乾旱時期用水分配模式之研究---子計畫：乾旱時期地面水與地下水聯合運用作業模式研發(I)

(1)

行政院國家科學委員會專題研究計畫成果報告

乾旱時期用水分配模式之研究--子計畫:乾旱時期地面水與

地下水聯合運用作業模式研發(I)

研究成果報告(完整版)

計畫類別：整合型計畫編號： NSC 98-2625-M-009-005- 執行期間： 98 年 08 月 01 日至 99 年 07 月 31 日執行單位：國立交通大學土木工程學系（所）計畫主持人：張良正計畫參與人員：博士後研究：陳宇文博士後研究：葉明生報告附件：出席國際會議研究心得報告及發表論文處理方式：本計畫可公開查詢

中華民國 99 年 10 月 31 日

(2)

�

��

�

�� . . . I �� . . . II Abstract . . . III �� . . . 1 �� . . . 1 2.1 �� . . . 3 2.2 �� . . . 3 2.3 �� . . . 3 �� . . . 3 3.1 �� . . . 3 3.2 �� . . . 4 3.3 �� . . . 4 3.4 �� . . . 11 3.5 �� . . . 12 �� . . . 17 4.1 �� . . . 17 4.2 �� . . . 19 4.3 �� . . . 30 4.4 �� . . . 32 4.5 �� . . . 37 �� . . . 38 5.1 �� . . . 38 5.2 �� . . . 39 �� . . . 39

(3)

�

��

�

�� MODFLOW 2000 �� 89 �� 96 �� DRASTIC �� DRAS-TIC

(4)

Abstract

In the past decade, the rapid growth of industry and agriculture have also increased the quantity of water requirement in Chou-Shui Alluvial Fan. To fulfill the water demand becomes a great challenge for the present water supply system. Since the surface water is influenced significantly by the hydrological condition that varies greatly in time, depending only on surface water system has high water deficit risk during drought period. On the other hand, groundwater is a much more stable resource than surface water and it is a good alternative water resource to reduce water deficit during drought period. Therefore, the long term purpose of this study is to develop a surface and subsurface conjunctive-use management model that can facilitate to increase the water allocation capability and reduce the water deficit risk in Chou-Shui Alluvial Fan during drought period. To do so, the groundwater characteristic of Chou-Shui Alluvial Fan should be familiar first. Two steps include the development of steady numerical model and the estimation of the recharge potential have been purposed.

First, MODFLOW 2000 which is one of the most famous models for groundwater simulation is used to construct a steady model of Chou-Shui Alluvial Fan. The average observed water level are used in the calibration process. After calibration, the spatial distribution of the calibrated recharge quantity mostly centralizes at the proximal area on the top aquifer. The spatial distribution of the calibrated pumping quantity is related with the distribution of the manufacturing districts. Second, the estimation of the recharge potential is determined by using a factor based method which modified from DRASTIC, a groundwater pollutant potential model and seven factors including land use, surface soil, river density, average annual rainfall, correlation between rainfall and groundwater storage, variation of storage for unit aquifer area and hydraulic conductivity are used. The estimation result indicates that the high recharge potential zone locates in the proximal area include, Shetou township, Tianzhong township, Xizhou township, Ershui township, Citong township and Linnei township. The distribution of the estimation is similar with the one of the steady numerical model.

