Modeling Water Supply Reliability following a major earthquake

Transcription

1 PNWS AWWA May 10, 2013 Modeling Water Supply Reliability following a major earthquake Anne Symonds, PE AECOM David Myerson, PE SFPUC AECOM AGS, A Joint Venture

2 Agenda AWSS and Background Level Of Service (LOS) Goals Modeling Ignitions and Demands SynerGEE GIRAFFE Project Development and Assessment Program Recommended to SFPUC Next Steps Page 2

3 AWSS and Background San Francisco Following the 1906 Earthquake and Fires

4 AWSS Overview Separate high pressure water system originally constructed in Miles of 10 inch and larger Cast Iron Pipe Two Salt Water Pump Stations Three gravity storage facilities fed by domestic system Total system consists of high pressure pipe, cisterns, suction connections, saltwater pumps, fireboats, reservoir and storage tanks Pipe construction with extra thick walls Double lead and restrained joints Limited connections Image Source: Page 4

5 AWSS System Map Page 5

6 AWSS Operations Over time has been extended to cover more of the City, now 135 miles, 80% cast iron pipe Used by the San Francisco Fire Department (SFFD) and owned and maintained by the SFPUC City Distribution Division (CDD) Normally operates in three pressure zones Designed to allow one pressure zone with pressures up to 340 psi Page 6

7 Modeling Process

8 AWSS Capital Planning Study (CS-199) Goal: To maximize the likelihood that the AWSS will effectively provide required fire fighting capabilities after a major seismic event. Recommended LOS Objectives: AWSS will reliably provide water to supply the probable fire demands after a M7.8 San Andreas earthquake Each fire response area will be XX% reliable in supplying probable demands. AWSS will be YY% reliable in supplying probable demands Citywide. Page 8

9 Modeling Overview - Modeling objective: To determine the amount of water the AWSS can deliver given a set of demands - Hydraulic and reliability modeling to determine project hydraulic benefits - Tools used: SynerGEE GIRAFFE (Cornell University) EPANET ArcGIS R script Page 9

10 Estimating Fire Demands

11 Fire Demand Formulation 1 Monte Carlo Fire Flow Data by Block (Scawthorn) Group Blocks by Fire Response Area (FRA) Delineation of FRAs 2 Select Likely Ignition Location for Each FRA (review most likely and largest fire locations) 3 Calculate Magnitude of Fire Flows for each FRA 4 5 Result: Representative Fire Demand per FRA. Input into hydraulic and reliability modeling at closest hydrant location Modeling Page 11

12 1 Delineation of Fire Response Areas Based on SFFD Response Districts Further refined based on fire density Page 12

13 2 Ignition Location Selection Red Block with Highest Occurrence of Ignitions Blue Top 5 Demands per FRA Page 13

14 Average Total Demand 3 Fireflow Magnitudes Average Total Demand per Quintile Stochastic set of ignitions (1000 iterations) 60-minute 3 rd quintile selected as representative demand set Suppression accounted for by removing all fires at first minute Quintile Page 14

15 4 Model Demand Locations Demand node highlighted in blue Location based on nearest network node to locations found in step 2 Page 15

16 Estimating System Reliability

17 Hydraulic Model (SynerGEE) AWSS Existing Condition Model 6,274 junctions 6,312 pipe segments 178 valves 10 pumps 5 tanks Model reviewed and calibrated with flow test data Page 17

18 SynerGEE Model Results Page 19

19 EPANET Model (Converted from SynerGEE) Page 21

20 Graphical Iterative Response Analysis for Flow Following Earthquakes (GIRAFFE)

21 GIRAFFE Overview Developed at Cornell University by Tom O Rourke and colleagues Uses open source EPANET engine Deterministic and probabilistic simulations 5 modules: System Definition [Input] Seismic Damage [Input] Earthquake Demand Simulation [Module] Hydraulic Network Analysis [Computation] Results Compilation [Output] Page 23

