1 A Memetic Algorithm for Water Distribution Network Design R. Baños* , C. Gil*, J. I. Agulleiro*, J. Reca † * Dpt. Computer Architecture and Electronics, University of Almería (Spain) † Dpt. Rural Engineering, University of Almería (Spain) 11th Online World Conference on Soft Computing in Industrial Applications – September-October , 2006
17
Embed
A Memetic Algorithm for Water Distribution Network Design · The memetic algorithm here proposed works better than other heuristics. When the problem instance grows, the memetic algorithm
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
A Memetic Algorithm for WaterDistribution Network Design
R. Baños* , C. Gil*, J. I. Agulleiro*, J. Reca†
* Dpt. Computer Architecture and Electronics, University of Almería (Spain)† Dpt. Rural Engineering, University of Almería (Spain)
11th Online World Conference on Soft Computing in Industrial Applications – September-October , 2006
2
Summary
Water distribution network design (WDND)Description of the problemFormulation
Experimental analysisTest networksParameter settingsResults at Alperovits and Shamir’s networkResults at Hanoi network
Conclusions and future work
3
Water distribution network design (WDND)
Goal: find the best way in term of investment cost ofconveying water from the sources to the users, satisfyingtheir requirements.
Variables imposed in the model:Network connectivity, Capacity of the tanks,Power of the pumps,Pressure required.
Decision variables:Pipe diameters.
Description of the problem
4
Water distribution network design (WDND)
WDND: minimize fitness function “F”
where: F is the cost function,m is the number of pipe diameters,ci is the cost of the pipe with diameter i per unit of length,Li is the total length of pipe with diameter i in the network,cp is a penalty coefficient, hrj is the required presure head in the node j,hj is the current presure head computed by EPANET for node j.
Formulation
5
Linear programming techniques,Non-linear programming techniques,Heuristic methods:
Experimental Experimental analysisanalysisResults in Alperovits and Shamir’s network
All the methods reach the minimum cost.On average of 10 runs: MA outperforms to the other methods (all the configurationsreach 419.000 monetary units). Difference among methods is, on average, lesser than 1.7%.
13
Experimental Experimental analysisanalysisResults in Alperovits and Shamir’s network
All the methods reach 419.000 monetary units, although SA converge faster
14
Experimental Experimental analysisanalysisResults in Hanoi network
MA obtains the best investment cost (6295909) while other methods are more expensive.On average of 10 runs: MA outperforms to the other methodsDifference among methods is, on average, lesser than 1.95%.
15
Experimental Experimental analysisanalysisResults in Hanoi network
MA obtains best results than other methods.Although the difference, in percentage, is small, in layout problems it becomes important.
16
ConclusionsNew memetic algorithm for Water Distribution Network Design (WDND).
Comparative study of memetic algorithms with other heuristic approaches.
The memetic algorithm here proposed works better than other heuristics.
When the problem instance grows, the memetic algorithm performs better.
Future workMulti-objective treatment of this problem considering reliability.
Extend the formulation to consider other designing aspects (connectivity, etc.)
Conclusions and future work
17
QUESTIONS?
COMMENTS?
11th Online World Conference on Soft Computing in Industrial Applications – September-October , 2006