Keywords�Chou-Shui Alluvial Fan, groundwater characteristic, steady numerical model con-struction, model calibration, recharge potential and DRASTIC

(5)

�

��

�

��

�

��

�

�� 1: ��-��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

�� 1, 800 �� 0 � 100 ��

(6)

(7)

2.1 �

�

��

�

��

�

��

�

��

�

��

�

�� 88 �� 72 �� 300 �� 8 �� 250 �� (F1) �� (T1) �� (F2) �� (T2) �� (F3) �� (T3) �� (F4) �� (T4) �� 1 ��

2.2 �

�

��

�

��

�

��

�

��

�

�� (81 �� 87 ��) �� 82 � (203 �) �� 28 � (55 �) �� 88 �� 2 ��

2.3 �

�

��

�

��

�

��

�

��

�

��

�

�� (K) �� 10−3 _{� 10}−5 _{(m/s) �� (T ) �� 0.01 � 4.19 (m/min) �� (Q/s)�} � 0.34 � 208.33 (m3_/(hr_{· m)) ��}

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

�� 3.1 �� MODFLOW �� 3.2 �� 3.3 ��

3.1 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

�� 1950 �� 1970 �� MODFLOW �� (U.S.G.S) �� MOC3D � MT3D � MT3DMS �� MODFLOWP � PEST � UCODE �� (Inverse modeling) �� MODFLOW �� 1983 �� MODFLOW �� 1987 �� MODFLOW 88 �� 1996 �� MODFLOW 96 �� MODFLOW 2005 � MODFLOW ��

(8)

�� Processing MODFLOW � Visual MODFLOW � Groundwater Vistas � GMS ��

�� Groundwater Vistas 5.0 �� MODFLOW �� Groundwater Vistas � Environmental Simulations International �� 3D �� MODFLOW 2005�� Groundwater Vistas �� MODFLOW �� MODFLOW �� Groundwater Vistas ��

3.2 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

MODFLOW �� (�� ) �� (�� 3 ��) � � 3: ��

3.3 �

�

��

�

��

�

��

�

��

�

��

�

��

�

�� • �� 1. ��

(9)

MODFLOW �� 88 �� 72 �� (�� 300 ��)�� 8 ��(�� 250 ��)�� 4 �� 1 2 3 2 Tree Tree Tree Tree Tree Tree Tree Tree Tree Tree 1 3 4 � 4: �� 2. �� (1999) �� 5 �� ∗ AB �� 10 �� (No flow boundary) �� (Pointsource) � ∗ BC �� BC �� ∗ CD ��

(10)

� 5: �� ∗ DEF �� BC �� EF �� ∗ FG �� ∗ GA �� (Groundwa-ter divide)� 3. �� 1(km) × 1(km) �� 85 �� 75 �� 6 �� (�� 7 ��)�� 5 ��(� � 8 ��)�� 9 �� 47 �� 30 �� • ��

(11)

N E S W 0 6000 12000 24000 36000 � 6: �� 7: �� 1 ��

(12)

� 8: �� 2 � 7 ��

(13)

�� MODFLOW �� : 1. �� 10 � 13 �� 1 � 3 � 5 � 7 �� 2 � 4 � 6 ��

Schwartz � Zhang (2003) �� 4× 10−3 _{(m/day) ��}

�� 1/10 � 2. �� 1/4 � 1/5 �� 3. �� 4. �� 91 �� (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) ₍₁₎ (1) (1) (1) (1) (1) (1) (1) (1) (1) (1) N E S W 0 6000 12000 24000 36000 � 10: �� 1 ��

(14)

(2) (1) (2) (2) (2) (2) (2) (2) (2) (1) (2) (2) (2) (1) (1) (2) (1) (2) (2) ₍₂₎ (2) (2) (1) (1) (2) (1) (1) (2) (1) (1) (1) (1) (1) (1) (1) (2) (1) (1) (1) (1) (2) (1) (1) (1) (1) (1) N E S W 0 6000 12000 24000 36000 � 11: �� 3 �� (3) (2) (2) (2) (2) (2) (3) (2) (2) ₍₂₎ (2) (3) (2) (2) (3) (3) (3) (2) (2) (2) (2) (1) (1) (2) (2) (3) (2) (2) ₍₃₎ (2) (3) (3) (3) (3) (3) (3) (3) (3) (4) (3) N E S W 0 6000 12000 24000 36000 � 12: �� 5 ��

(15)

