 1-1 1.2 Objective and Scope of the Study 1-2 1.3 Thesis Outline 1-5 2. Literature Review 2-1 2.1 Current Comprehensive Watershed and Reservoir Models 2-1 2.1.1 Watershed Models 2-1 2.1.2 Reservoir or Lake Models 2-8 2.2 Application of Optimization Techniques in Stormwater Management 2-9 2.3 BMPs Related Studies 2-12 2.3.1 Construction and Maintenance Cost of BMPs 2-12 2.3.2 Treatment Efficiency of Structural BMPs 2-14 3. Model Development 3-1 3.1 Study Area 3-1 3.2 Stormwater Quantity and Quality Simulation 3-5 3.3 Post Treatment Model for BMPs Placement 3-14 3.3.1 BMP Models 3-14 3.3.2 Channel Routing Model 3-18 3.3.3 Reservoir Water Quality Model 3-19 4. Genetic Algorithms 4-1 4.1 Objective Function and Constrains 4-3 4.2 Encoding 4-4 4.3 Selection 4-7 4.4 Crossover 4-8 4.5 Gaussian Mutation 4-9 4.5 Penalty Function Implementation and Elitist 4-10 5. Simulation and Optimization Results 5-1 5.1 Post Treatment Program 5-1 5.2 Scenarios Assumptions and Results 5-6 5.2.1 Short term storm period 5-8 5.2.2 Long term/ yearly period 5-9 5.2.3 Single design storm 5-10 5.3 Comparisons with Related Research 5-13 6. Conclusions and Recommendations 6-1 6.1 Conclusions 6-1 6.2 Recommendations 6-3 REFERENCES Table 2-1 Summary of current comprehensive watershed models 2-3 Table 2-2 Annual maintenance costs 2-13 Table 2-3 Capital cost functions for wet-weather controls 2-13 Table 2-4 Structural BMP expected pollutant removal efficiency 2-14 Table 2-5 Structural BMP expected pollutant removal efficiency 2-15 Table 2-6 Summarized BMP treatment efficiencies 2-15 Table 3-1 Landuse changes between 1998 and 2004 3-4 Table 3-2 Pollution loads in Fei-Tsui Reservoir watershed 3-4 Table 3-3 Monthly averaged flow discharge during 1997-2004 3-4 Table 3-4 List of Fei-Tsui reservoir watershed division 3-7 Table 3-5 The estimated pollutant loading rates of VAST model for the entire Fei-Tsui Reservoir watershed 3-9 Table 3-6 Conversion of total cost to annual cost of BMPs 3-18 Table 4-1 Water quality standard of TSS in water bodies 4-3 Table 4-2 Water quality standard of TP in reservoirs 4-4 Table 4-3 Pre-definition results of each region 4-6 Table 5-1 Results of different initial chromosome population 5-5 Table 5-2 Three scenarios modeled in this study 5-7 Table 5-3 Short term storm period optimization results (03/28-04/14, 2003) 5-8 Table 5-4 Long term storm period optimization results in 2002 5-9 Table 5-5 Long term storm period optimization results in 2003 5-10 Table 5-6 Placement strategy evaluated by the 2-year/2-hour storm 5-11 Table 5-7 Placement strategy evaluated by the 5-year/4-hour storm 5-11 Table 5-8 Related research on BMPs placement 5-13 Table 5-9 Steady-state model used annual averaged inflow and pollutant loads of 2003 and discrete differential dynamic programming (DDDP) 5-14 Table 5-10 Steady-state model used annual averaged inflow and pollutant loads of 2002 and discrete differential dynamic programming (DDDP) 5-14 Table 5-11 Steady-state model used annual averaged inflow and pollutant loads for 12 years and discrete differential dynamic programming (DDDP) 5-15 Figure 1-1 System-wide procedure of this study 1-4 Figure 2-1 Flow chart of WinVAST 2-15 Figure 3-1 Fei-Tsui Reservoir watershed 3-2 Figure 3-2 Landuse distribution of Fei-Tsui Reservoir watershed 3-3 Figure 3-3 Water quality sampling points of Fei-Tsui Reservoir watershed 3-3 Figure 3-4 Binary tree of Fei-Tsui Reservoir watershed 3-6 Figure 3-5 Calibration and validation results of VAST model during 3/28/2003 - 4/9/2003 3-8 Figure 3-6 The interface and settings of BASINS model in Fei-Tsui Reservoir watershed 3-10 Figure 3-7 Stormwater runoff simulation results in Fei-Tsui Reservoir watershed during 2002/1/1 - 2003/12/31 3-11 Figure 3-8 TSS simulation results in Fei-Tsui Reservoir watershed during 2002/1/1 - 2003/12/31 3-12 Figure 3-9 PO4–P simulation results in Fei-Tsui Reservoir watershed during 2002/1/1 - 2003/12/31 3-13 Figure 3-10 Typical wet detention pond design 3-15 Figure 3-11 Grass swale in Seattle, U. S. A. 3-16 Figure 3-12 Schematic design of a buffer strip 3-17 Figure 4-1 Genes permutation in a chromosome 4-5 Figure 4-2 One and multipoint crossover 4-8 Figure 5-1 A simple interface of the program 5-2 Figure 5-2 The format of program setting file – config.dbf. 5-2 Figure 5-3 Storm runoff and pollutant concentration input format (the response matrix) 5-3 Figure 5-4 Comparison of the simulation results of TSS concentrations using unsteady, variable volumes Vollenweider model with observed data in 2002 5-4 Figure 5-5 Comparison of the simulation results of TP concentrations using unsteady, variable volumes Vollenweider model with observed data in 2002 5-4 Figure 5-6 Convergence of different initial chromosome 5-5 Figure 5-7 Critical storms of Fei-Tsui Reservoir watershed (Hsieh et al., 2003) 5-7 Figure 5-8 Trade-off relationships between BMPs costs analyzed in this study and drinking water treatment cost 5-122703884 bytesapplication/pdfen-US最佳配置策略水區暴雨模式最佳管理作業模擬河道演算水庫水質模Optimal placement strategyWatershed stormwater modelBest management practice modelChannel routingReservpor water quality modelGenetic algorithmthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/50530/1/ntu-96-D90521002-1.pdf