Finite state machine control optimized by simulated annealing for reducing combined sewer overflows

Abstract Urban drainage networks that combine stormwater and wastewater can experience combined sewer overflows (CSOs) during intense rainfall, discharging untreated effluent to receiving waters. This paper proposes a supervisory control strategy based on finite state machines (FSMs) that coordinate pump and valve actions in sewer storage tanks. The FSM parameters (level thresholds and a flow-distribution factor) are calibrated offline using simulated annealing over bounded, discretized parameter sets to minimize the cumulative overflow volume. The tuned controller is evaluated in the Benchmark Simulation Model for Integrated Urban Wastewater Systems under multiple rainfall scenarios. The optimized FSM reduces CSO volumes by up to 56.9%, decreases pollutant loads by up to 61%, and lowers overflow frequency by up to 40%, while improving dissolved oxygen conditions in the receiving river. After calibration, the controller runs online with fixed parameters as a set of state-transition rules, requiring negligible runtime computation and remaining compatible with resource-limited automation infrastructures.