Transcript ModelCare 2005
Well flow & Pumping tests
new techniques → new possibilities → new software
MLU for Windows
• • •
MLU in broad outline
Development of the solution technique Possibilities and limitations Differences with similar software
How does MLU work
(demonstration) • • Multi-aquifer representation and data entry Presentation of results • • • •
Calibration of analytical models
• Pumping test analysis Parameter optimization Analysis of the results
Some practical cases
Recovery test, slug test, etc.
Aquifer thermal energy storage
MLU in broad outline
Kick Hemker
1
- Development of the solution technique Analytical solutions for well flow in layered aquifer systems
1980 → NOW 2 -
Possibilities and limitations aquifers / systems maximum model capacity input and output
3 -
Differences with similar software with respect to classical p.test software with respect to numerical models
Well flow in layered aquifer systems
1979
Pumping test near Lexmond 1 2, 3 4
Analytical solutions for well flow in layered aquifer systems
1980
steady state well flow aquifer: T; aquitard: c recharge from the top (and base) 1 fully penetrating well drawdown s = f(layer number, r)
Analytical solutions for well flow in layered aquifer systems
1985
transient well flow aquifer: T, S; aquitard: c recharge from the top (and base) 1 fully penetrating well drawdown s = f(layer number, r, t)
Analytical solutions for well flow in layered aquifer systems
1987
aquitard storage aquifer: T, S; aquitard: c, S’ top and base: no drawdown / no flow 1 fully penetrating well drawdown s = f(layer number, r, t)
Analytical solutions for well flow in layered aquifer systems
1997
multi-screened well well radius, well storage s well , skin effect uniform gradient at well screens (layer no., t), s(layer no., r, t)
Analytical solutions for well flow in layered aquifer systems
1999
stratified aquifer well radius, well storage s well , skin effect partially penetrating well uniform drawdown at well screens (t), s(layer, r, t), flux well (layer, t)
MLU (DOS)
→
MLU for Windows
2007
theory → implementation 1 → 300 pumping/injection wells 1 → filters in any layer flux( 100 pumping periods
t
) for each filter + delayed observation well response
MLU for Windows Features of MLU version 2.25
• layered aquifers / aquifer systems • leaky, confined and unconfined • up to 40 aquifer layers • up to 300 pumping- and injection wells • up to 100 pumping periods for each well • up to 100 observation wells • up to 1000 measurements per obs. well • up to 16 parameters to be optimized • time conversion: sec, min, h, d and yr • copy & paste (spreadsheets) • time graphs (drawdown, head, flux) • contour plots (with animation) • bitmap and vector output • data output as fth, xyz and fem
MLU for Windows
Limitations
• infinite areal extent • each layer homogeneous and isotropic • only well flow • only Darcy flow • superposition of wells
which means:
• no drawdown cone in unconfined aquifer • no seepage face in phreatic well • no mutual effects of pumping wells • no sheet pile walls, etc.
and obviously also no:
• Noordbergum effect.
MLU for Windows Compared to classical pump.t.software
• Multi-layer • Multi-well • Multi-screen • Multi-Q (variable discharge) • Aquitard storage • Same interface for all tests: pump, recovery, slug- and st-drawdown tests, etc.
Compared to 3D numerical models
• No finite element or finite difference grid • No time steps • No multi-screen problem • Drawdown in pumping well (radius, storage, skin) • Delayed observation well response • Easy to design/adept well fields • More accurate results • Faster optimization
MLU Interface
Input 1/4:
General info tab
Time units Length units
MLU Interface
Input 2/4:
Aquifer system tab
Upper and Lower boundary conditions Aquifer (Yellow) Aquitard (Orange)
MLU Interface
Input 3/4:
Pumping wells tab
Check boxes to temporarily exclude individual pumping wells and periods Pumping wells may be screened in any selection of layers
MLU Interface
Input 4/4:
Observation wells tab
Screened in one layer only
MLU Interface
Output 1/3:
Optimization results etc.
