(497f) Two-Point Constraint Control of Water Quality in Distribution Networks

Providing potable drinking water with sufficient free residual chlorine (used as disinfectant worldwide) to prevent microbial regrowth and contaminant by-product formation is of utmost importance for municipal authorities. The work presented in this paper aims towards water quality control as it transports through water distribution networks. Maintaining free residual chlorine concentration levels in water is a critical task as chlorine reacts while it transports through the pipe network with incoming dissolved organic matter (DOM) in bulk flow as well as in biofilms present at the pipe surfaces. Presence of excessive chlorine in water causes the generation of carcinogenic disinfectant by-products such as tri-halo methane (THMs). On the other hand, lower chlorine levels in water results in microbial contamination. Hence, this is a two-point constraint control problem of maintaining free residual chlorine levels under the maximum and minimum bounds by suitably optimizing the disinfectant dosage at booster stations. However, transport delay, and complex interrelations present amongst the nodes in large water distribution network makes it difficult to design a global feedback control system. Therefore, in this work we have proposed to decentralize the water distribution system by clustering demand nodes of the network based on knowledge of interactions between manipulating and control variables using partial correlation analysis. In such multivariable control, due to the presence of interactions, the stability of system might get affected. Knowing the feasibility, stability and integrity of control structure by clustering variables in the above manner constitutes an important design aspect. To address this aspect, we have proposed efficient use of effective relative gain array analysis (ERGA) which accounts for transport delay also instead of regular steady state relative gain array analysis (RGA). The partitioning and control strategy presented in this work is based on concentration sensitivity matrix analysis by using PCA (Principle Component Analysis), Partial Correlation Analysis and ERGA stepwise.We also considered the results of partial correlation analysis to generate communication links which will be implemented for higher level coordinated decentralized control to account for dynamic, non-negligible interactions as disturbances to local control units. The proposed three step clustering-mapping strategy and mathematical formulation for two point control by optimized dosage is successfully verified for the steady state case on a prototype distribution example network 2 of EPANET (USEPA, 2000).