State Estimation Network Design for Water Distribution Systems

Abstract

State estimation (SE) involves estimating state variables of interest that cannot be directly measured by using measurable variables. In water distribution system (WDS) SE, nodes are often aggregated to reduce the number of unknowns. To achieve high SE accuracy, the optimal observation locations in the WDS should be determined. This paper proposes an optimal meter placement and node grouping (OMPNG) model for WDS demand estimation (DE). The nonlinear Kalman filter (NKF) method is used to estimate the nodal group demand (NGD) from pipe flow measurements at meter locations. A k-means clustering method is introduced to generate the initial node grouping for the proposed OMPNG model. An elitism-based genetic algorithm is employed to minimize the sum of the NGD root-mean-square errors (RMSEs). The proposed OMPNG model was applied to the modified Austin network DE problem, and the results were compared with those obtained by optimizing node grouping with fixed meter locations based only on engineering sense. The results showed that the proposed OMPNG model significantly improves the DE accuracy and reliability.