Reducing the cost of operational water on military bases through modeling, optimization, and control

Abstract

Military municipal water systems provide safe and clean water to the surrounding community while also supporting the intense and often unpredictable training schedules of the tenant units. Much like their civilian counterparts, military water systems are also consumers of great amounts of energy and capital. As a part of the Army Net Zero program in 2011, an annual water inventory conducted on eight U.S. Army installations concluded that consumption was 5.5 billion gallons. Using the Environmental Protection Agency’s average national estimate of 1,500 kWh of energy consumed for every 1,000 gallons of treated water, it is readily apparent that the department of defense is a heavy consumer of both water and energy. Because the scale of the military’s usage is so vast, so too is their waste. Waste in water systems is common and commonly neglected, as many were initially constructed decades ago and the commodity that they transport is relatively inexpensive. However, recent droughts affecting regions of the United States highlighted the need to conserve and avoid waste, regardless of the commodity price. The efficiency of water systems is highly dependent upon developing accurate models and using those models to accurately deal with disturbances such as demand and chlorine concentration. This work extends water distribution system modeling, optimization, and control to a military setting where constraints are tighter for resiliency purposes, demands are often unpredictable, and saving money and water improves defense capabilities. First, a discretized nonlinear, equation based model of a known system at an existing U.S. Army installation that accurately predicts system behavior under typical demand considerations. The model is calibrated for accuracy using actual system data from a military installation and employed in a nonlinear optimization program to study reduction of costs, minimizing waste, and improvements in energy efficiency. Demand profiles were constructed from residential data and scaled to better represent demand on military bases. With very little adjustment, this model can be used to optimize similar systems in the military inventory. Water and energy savings exceed 10% in the optimized system, which predicts the Army could save greater than $1.5 million per year in the continental United States if rigorous optimization was conducted on storage and pumping at every base. It is shown that a reduced order empirical model is a viable alternative to the computationally expensive equation based approach. The empirical model is used to implement model predictive control, providing the system protection against large and unpredictable disturbances. This method adds an additional manipulated variable, chlorine injection, to ensure efficient constraint compliance. Experimental results show this method further supports the aforementioned savings in the optimized system alone, while efficiently handling disturbances. This research closes previous gaps in research, particularly on military installations. First, it serves to minimize the system volume, or excess water on hand, while meeting all demands and strict system constraints dictated by resiliency and emergency preparedness. Secondly, this work uses a nonlinear model predictive control structure to deal with large and unpredictable disturbances that occur uniquely on military installations. The feedforward control action integrated into the controller is particularly effective at minimizing disturbances on inlet concentration.