MLU Interface
Output 2/3:
Time graphs
MLU Interface
Output 3/3:
Contour plot
Copy Time graph Copy contour plot Save curve data Save contour data Save model as FEM MLU Interface
MLU Help MLU Interface
Calibration analytical model
Kick Hemker
1 –
Pumping test analysis with MLU differences with graphical methods
2 -
Parameter optimization the parameters least-squares solution non-linear regression
3 –
Analysis of the results i s the result a proper solution ?
the accuracy
Pumping test analysis with MLU
Classical pumping test software
Based on graphical methods: - curve-fitting or Searching for a best-fit straight line - limited to 1 – 4 parameters - little information about the accuracy.
MLU → Calibration analytical model
Non-linear regression technique: - parameter optimization based on least-squares method - graphical inspection of the model fit - statistical information on the accuracy of the results.
Parameter optimization
What values can be optimized ?
Make a selection of: • T- and S-values of all aquifers • c- and S’-values of all aquitards using any code (1-9, a-z, A-Z) in the # column.
Use the same code to group two (or more) values as a single parameter for optimization
Parameter optimization
What values can be optimized ?
Hydraulic properties + also: • r c , r w and skinfactor of all pumping wells using any code (1-9, a-z, A-Z) in the # column.
Actual optimization parameters are dimensionless
0
: log (hydraulic property value/starting value)
1
: pumping well property/starting value Starting values must be larger than zero.
Computed hydraulic property = starting value * exp(parameter value) Computed pumping well property = starting value * parameter value
Parameter optimization
Least squares solution: Residual error = difference between the computed and the measured drawdown Sum of squares of residuals is minimized linear: log: residual error = computed – measured drawdown residual error = log (computed) – log (measured) Least-squares solution is obtained
iteratively
(Levenberg-Marquardt algoritm) each iteration step the sum of squares is reduced
stopping-criterion:
improvement sum < Rel * sum + Abs * Abs (ft 2 )
Parameter optimization
Non-linear regression Test case: Schroth.mlu Schroth & Narasimhan: GroundWater (35) 2, p.371-375 2 aquifers 1 pumping well 3 observation wells Log drawdown curve fitting
Parameter optimization
Analysis of the results
Has the optimization procedure been successful yet ? Two prerequisites: 1. Iterative process -> “parameters found” 2. Inspection of Time graphs -> “good fit” Only if both OK -> See “Optimization results” Values + accuracy
===================================================== M L U A Q U I F E R T E S T A N A L Y S I S For Unsteady-State Flow in Multiple-Aquifer Systems ===================================================== THE CALCULATED LEAST SQUARES SOLUTION Parameter value + Standard deviation T 1 58.0 + 3.3 ( 6 % ) T 2 5.156E+00 c 2 248.6 S 1 4.865E-04 S 2 1.678E-05 S' 2 1.467E-04 rc 1 5.277E-02 + 6.381E-02 ( 1 % ) + 6.2 ( 3 % ) + 5.231E-05 ( 11 % ) + 1.030E-06 ( 6 % ) + 9.490E-06 ( 6 % ) + 4.385E-04 ( 1 % )
Analysis of the results
Tab “Optimization results”
======================================================= ….
….
