Roland Löve, DTU
Opløsningen af hydrauliske modeller bliver ofte valgt efter praktiske hensyn. Vi har udviklet en algoritme til automatisk simplificering af 1D modeller og testet i et 45km2 opland i Melbourne, Australien hvordan forskellige simplificeringsgrader påvirker 1D-2D beregningsresultater på lokalt og oplandsniveau.
1. Simplification of one-dimensional
hydraulic networks by automated
processes evaluated on 1D/2D
deterministic flood models
Steffen Davidsen, Roland Löwe, Cecilie Thrysøe,
Karsten Arnbjerg-Nielsen
2. 11 May 2017DTU Environment, Technical University of Denmark
Purpose
• obtain stable 1D-2D model and reduce simulation times
• work towards standardized way of model reduction
• document effects on simulation results
2
3. 11 May 2017DTU Environment, Technical University of Denmark
Algorithm – Trim vs. Merge
3
Trim
Merge
4. 11 May 2017DTU Environment, Technical University of Denmark
Algorithm – Flow Chart
4
5. 11 May 2017DTU Environment, Technical University of Denmark
Algorithm – Travel Time Compensation
Apply travel time as delay on the rainfall input for each catchment
(in increments of 2min)
5
𝑣 = 𝑀 ⋅ 𝑅ℎ
2
3
⋅ 𝑆
1
2 velocity in deleted pipe
IF 𝑣 < 0.15
𝑚
𝑠
𝑣 = max(𝑣 𝑢𝑝𝑠𝑡𝑟𝑒𝑎𝑚, 0.15
𝑚
𝑠
)
6. 11 May 2017DTU Environment, Technical University of Denmark
Algorithm – Volume Compensation
Compute volume of deleted nodes and pipes
Increase downstream node diameter
6
𝐷 𝑛𝑒𝑤 =
𝑉𝑛𝑜𝑑𝑒 + 𝑉𝑑𝑒𝑙𝑒𝑡𝑒𝑑
ℎ 𝑛𝑜𝑑𝑒 ⋅ 𝜋
7. 11 May 2017DTU Environment, Technical University of Denmark
Technical Implementation
• Python Code
• Perform simplification on arrays describing the network structure
• Update the MIKE Urban database to match the simplified structure
• Update the 1D-2D coupling file
7
8. 11 May 2017DTU Environment, Technical University of Denmark
Case
• Elster Creek Catchment
• 45km2,
~100,000 households
• stormwater network with
outlet to the sea, modelled
in MIKE 2014 FM
• 11,000 pipes, mesh with
415,000 faces (10x10m)
• simplification takes ~1
minute
8
9. 11 May 2017DTU Environment, Technical University of Denmark
Results – Computation Time
9
10. 11 May 2017DTU Environment, Technical University of Denmark
Results - Hydrographs
Especially merging leads to
higher peaks:
• fewer nodes higher water
levels in the system faster
flow?
• smaller energy losses when
removing manholes?
10
merged
trimmed
11. 11 May 2017DTU Environment, Technical University of Denmark
Results – Flow 1D2D (2 year event)
11
Baseline
trimmed
merged merged+trimmed
merged without
increasing node
diameter
12. 11 May 2017DTU Environment, Technical University of Denmark
Results- Flooded
Area
12
trimmed
merged
merged+trimmed
Baseline, T100
13. 11 May 2017DTU Environment, Technical University of Denmark
Results - Flooded Area
13
% pixels correctly
modelled as flooded (Hits)
% pixels falsely modelled
as flooded (false alarms)
% pixels falsely not
modelled as flooded
(misses)
CSI skillscore
14. 11 May 2017DTU Environment, Technical University of Denmark
Results - Expected Annual Damage (EAD)
14
from: SVK Skrift 31
15. 11 May 2017DTU Environment, Technical University of Denmark
Results - Expected Annual Damage (EAD)
15
16. 11 May 2017DTU Environment, Technical University of Denmark
Conclusions
1. 1D simplification had quite significant impact on simulation times
(in a model with coarse surface resolution…)
2. Trimming affects flooded areas strongly
3. Somewhat higher peak flows when merging pipes
4. Little effect from travel time and volume compensations.
Increased node diameter leads to large flows from 1D to 2D model.
16
17. 11 May 2017DTU Environment, Technical University of Denmark
https://gitlab.gbar.dtu.dk/users/rolo/projects
(tools for flood damage calculation, coupling 1D-2D models,
parsing ERF files, 1D network simplification + clipping MU
models
released under GNU GPLv3)
Davidsen, S., Löwe, R., Thrysøe, C., and Arnbjerg-Nielsen, K. (2017) Simplification of
one-dimensional hydraulic networks by automated processes evaluated on 1D/2D
deterministic flood models. Journal of Hydroinformatics, accepted.
17