WDSA / CCWI 2018 Joint International Conference

WDSA / CCWI 2018 Joint International Conference
WDSA / CCWI 2018 Joint International Conference

Battle of Post-Disaster Response and Restoration (BPDRR)

EPANET model

The Battle of Post-Disaster Response and Restoration (BPDRR) is the eighth battle competition of the Water Distribution Systems Analysis conferences. This competition is open to individuals or teams from academy and industry. Participants are invited to propose a solution to the problem to be presented in a special session during the WDSA / CCWI 2018 conference at Kingston, Canada.

For this version of the battle competition, participants will deal with the restoration of a water distribution system that has been damaged after an earthquake disaster. This problem represents some of the decisions that a water utility must take in these kinds of disasters, and it is intended to explore what is the best way to use available resources for the restoration and to identify ways in which a water utility can be prepared for such scenarios.

How to cite the problem

To cite this problem and models in your conference/journal paper or report please use the following citations:

Key Dates

  • October, 2017 - Publication of problem description & Call for abstracts
  • January 31st, 2018 - Abstract submission deadline
  • February 15th, 2018 - Notification of acceptance of abstracts
  • June 1st, 2018 - Full paper submission deadline


Frequently Asked Questions (FAQs)

1. Should criterion 6 (C_06) be evaluated over 5 days or 7 days? The equation in the annex uses 5 days, whereas the description on p.3 states 7 days.

There was a typo in the Annex. All criteria will be evaluated for the next 7 days. Thank you for noticing the mistake. It was fixed on February 12th, 2018 in the problem description.

2. When modelling an action, at what time does it take effect? E.g. if a valve closure takes 15 minutes, is the closure implemented in the model at the end of the 15 minutes? If two valves are closed as part of the same action (e.g. isolate P231), is their closure modelled sequentially (and if so, in the order in which they are listed in the spreadsheet?) or simultaneously?

Actions will only take effect AFTER the duration of the activity has passed. It means that if a crew is sent at 6:30am to isolate pipe 3897 (which has valves at both of its ends), then the pipe will be modeled as close only after 7:00am, once the 30mins (2 valves x 15mins/valve) have passed. This includes the case of pipes that require more than 2 valves to be isolated, so if a crew is sent at 6:30am to isolate pipe 231 (which requires 4 valves in different pipes to be isolated), then the whole activity will take 1hr, meaning that only after 7:30am the pipes 231, 217,234 and 235 will be modeled as closed - simultaneously.

3. Should the reopening of valves be included in the schedule? It can’t be assumed that valves are reopened immediately following replacement of a pipe as a single set of valve closures may be used to isolate more than one pipe.

The task of reopening pipes/valves in the schedules has been included in the problem description (February 12th, 2018). For convenience, we will leave this task with a zero duration.

4. Is any time allocated to the reopening of valves?

No, this is a zero duration task.

5. In the inp files the reporting time step is 30 min and the total duration is 24 hours. According to the text it should be 15 min and 168 hours, respectively. Please clarify.

For the evaluation of solutions the time step will be 15 mins and the duration will be 168 hours. The participants need to modify this values accordingly. However, they should keep in mind that getting the Status Report may take a considerable time when using EPANET2 interface.

6. Should C_03 and C_05 be calculated with 15 min time steps ?

Yes. The time step is 15 mins for all consideration of the problem unless it is stated otherwise.

7. Description states that restoration time should be reported in hrs (rounded to the lowest hour). How should this assumption be understood in case one valve needs to be closed to repair more than one pipe? How can isolation time get calculated? For which breaks should this time be appended?

When the description states that the time needs to be rounded to the lowest hour, that only applies for ‘Replace’ and ‘Repair’ after using the equations from the table (i.e. 0.223*D^0.577 and 0.156*D^0.719). ‘Isolate’ is a separate activity (i.e. different row in the Excel file) that accounts only for the sum of the time spent in closing the necessary valves, and its duration is equal to 15mins x Number of valves to be closed.

If a valve needs to be closed to repair more than one pipe you just need to have an ‘Isolate’ activity closing that valve and any other valve required for the isolations; then the necessary ‘Repair’ or ‘Replace’ activities, one for each damaged pipe; and then ‘Reopen’ activities to open the closed valves whenever considered convenient.

8. Annex presents six criteria to evaluate solutions. Also weights for scenarios have been declared.  That means we obtain 6 values of criteria (one aggregated value for each criterion). Is there any aggregation method for the evaluation of all criteria to get one aggregate mark for our solution?

The Battle of the Network Models Committee suggested the use of a non-unified set of criteria such that no ranking or absolute winner for the battle can be declared. This means that participants can treat the different objectives in the way the consider more realistic and convenient for the city.

9. Can we isolate a certain area of the network (like a DMA, for example), and repair all the pipes within this area at once? Or, we must isolate only the valves according to the pipe we are going to replace/repair?

You are allowed to close any set of valves even if they do not isolate a damaged pipe or if together they isolate a whole sector. To do so, include in the schedules 'Isolate' activities for each pipe that contains a valve that you want to close. For the 'Pipe' column write the ID of the pipe that contains the valve, and for the 'Pipes closed for isolation' column just write the same ID of that only pipe so it can be modeled as a CLOSED pipe in EPANET.

10. Will pumps be operating as given or can we optimise them? Can pumps be isolated as links?

Pumps cannot be optimized per se since they are always turned on. However, since you are allowed to close and open independent valves as desired, you can close the closest valve upstream or downstream of a pump to simulate a pump turned off. You just need to include an ‘Isolate’ activity in one of the schedules, for the pipe with the valve that you want to close (e.g., to turn off pump 6068 you can include the activity ‘Isolate Pipe 4384’) . You are allowed to ‘Isolate’ and ‘Open’ a pipe as many times as you want, but keep in mind that each closure takes 15 minutes.

11. Pipe 6062 is closed in the model but has the code valve 0-0. Does this mean that there are pipes that can become closed/open in simulation but have no isolation valves in the response scheduling?

The model includes 11 pipes that are closed from the beginning of the simulation. They represent boundaries from a prior sectorization of the network. They must be kept closed during the entire simulation even if they do not have valves in it. To isolate any other pipe in the system, the valves from its correspondent hydraulic segment must be closed.

12. What are the units for criteria C_03 and C_06. (In the solution spreadsheet, the units given for ‘Volume of water lost during the next 7 days after the event’ (C_06) are L/s, which is a flow rate not a volume)

The Solution Template has been modified. For C_03, please use [minutes*%] (since the functionality is expressed as a percentage and not as a fraction) and for C_06 specify total volume in [L].