Improvements in water supply systems based on optimization and recognition of consumption patterns

Water supply systems consume large amounts of energy because of the pumping processes involved. The operational strategy of using frequency converters enables the system to work with better adjusted discharge rate to meet demand. In this case, an optimization strategy can establish an optimal procedure in order to schedule the rotational speed of pumps over a period and guarantee a volume of water in the supply tank. This work presents and solves an optimization problem that provides the optimal schedule for the rotational speed of pumps in a real water supply system considering minimizing the use of electricity and the cost thereof and maintenance. The optimization problem is based on two Artificial Neural Networks (ANN) models that provide the total power consumption in the pumping system and level of water in the tank. Pattern recognition techniques in univariate time series based on the real data are used to forecast the demand curve according to the season ofthe year. The results show the potential savings generated by the proposed method and show the feasibility of scheduling the rotational speed of the pumps to ensure the minimum energy cost without affecting hourly demand and the security of the supply system