(2) (2) (2) (2) (2) (3) (2) (2) (2) (2) (1) (1) (2) (2) (4) (2) (3) (2) (3) (2) (2) ₍₂₎ (2) (3) (2) (3) ₍₃₎ (3) (3) (3) (4) (3) (4) (4) N E S W 0 6000 12000 24000 36000 � 13: �� 7 ��

3.4 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�� MODFLOWP � UCODE �� MODFLOWP �� (U.S.G.S) �� MODFLOW �� MODFLOW �� MODFLOWP �� -�� (Modified Gauss-Newton method) �� (Conjugate-direction method) �� UCODE �� (U.S.G.S) � IGWMC �� MODFLOW �� MODFLOW ��

(16)

��

�� (Zonation method) �� Emselln � de Marsily (1791)�� Cooley (1977�1979)�� S �� MODFLOW �� UCODE �� : 1. �� 89 �� 98 �� 2. �� (Zonation) �� 10 � 13 �� 3. � UCODE �� K �� 4. �� UCODE ��

3.5 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

� 14 � 16 �� 17 � 19 ��

(17)

(18)

(19)

(20)

� 17: ��(��)

(21)

� 19: ��(��)

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

�� (GIS) �� 0 � 100 ��

4.1 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

� 1: �� K ��

(22)

� 20: �� 20 �� (� � 1 ��)��DRASTIC �� 0 � 100 �� 1km × 1km �� AW Acno(i) = �Mcno j=1 Area (j)× RFi(j) �Mcno j=1 Area (j) (1)

(23)

�� Area (j) �� j �� AW Acno(i) �� i �� cno ��

�� (1) �� cno ��

T RFcno =

�N

i=1ω (i)× AW Acno(i)

�N i=1ω (i) (2) �� ω (i) �� i �� N �� T RFcno �� cno �� (2) ��

4.2 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

�� 1km × 1km �� (Polygon) �� (� (1) ) �� (Point) �� (Inverse Distance to a Power) ��

(1) ��

�� (Bou Kheir et al.,2003) �� (1) � ��(2) �� (3) �� (Shaban et al., 2006) �� (�� 21 ��)� (2) �� 0 − 30 (cm) � 30 − 60 (cm) � 60 − 90 (cm) � 90 − 150 (cm) � �� 22 � 25 �� (3) �� (K, hydraulic conductivity) �� K �� (��2.2-1)�� (F1 � F4) �� (T1 � T4) �� (F1) �� 42 �� 26 �� K �� 0.12 � 0.13(m/min) �� (4) �� 1993 � 2008 �� 27 ��

(24)

� 21: ��

(25)

� 23: 30 − 60 (cm) ��

(26)

� 25: 90 − 150 (cm) ��

(27)

� 27: ��(mm) (5) �� (�� 28 � �)� (6) �� 20 �� 20 �� 40 �� 29 �� 33 �� 30 �� R2 _{�� R}2 _{��} �� 20 ��

(28)

� 28: �� (3) (3) (2) (1) (2) (2) (4) (2) (2) N E S W 0 6000 12000 24000 36000

(29)

� 30: �� (1) �� 31 � 33 �� 0.804 � 0.691 � 0.519 � � 34 �� 3 �� R2 _{�� 35 �� R}2 _�� 1 �� (�� 36 ��)�� (7) ��

(30)

0 200 400 600 800 1000 Precipitation (mm) 0 1 2 3 4 5 G W _ le ve lC h a n g e (m ) y=0.004481x+0.162857 y=0.005033x R2_=0.804 R2_=0.724 � 31: �� (1) �� (��) 0 1000 2000 3000 Precipitation (mm) 0 1 2 3 4 5 G W _ le ve lC h a n g e (m ) y=0.001621x+0.241705 y=0.001864x R2_=0.691 R2_=0.621 � 32: �� (1) �� (��)

(31)

