Evolutionary optimization of combined sewer overflow control

Model predictive control (MPC) is coupled with a real-coded Genetic Algorithm to predict a decision sequence that minimizes combined sewer overflow (CSO) volume for a 3-hour rainfall event over a hypothetical sewer system. Rainfall is transformed to overland runoff through the cell model which depicts each sewershed (draining to an overflow dropshaft) by two linear reservoirs in series, and water entering the interceptor is routed downstream to establish water levels at the dropshaft connections. A pumping rate at the most downstream end of the interceptor plus one sluice gate position for each dropshaft connection will be altered to produce the best control strategy. Resulting management scenarios disperse overflows differently throughout the sewer, but may yield similar overflow volumes. This paper describes the simulation approach taken and displays the overflow distribution for favorable control sequences.