22 GIRAFFE GUI (Windows XP) Page 24

23 GIRAFFE Model Inputs Pipeline Fragilities Based on PGV values by block PGV estimated from regressions between block centroid distance to San Andreas fault and grouped by shear wave velocity (Vs30) System Information File (EPANET file) Page 25

24 GIRAFFE Process Flow 1 Read System Input and Repair 2 Generates random pipe Rate Files breaks and leaks 3 Reconfigures system network 4 5 Solves hydraulic network Identifies nodes with negative pressures Multiple Iterations 6 Removes node with highest negative pressure 7 Computes serviceability Page 26

25 GIRAFFE Model Controls Simulation type: Deterministic (single) Monte Carlo Fixed Monte Carlo Unfixed Convergence Criteria Break/Leak Ratio Leak type probabilities Random seed generator Page 27

26 GIRAFFE Output Pipe break/leak list System damage file Serviceability result Information by time-step for: Nodes Pipes Pump stations Valves Tanks Output summary Page 29

27 Solution to Output Limitation 1 2 Take damage file from each damage scenario Determine demand nodes with negative pressures 3 Systematically decrease 4 demands with negative pressures until no negative pressures on demand nodes Determine remaining system negative pressures are left 5 6 Systematically decrease ALL demands until minimal negative pressure remain Determine amount of water system can provide Page 31

28 Sensitivity Analyses Performed Demand location and magnitudes Fireboat assumption Partial demands Infirm zones Page 32

29 Cistern, Suction Connection, Alternate Water Modules

30 5 Modeling Summary Stochastic Set of Fire Demands Aggregated Demands by Fire Response Area High Pressure System Reliability (GIRAFFE) Cisterns, Suction Connections, and Alternative Water Source Reliability (Computational Modules) FRA Reliability Score Citywide Reliability Score Page 34

31 Cisterns Page 35

32 Suction Connections - Green: existing suction connection - Red: proposed suction connection (Alternative C only) Page 36

33 Alternate Water Sources Palace of Fine Arts Pond Suction Line Laguna Honda Lake Merced Page 37

34 Reliability Scoring

35 Model Results Terminology - Reliability: Available supply / demand requested - Citywide reliability: Average of each Fire Response Area s reliability Page 40

36 Reliability Score Calculation Example FRA FRA demand: 5,000 gpm Total water available: 4,170 HPS contribution: 3,850 gpm (average from 15 iterations) Cistern contribution: 320 gpm (average over 1000 sets) Suction connection contribution: 0 gpm Alternative water contribution: 0 gpm FRA reliability: 83% (4,170/5,000) Page 41

37 Reliability Score Context System tested at 3 rd quintile demands with 46 simultaneous ignitions Considers initial fire department response but doesn t model response or resources required HPS evaluated with an aggregated demand for each FRA while other water sources are evaluated by block Reliability index scores are a relative representation of system performance Page 42

38 Alternative Programs and Assessment

39 Alternative Program Development Developed 3 Program Alternatives each composed of multiple projects - Alternatives A & B: new pipe extensions and water supply and some cisterns - Alternative C: all cisterns Performed Pairwise comparison of the Alternatives Recommended Preferred Alternative for further consideration Page 44

40 Evaluation Criteria Program Alternative Scoring Alternative Delivery Reliability Firefighting Capability Ranking B A C A/B tie Cost B A C Schedule A/B tie C Operations and Maintenance Insurance Premiums Environmental / Community Impacts B A C A/B/C tie B A C Final Ranking B A C C Page 45

41 Next Steps

42 Next Steps for AWSS Evaluation of other potential combinations of systems to meet potential fire demands Improvements to Potable Water system Construction of Multiuse or hybrid pipes Evaluation of relative risk of event Construction of projects funded by 2010 ESER bond Recommendations for future bond election Page 47

43 Thank You Modeling Water System Reliability