0 500 1000 1500 2000 2500 Precipitation (mm) 0 2 4 6 G W _ le ve lC h a n g e (m ) y=0.001979x+0.315834 y=0.002338x R2_=0.519 R2_=0.403 � 33: �� (1) �� (��) � 34: �� R2 _��

(32)

� 35: �� R2 _��

(33)

� 37: �� (m)

�� 37 ��

(34)

4.3 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

� 38: �� 2: �� ₂ ₁ _2.5 _19% �� 3 0 3.0 22% �� 1 0 1.0 7% �� ₂ ₀ _2.0 _15% �� 0 1 0.5 4% �� 2 0 2.0 15% �� 2 1 2.5 18% �� 13.5 ��(1.0); ��(0.5) ��Shaban �� (2006) �� Shaban et al. (2006) � Yeh et al. (2008) �� 38 ��

(35)

�� 38 �� 1.0 �� 0.5 �� 2.5 �� 2 ��

(36)

4.4 �

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

��

�

� 3: �� (�) �� ₁ �� ₁₀₀ 2 �� 90 3 �� 70 4 �� 60 5 �� 40 6 �� 20 7 �� 0 �� 11 �� 100 1 �� 90 2 �� 80 3 �� 70 4 �� 60 5 �� 50 6 �� 40 7 �� 40 8 �� 30 9 �� 20 10 �� 10 �� 1 0.22− 0.33 20 2 0.33− 0.45 40 3 0.45_{− 0.56} 60 4 0.56− 0.68 80 5 0.68− 0.79 90 6 0.79_{− 0.92} 100 �� 1 1114_{− 1330 20} 2 1330_{− 1545 40} 3 1545− 1760 60 4 1760_{− 1976 80} 5 1976_{− 2192 100} �� ₁ ₀_{− 0.19} ₂₀ 2 0.19_{− 0.38} 40 3 0.38_{− 0.58} 60 4 0.58− 0.77 80 5 0.77_{− 0.97} 100 �� (�� 20 ��2)�� 0 � 100 �� 3 � 4 �� 39 � 45 �� 1km × 1km �� 39 � 45 ��

(37)

� 4: �� (�) �� ₁ ₀_{− 0.02} ₂₀ 2 0.02_{− 0.04} 40 3 0.04_{− 0.06} 60 4 0.06− 0.9 80 5 0.9_{− 0.12} 100 �� ₁ ₀_{− 7} ₁₀ 2 7_{− 14} 20 3 14_{− 21} 30 4 21− 28 40 5 28_{− 35} 50 6 35_{− 42} 60 7 42− 49 70 8 49_{− 56} 80 9 56_{− 63} 90 10 63− 72 100 N E S W 0 6000 12000 24000 36000 grade grade 0 - 3.79 0 - 3.79 3.79 - 7.59 3.79 - 7.59 7.59 - 11.39 7.59 - 11.39 11.39 - 15.19 11.39 - 15.19 15.19 - 19 � 39: ��

(38)

N E S W 0 6000 12000 24000 36000 SUM_GRADE SUM_GRADE 0 - 4.11 4.11 - 8.23 8.23 - 12.35 12.35 - 16.47 12.35 - 16.47 16.47 - 20.59 16.47 - 20.59 � 40: �� N E S W 0 6000 12000 24000 36000 VALUE VALUE 1 3 4 6 6 7 � 41: ��

(39)

N E S W 0 6000 12000 24000 36000 VALUE VALUE 3 6 9 12 15 � 42: �� N E S W 0 6000 12000 24000 36000 GRADE GRADE 1 2 2 3 4 � 43: ��

(40)

N E S W 0 6000 12000 24000 36000 VALUE VALUE 3 6 9 12 15 � 44: �� N E S W 0 6000 12000 24000 36000 PERCENT PERCENT 2 4 5 7 9 11 13 14 16 18 � 45: ��

(41)