M L U A Q U I F E R T E S T A N A L Y S I S For Unsteady-State Flow in Multiple-Aquifer Systems ======================================================= Initial sum of squares is Residual sum of squares is Residual sum of squares (m²) 0.6031
Improvement last iteration 2.1E-12 Number of iterations Condition number 2.7390
0.0131
6 463.4
Correlation matrix (%) T 1 100 T 2 c 2 S 1 S 2 S' 2 rc 1 1 60 -82 -39 70 1 100 21 100 -14 53 3 -25 -78 100 -3 55 -6 43 100 -90 3 -67 100 -31 -1 100
Very high condition number (about 1e9 or higher) High correlations (near +/-100%) reduce the number of parameters
Some examples
Kick Hemker, Benno Drijver
1 –
Different tests pumping test recovery test slug test step-drawdown test
2
– Aquifer thermal energy storage (ATES differences with normal pumping well fields design of ATES well fields practical example
MLU for Windows Test cases 12 tests are available in the directory “examples”
layers pumping wells obs. wells 1 2 6 1 2 2 1 2 1 1 4 1 1 1 1 6 4 2 2 4 1 3 1 6 parameters 3 T1 c1 S1 6 T2 c2 S2 S’1 rc1 sk1 0 3 T1 S1 sk1 3 T1 S1 sk1 7 T1 T2 c2 S1 S2 S’2 rc1 2 T1 S1 7 T1 sk1
up to
sk6
Example 1: Pumping test The classical example “DALEM“ (Kruseman & de Ridder) Leaky aquifer Curve fitting: LOG-drawdown Linear-drawdown T 1780 ( 3 %) c 539 (36 %) S 1.6 10
-3
( 5 %) 1676 ( 3 %) 328 (22%) 1.8 10
-3
( 6%)
Example 2: Recovery test Pumping station Hardinxveld-Giessendam (Dec. 1981) pumping well radius = 0.155 cm LOG-drawdown curve fitting: without and with skinfactor T 854 ( 4%) S 6.1 10
-5
(10%) skin = sum of squares 1.5030 m
2
1321 ( 1%) 2.8 10
-4
(13%) 6.1 ( 4%) 0.0025 m
2
Example 3: Slug test Classical test example of “Cooper et al. 1967” Slug = 10.16 litre In MLU modeled as: for 0.1 sec a discharge of 0,1016 m
3
/s.
Cooper: T = 45 m
2
/d , S ~ 10
-3
MLU : T = 40,6 ( 4%) S = 1.9 10
-3
(29%)
Example 4: Step-drawdown test Classical test example “Clark 1977” Q increases from 1306 till 5019 m
3
/d in 6 steps (each 3 hr) Skinfactor increases with Q In MLU: 6 pumping wells, all at the same spot MLU : KD = 396 m 2 /d ( 1%) Sk 1 = 1.28 ( 7%) Q= Sk 2 = 1.69 ( 5%) Sk 3 = 2.07 ( 5%) Sk 4 = 2.42 ( 4%) Sk 5 = 2.77 ( 4%) Sk 6 = 3.19 ( 3%) 1306 m
3
/d 1693 2423 3261 4094 5019 + recov.
Aquifer thermal energy storage (ATES) Discharge = infiltration (net discharge = zero) Area of influence much smaller than a “normal well field” Effects of major heterogeneities in area of influence small An analytical model is justified in most cases
Design ATES Distance between Cold and Warm wells: larger: improves thermal operation smaller: reduces hydrological effects EXAMPLE
Properties ATES
Total capacity: Total volume: No. of wells: Filter depth:
Cross section
200 m³/hr 480.000 m³/season 8 (4 Cold and 4 Warm) 25 – 48 m
Single pumping well versus ATES Pumping well Drawdown cone after 50 days of a continuous discharge of 200 m³/hr (grid spacing: 1 km) ATES Hydrological effects after 50 days of a continuous discharge and infiltration of 200 m³/hr
Design ATES Some possible well configurations configuration A: relatively unfavourable
Design ATES Hydrological effects (5 cm contours at filter depth)
A B
grid spacing 500 m
C
Large capacity site model: Agriport A7 Total capacity: 40.000 m³/hr Pumped volume: 190 million m³/yr
Benno
Conclusions
1 – 2 3
-
– 4
“New” analytical solution for non-steady state well flow in layered aquifer systems (publications available as pdf) Superposition in space and time Parameter optimization technique Simple interface
Analytical model
• All sorts of aquifer tests • Design well fields Graphical + digital output hydraulic heads Statistical results of optimization Transfer to numerical model
MLU for Windows Version 2.25
A
MLU information + documentation www.microfem.nl/products/mlu.html
www.microfem.nl/download mlu-set.zip mlu.pdf mlu-user.pdf
= users guide mlu-tutorial.pdf = multilayer approach update.txt
= update history since 2008 mlu.pps
= MLU-LT software = fact sheet = powerpoint presentation
B
MLU office license, updates, support www.microfem.nl/order
C
Email: [email protected]