Controlling Smart Water Distribution System using Model Predictive Control Algorithm

With the increase in water demand, water distribution networks (WDNs) become more complex as the requirements to meet operational and economic standards increase, which results in the need for a complex control system. One of the most effective strategies to accommodate such system is Model Predictive Control (MPC) which allows the prediction of future plant behavior based on past performance and a set of control objectives. This study proposes a state space-based model predictive controller for WDNs in Middletown, Pennsylvania. The aim is to achieve the optimal performance of the WDNs by following a given set-points accounting for a 24-hour water demand and to minimize the power consumption of the pump by manipulating the flow rates, thereby directly minimizing the cost. The system is comprised of a sequence of pumps that delivers water to two water tanks in series that are connected to the end-users. The model applies Multiple Input and Multiple Output (MIMO) systems with two control inputs: actuator flow rates from reservoir and water tank, and two observable outputs: water levels in two tanks. Multi-objective functions are minimized to stabilize the control and guarantee water supply, subject to hard constraints. The results show that the output reached stabilization and tracked the references with an average deviation of 3.5x10-6, meeting the water demand. Total power consumption and daily demand charge from the optimized actuators are also identified. In addition, system response comparisons between MPC and Linear Quadratic Regulator (LQR) are presented to illustrate the effectiveness of the proposed model.