Stochastic characterization of flow and transport in networks
MetadataShow full item record
The ultimate objective of this thesis is to analyze the joint uncertainty distributions of multiple water quality parameters in order to statistically characterize the baseline variability of water quality in a distribution network. The water quality modeling capability of EPANET is employed, and the stochastic variability in demand pattern is stochastically generated for the network simulations. The resultant stochastic variations in water quality parameters at several locations within the network are analyzed using geostatistical tools. We found that the chlorine concentration at nodes along a straight pipe segment show similar temporal profiles. We also observed that network operating conditions such as the presence of storage tanks and their volumetric capacity influenced the pattern of variability and that manifested in the computed spatio-temporal variograms. Further analysis showed that the system flow conditions such as the aggregate base demand, demand patterns, initial water quality at the source and all nodes, and storage tanks capacities are the major factors affecting the quality and variability of chlorine in the distribution network. Other factors include network connectivity and operating conditions (pumps and tank cycling). These induce flow reversibility in the network that induces more variability. Principal component analysis (PCA) is demonstrated as a tool to identify grouping of nodes that exhibited similar characteristics of water quality. The analysis confirmed our hypothesis that a robust statistic (variogram) can be computed by pooling all the data at nodes that link to the fundamental nodes identified by the PCA. Joint simulation of several water quality variables such as pH, temperature and TOC, taking into account their influence on chlorine concentration is demonstrated using the beta version of the multi-species extension to EPANET (MS-EPANET). The simulated results exhibit realistic correlation between the various water quality variables and the resultant chlorine concentration. Refining the multi-species simulation procedure by using physically realistic values for bulk chlorine decay constant (KB) and extending the PCA analysis procedure including more types of water quality variables such as pH, electrical conductivity and total organic carbon (TOC) are suggested as items for future research.