4.5 �

�

��

�

��

�

��

�

��

�

��

�

��

�

�� 4.4 �� 4.3 �� 5 �� 46 �� 46 �� 47 �� 5: �� 80 ��(Very High, VH) ��70�80�� (High, H) ��60�70�� (High Moderate, HM) ��50�60�� (Moderate, M) ��40�50�� (Low Moderate, LM) ��30�40�� (Low, L) ��30 ��(Very Low, VL)

H

0 2650 5300 10600 15900 公尺圖例補注潛勢分數總和 VL L LM M HM H VH � 46: ��

(42)

龍井鄉 _北屯區伸港鄉大肚鄉南屯區西區和美鎮線西鄉烏日鄉南區鹿港鎮大里鄉彰化市霧峰鄉秀水鄉福興鄉花壇鄉芬園鄉芳苑鄉埔鹽鄉草屯鎮大村鄉二林鎮員林鎮溪湖鎮埔心鄉南投市永靖鄉社頭鄉埤頭鄉田尾鄉大城鄉田中鎮北斗鎮竹塘鄉名間鄉溪州鄉二水鄉麥寮鄉竹山鎮二崙鄉崙背鄉 _西螺鎮莿桐鄉林內鄉台西鄉斗六市土庫鎮虎尾鎮褒忠鄉東勢鄉斗南鎮四湖鄉古坑鄉元長鄉大埤鄉口湖鄉水林鄉北港鎮 _溪口鄉大林鎮新港鄉梅山鄉民雄鄉六腳鄉竹崎鄉東石鄉太保鄉 _嘉義市番路鄉朴子鎮

H

0 2650 5300 10600 15900 公尺圖例台灣鄉鎮圖補注潛勢分數總和 VL L LM M HM H VH � 47: ��

�

��

�

��

�

��

�

��

�

��

�

��

�

5.1 �

�

��

�

1. �� 2. �� 89 �� 96 �� 3. ��DRASTIC�� 4. �� Shaban �� 5. ��

(43)

5.2 �

�

��

�

1. ��

�

��

�

��

�

��

�

��

�

��

�

1. Mondal, N. C. and V. S. Singh (2004) A new approach to delineate the groundwater recharge zone in hard rock terrain,Current Science, Vol. 87, 5, 10.

2. National Water Well Association,”DRASTIC: A Standardized System for Evaluating Ground Water Pollution Potential Using Hydrogeologic Settings”,USEPA,May 1985. 3. Schwartz, F.W., and Zhang, H. (2003) Fundamentals of ground water, John Wiley Sons,

New York.

4. Shaban, A., M. Khawlie, and C. Abdallah (2006) Use of remote sensing and GIS to determine recharge potential zone: the case of Occidental Lebanon, Hydrogeology Journal, 14: 433-443.

5. Yeh, H.F., Lee, C.H., Hsu, K.C., and Chang, P.H. (2008) GIS for the assessment of the groundwater recharge potential zone. Environ. Geol., Vol. 58, 185-195.

6. ��1997�� 7. ��1994�� 8. ��1995�� 9. ��1996�� 10. ��1998�� 11. ��1996�� 12. ��1996�� 207�221�� 13. ��2005�� 1-28�� 14. ��1996�� 165-179��

(44)

15. ��1997�� 96P� 16. ��1998�� 17. ��2000�� 18. ��1957�� 206�� 19. ��1973�� 20. ��2003�� 125-139�� 21. ��2009�� 22�� 22. ��1996�� 23. ��1999a�� pp.369- 376� 24. ��1999b�� pp.65-76 25. ��1996�� 181-206�� 26. ��1998�� 69P� 27. ��1995��102�� 28. ��1999��129�� 29. ��1996�1997��-�� 30. ��1999��

(45)

41

1. Ŷ

Ƒ

100

2. Ƒ

Ƒ

Ŷ

Ƒ

Ŷ

Ƒ

Ŷ

3.

