An artificial intelligence approach for optimizing pumping in sewer systems

This paper presents details of a fuzzy logic system developed for the control of sewer pumping stations for energy costs savings. This is part of an ongoing collaborative project between Anglian Water and the University of Sheffield. The model rules and operation are developed for one representative pumping station in order to enable the identification of potential benefits and inhibitors to Anglian Water. Results are included that demonstrate the potential for energy cost-savings by application to a single pumping station for dry weather flow conditions and through comparison to current on/off switching rules. The fuzzy system is shown to be robust to changes in flow pattern (using both modelled inflow data and real data from a flow survey), but sensitive to changes in price structures. Application of a genetic algorithm (GA) search technique was used to adjust the parameters that define the membership functions in the fuzzy rules, in order to provide automated minimization of the energy costs towards an optimal solution. The GA system is shown to be transferable to another pumping station with different pump sizes, wet well capacity and inflow pattern. The GA solution outperformed the base case in terms of energy costs and switching totals.