On Reduced Observer-Bank Synthesis and Edge Implementation for Residual Chlorine Concentration Soft Sensors in Water Distribution Networks

In the present paper, a reduced observer-bank synthesis and edge-oriented implementation framework for a model-based residual chlorine concentration soft sensor in water distribution networks is presented. For a benchmark network described through nonlinear hydraulic and chlorine transport–reaction dynamics, local discrete-time linear approximants are derived around admissible operating points. Based on these approximants, a finite bank of full-order Luenberger-type observers is designed for the estimation of nonmeasurable chlorine concentration variables. The observer parameters are selected through a metaheuristic multicriterion tuning procedure that combines discrete-time pole-placement requirements with estimation-performance objectives. A central contribution of the paper is an observer-bank construction algorithm that eliminates redundant target operating areas. The algorithm adaptively covers the normalized input domain by generating each observer operating area after nonlinear steady-state computation, local linearization, observer tuning, and validation for each operating point. The resulting observer bank is combined with a switching supervisor based on target-area overlap and measurable-output convergence. The resulting scheme is implemented in a Node-RED edge runtime, where multiple observers execute in parallel and the supervisor/switching mechanism selects the active observer that provides the appropriate estimate. Deterministic telemetry generated by a MATLAB R2025b simulation of the water distribution network is fed to the Node-RED instance, enabling repeatable evaluation of acquisition, estimation, switching, single-stream telemetry arbitration, and runtime adaptation. Computational and edge runtime experiments demonstrate the feasibility of deploying the proposed switching-observer soft sensor for real-time chlorine monitoring in water distribution networks.