The Research On Contamination Source Identification In Water Distribution Netwoks Based On PSO

In terms of solving the problems of drinking water contamination caused bysecondary pollution and artificial poisoning in water distribution networks, thecontamination source tracking model can identify the source location s to isolate thecontaminated area and minimize its hazards, which is significant to the drinking watersecurity emergency handling.According to the experiment of Cr(VI) solution intrusion into a laboratory waterdistribution network, the changes of contaminant Cr(VI) concentrations with time andspace in the pipe network are analyzed. On this basis, the hydraulic and water qualitymodels, which are coincident with the actual experimental scenario in the experiment--tal network, are established and calibrated.The simulation-optimization model for contamination source identification inwater distribution networks is established as a program complied in Matlab, with thesum of square differences between simulated contaminant concentrations andmeasured ones at monitoring sites as the objective function and the EPANET toolboxas the embedded simulation engine. Besides, Particle Swarm Optimization algorithmis used to obtain the information of contamination source location, contaminantintrusion starting time and intrusion speed.Based on the Cr(VI) intrusion experiment and the simulation-optimization model,sets of simulated contaminant concentrations and measured ones are respectively usedas data source to identify contamination sources, which shows that the model isaccurate and efficient. Through the comparison analysis, the model is less accuratewhen measured contaminant concentrations are used as data source. In addition, it ismore accurate and efficient to be used to identify a single contamination source thanto identify multiple sources.By setting the parameters influencing the model and comparing the output results,the effects of influencing factors, such as network topologic structure, PSOparameters, forward model error, are analyzed. On the basis, the parameters aboutinfluencing factors are reasonably set. Results show that computational accuracy andefficiency of the simulation-optimization model could be high if the followingconditions are achieved: network structure is similar to the actual one, few joints hassimilar downstream to the real contamination source, the forward model error forwater distribution networks is very small, monitoring sites are reasonably arranged, and PSO parameters and the simulated-optimization time period are reasonable.