Dynamic soft sensor based parameters and demand curve estimation for Water Distribution System: Theoretical and Experimental cross validation

Abstract

The ostensive process of each Water Distribution System (WDS) is manipulated by the hydraulic parameters such as dimensional features, working state of the system components and the demand profile of the end user. In general cases, the access of those factors is unobtainable due to the shortage of sensors and more capital expenditure on the measurement methods. In such circumstances, the crucial parameters have been foreseen with the obtainable measurements of WDS, which are often contaminated with measurement noises. In this study, a hybridized version of metaheuristic based Grey Wolf Optimization (GWO) is proposed using static Kalman Bucy (KB) mechanism as Hybridized Grey Wolf Optimization (HGWO). This proposed hybridization suppresses the effect of measurement noises over the parameter estimation. Further, to make the HGWO adaptable for any fault occurrence or dynamical changes, a statistical based fault diagnosis is also proposed. To assess the efficiency of the proposed algorithm under fault occurrence situation a synthetic WDS system considered as the first case study, where faults purposely injected into the system and performance of the estimation algorithm studied. The second case study is from eastern WDS section from Peroorkada town, Trivandrum City, India given by Kerala Water Authority (KWA) consists of 110 unknown parameters and 6 unobservable demand profiles. The credibility of estimation algorithm also tested in the Hardware in Loop platform (HIL) for field applications. For both the case studies, parameter estimation is carried out using and compared with the related algorithms, viz., Particle Swarm Optimization (PSO) and GWO. The obtained result shows that the proposed HGWO provides better estimates of the factors and the unobservable states for both the case studies.