500

(46)

99 10 31

z 12/14

Keynote

Bibring

z 12/15

z 12/16

12/17

NSC 98 2625 E 009 005

--

:

(I)

99

12

14

99

12

18 (

)2009 AGU

(

)AGU Fall Meeting 2009

(

)

(

) Integrating Evolution Algorithms with Linear Programming to the

Conjunctive Use of Surface and Subsurface Water

(47)

43

AGU

16,000

AGU

(

)

1.

2.

3. ( )

1. American Geophysical union 2008 Fall Meeting

:

AGU 2009

2. Fall Meeting

3. 2.

AGU 2009 Fall Meeting

(48)

Integrating Evolution Algorithms with Linear Programming to the Conjunctive Use of

Surface and Subsurface Water

Liang-Cheng Chang Yu-Wen Chen Chen-Che Pan

Department of Civil Engineering, National Chiao Tung University, Taiwan, ROC

ABSTRACT

Surface water resources are strongly influenced by hydrological conditions, and using only

surface water resources as water supplies may have higher shortage risk than before because

of the climate change caused by the global warming. Conjunctive use of surface and

subsurface water is one of the most effective water resource practices to increase water supply

reliability with minimal cost and environmental impact. Therefore, this paper presents a novel

stepwise optimization model for optimizing the conjunctive use of surface and subsurface

water resources management.

At each time step, a two level decomposition approach was proposed to divide the

nonlinear optimal conjunctive use problem into a linear surface water sub-problem and a

nonlinear groundwater sub-problem. Because of the two level decomposition approaches, a

hybrid framework is used for the implementation of the conjunctive use model. In the hybrid

framework, evolution algorithms, Genetic Algorithm (GA) and Artificial Neural Network

(ANN), and Linear Programming (LP) are used for model solving. GA and LP are respectively

used for determining the optimal pumping quantities and reservoir allocation, and ANN is

used for the groundwater simulation. In the groundwater simulation, this study uses an ANN

to simulate groundwater response and greatly reduce computational loading for unconfined

aquifers, unlike conventional “response matrix method” or “embedding method”. Because of

the very high performance of LP, the usage of LP for the linear surface water sub-problem can

significantly decrease the computational burden of entire model.

In this study, four cases have been demonstrated. Case #1 is a pure surface water case and

others are conjunctive use cases. In Case #2, “surface water supply firstly” is the supply

principle between surface water. In Case #3 and #4, the “Index Balance” theory is the supply

principle and different operation curves used in different cases respectively. The case result

has been compared as below. First, the case comparison between Case #1 and #2 indicates that

a small but reliable groundwater supply can significantly decrease the water supply risk. With

(49)

45

“surface water supply firstly” policy, an additional groundwater supply, whose maximum

supply capacity is just 8.67% of the total water demand, can significantly decrease the

shortage index (SI) from 2.93 to 1.54. Second, these comparisons between Case #2 and #3 or

#2 and #4 demonstrate that extending the principle of “rule curve” operation to groundwater

supply can dynamically balance the use of groundwater and surface water resources.

Therefore, the “rule curve” operation for groundwater also can further increase the operating

efficiency of water supply. Because of the higher operating efficiency of the proposed model,

the SI can significantly decrease to 1.13 or 0.79 depend on different rule curve of groundwater.

Finally, because of the usage of the hybrid framework and ANN, the computational efficiency

of proposed model is higher than other models with purebred framework or traditional

groundwater simulation. Therefore, the proposed model has ability to extend to solve a

complicated large field problem and also will be a valuable tool for conjunctive surface and

subsurface operations planning.

