Assignment # 4

Article Reference:
Byoung Ho Lee, Rolf A Deininger, January/February 2009, Optimal Locations of Monitoring Stations in Water Distribution System, Journal of Environmental Engineering

Summary:
US Environmental Protection Agency requires all the drinking water authorities to meet the drinking water quality standards in the distribution system. There are a set of regulations for selecting the monitoring locations, sampling frequency and the different water quality parameters to be monitored. For setting up a monitoring station, flow, flow pattern and pressures in the system need to be determined which is done using different hydraulic models. The paper describes about the methods of how to choose the monitoring stations thereby maximizing the coverage. Author explains how this problem was solved using the integer programming.

Author illustrates a small example. In the example the water fractions of each node were calculated and it is summarized in the matrix W, which shows how each node contributes to the sampling station. Based on the W matrix a knowledge based matrix K is derived which shows the decision of which node to be considered as the covered. Coverage criteria are selected to base the selection of the covered node. For this particular example they have considered 50%, which means the selected covered node should pass through at least 50% of the sampled water. Coverage matrix is obtained from the K, by converting all the non zero entries to ones in the matrix. Optimizing function of the problem can be formulated like below, where d=demand at the node, y=set which denotes if the demand at this node is covered, x=sampling station at node and a= transpose of the Matrix K.

One more example was considered for locating the monitoring stations in a medium sized city like Flint, Michigan. The city gets its supply from Huron River through a pipeline operated by Detroit Water and Sewerage board. The treated water is channelized through the pipeline till the location of original treatment plant which is not in use any more. As of now there are total of 14 monitoring stations which cover about 18.3% of total demand. For finding out the optimal solution of location of monitoring station first hydraulic analysis was carried out using COVER which formulates the coverage matrix W. COVTOIP is used to convert the coverage matrix to standard input file for IP package. It was found that after optimization total coverage was found to be 54%, when compared to 18.3% previously. As this was considered for only one scenario, author explains about how to handle the multiple scenarios situation. When considering multiple flow scenarios we need to take into account flow pattern which change based on hourly, daily or seasonal. By considering different demand patters, and the maximum coverage percentage to be achieved, the objective function and constraints are formulated which are then used to final an optimal solution. In similar grounds, an example of optimizing the location of monitoring stations for Cheshire city in Connecticut. By considering different flow patterns, different options of monitoring stations were found out. Final stations were selected based on the coverage percentage.

Discussion:
As of now as there are no guidelines for selecting the monitoring stations, combination of pathway analysis, coverage matrices and integer programming provides a first step for landing up at the monitoring stations. However for the large networks this method was very difficult to handle because of the large dimensionality of the problem. Further studies were carried out based on the methodology as proposed by Lee and Deininger one of them was by Muhammed and Khurram who tried to optimize the solution using Genetic Algorithm.

The article was very interesting as it gave an insight about how the monitoring stations are selected for finding out the water quality in the distribution system. Few of the points which was not clear was about how did they land with the assumption of taking 50% of coverage. How well can this are taken into consideration in other type of problems (cities). If this was my research topic I would like to explore how the location of the station might vary by considering different optimization techniques. Would like to to explore how such a technique be used to solve the problem for a big distribution system.