Presentation is loading. Please wait.

Presentation is loading. Please wait.

Effects of Measurement Uncertainties on Adaptive Source Characterization in Water Distribution Networks Li Liu, E. Downey Brill, G. Mahinthakumar, James.

Similar presentations


Presentation on theme: "Effects of Measurement Uncertainties on Adaptive Source Characterization in Water Distribution Networks Li Liu, E. Downey Brill, G. Mahinthakumar, James."— Presentation transcript:

1 Effects of Measurement Uncertainties on Adaptive Source Characterization in Water Distribution Networks Li Liu, E. Downey Brill, G. Mahinthakumar, James Uber, Emily M. Zechman, S. Ranjithan North Carolina State University

2 Contaminant Source Determination u Rapid identification of … ¡ Contamination source location ¡ Starting time ¡ Mass loadings at different time u When to stop the search and make final decision u Necessary information for threat management in water distribution systems

3 Challenges of Source Identification u Inverse Problem ¡ Ill-posed/Non-uniqueness u Under dynamic environments ¡ Dynamic system ¡ Dynamically updated observations u Under noisy environments ¡ Measurement error ¡ Uncertain demands ¡ Model error

4 Simulation-Optimization Method Hydraulic Simulation Water Quality Simulation EA-based Optimizer Observed Data C sim Source characteristics t C obs

5 Adaptive Dynamic Optimization Technique (ADOPT) u An EA-based search u Solves as information becomes available over time u Multiple solutions to assess non-uniqueness

6 Objective u Investigate the effects of sensor errors on source characteristics obtained using ADOPT

7 Assumptions u Deterministic demand values u Conservative contaminant u Contamination occurs at any one location in the network u Only sensor errors are considered

8 Scenarios with Sensor Error u Scenario 1: Sensor with continuous malfunction u Scenario 2: Sensor with intermittent malfunction u Scenario 3: Sensor activates after a lag time of first detection u Scenario 4: Sensor with systematic reading error

9 Contamination Case A Mass Loading Profile

10 Contamination Case A… Node 197 Node 184 Node 211 Node 115 Time Step (10 mins) Observed Conc. (mg/L) Time Step (10 mins)

11 Results for Case A with Perfect Data Node 197 Node 184 Node 211 Node 115 True source Best solution Prediction Error = 0.026 mg/L Observed Conc. (mg/L) Time Step (10 mins)

12 Case A : scenario 1 Node 115 True concentration Observed concentration Observed Conc. (mg/L)

13 Case A : scenario 1 Node 115 True concentration Node 184 Observed concentration Best solution Observed Conc. (mg/L) Time Step (10 mins) Observed Conc. (mg/L)

14 Case A: scenario 2, 3 & 4 Best solution True concentration Observed concentration Scenario 2 Scenario 3 Scenario 4 Observed Conc. (mg/L) Time Step (10 mins) Observed Conc. (mg/L) Node 115

15 Contamination Case B True Source Mass Loading Profile

16 Case B … Time Step (10 mins) Observed Conc. (mg/L) Time Step (10 mins) Observed Conc. (mg/L) Node 197 Node 184Node 211

17 Results for Case B with Perfect Data

18 Node 197 Node 211 Node 184 Results for Case B with Perfect Data

19 Case B: scenario 1 Time Step (10 mins) Observed Conc. (mg/L) Time Step (10 mins) Observed Conc. (mg/L) Node 197 Node 184Node 211

20 Case B: scenario 2 Time Step (10 mins) Observed Conc. (mg/L) Time Step (10 mins) Observed Conc. (mg/L) Node 197 Node 184 Node 211

21 Case B: scenario 3 & 4 Scenario 3 Scenario 4

22 Summary for results Number of alternative source locations Scenario #

23 Summary for results… Scenario # Mass Loading difference at true source location (g/min)

24 Final Remarks u Source characteristics identified by ADOPT are influenced by the type of sensor errors. u Investigate effects of demand uncertainty. u Update ADOPT to be robust under combined noisy conditions.

25 Acknowledgements This work is supported by National Science Foundation (NSF) under Grant No. CMS-0540316 under the DDDAS program.


Download ppt "Effects of Measurement Uncertainties on Adaptive Source Characterization in Water Distribution Networks Li Liu, E. Downey Brill, G. Mahinthakumar, James."

Similar presentations


Ads by Google