(50)

2 0

0 9

Fa

ll M

ee

tin

g

A

G

U

H23C-0974

Author : Liang-Cheng Chang 1 Yu-Wen Chen 2 Chen-Che Pan 3

I

ntegr

ating Evolution Algorithms with Linear Programming to the Conjunctive Use of Surface

and Subsurface Water

Liang-Cheng Chang

Professor of Department of the Civil Engineering, National Chiao Tung University, 1001

University Road, Hsinchu, Taiwan 300, ROC

email : [email protected]

Yu-Wen Chen

Post-Doctor of Department of the Civil Engineering, National Chiao Tung University, 1001

University Road, Hsinchu, Taiwan 300, ROC

email : [email protected]

Chen-Che Pan

PhD. candidate of Department of the Civil Engineering, National Chiao Tung University,

1001 University Road, Hsinchu, Taiwan 300, ROC

email : [email protected]

Introduction

1. The changing climate make the happen frequency of water supply

shortage increases than before.

2. Only using surface water resource is easily effected under extreme

climate condition and also associates with high supply risk.

3. On the contrary, because the relationship between groundwater

resource and rainfall quantity is not so close, conjunctive use

between surface and subsurface water can make water supply

stable.

Methodology

A hybrid algorithm is proposed for the conjunctive use model of

surfaceand subsurface model.

Model Architecture

Artificial Neural Network: Ggroundwater system (the groundwater

transfer function)

Linear Programming :Gsurface water system (determine the

reservoir

supply amount

)

Genetic Algorithm :Gthe bridge between surface water and subsurface

water system (determine the

pumping rates

)

figure1 : The model architecture for CUMM.

Model Objective Function (Original)

: means water supply quantity of the k-th linkage at the t-th time step and

also is the decision variable

: means groundwater pumping quantity of the k-th well and also is one of the

decision variables

: means the ratio of water shortage to water demand for the i-th demand node

: means index value discrepancy between the m-th and the n-th reservoir

: means index value discrepancy between surface water system and

groundwater system

: means ratio of spare volume to system capacity for the j-th reservoir

: means ratio of spare volume to system capacity for groundwater system

and

: are assigned to present the priority between different sub-objects

Modified Model Objective Function for Hybrid Algorithm

The entire problem can be divided into a primary sub-problem and secondary

sub-problem.

Primary sub-problem (solved by Genetic Algorithm)

Secondary sub-problem (solved by Linear Programming)

Groundwater ANN model development

Numerical Cases

A recursive ANN model is used to simulate the relationship between groundwater level and

pumping. The training and verification data for ANN is collected from a numerical model,

MODFLOW 96, and groundwater system structure is shown as Fig. 2. The structure of the ANN

model for training stage and prediction stage are shown as Fig. 3. The input nodes contain

state variables, , and decisionvariables , and the output nodes just contain state

variables, In the prediction stage, the structure of recursive ANN model needs to be

modified as Fig. 3(b). Similar with the traditional numerical models, the results calculated from

output nodes are recursively assigned to the input nodes for the next step simulation.

Three cases have been presented in this study (shown as Table 4. Case #1 is a case of surface

water independent operation and Case #2 and #3 are cases of conjunctive use operation. In

Case #2, the wells just supply while the surface system is empty by using "surface water supply

firstly" policy. In Case #3, the supply policy is determined by our proposed model, CUMM.

Cases Result

Comparing Case #1 and #2, an additional groundwater supply, whose maximum

supplycapacity is just 8.67\% of the total water demand, can significantly decrease

the shortageindex (SI) from 2.93 to 1.54.

Comparing Case #2 and #3, the higher operating efficiency of the proposed model

make the system SI significantly decrease to 1.13.

Because of the usage of the hybrid framework and ANN, the computational efficiency

of proposed model is higher than other models with purebred framework or traditional

groundwater simulation.

The proposed model has ability to extend to solve a complicated large field problem

andalso will be a valuable tool for conjunctive surface and subsurface operations

乾旱時期用水分配模式之研究---子計畫：乾旱時期地面水與地下水聯合運用作業模式研發(I)

行政院國家科學委員會專題研究計畫 成果報告