Adaptive Model Predictive Control for fire incidents in water distribution networks

Abstract

Water utilities in UK aim in network pressure reduction as a means to reduce leakage, in order to meet regulatory targets and avoid financial penalties. At the same time, however, they are not legally bound to guaranty specific fire flow rates at fire hydrants, thus posing a potential a risk to fire fighting. In the exiting literature, fire-flows in water distribution networks have been considered primarily with respect to design, vulnerability and capacity analysis, while control of pumps and valves has focused mainly on minimizing energy costs (and to a lesser extent pressure) under normal operating conditions. This study presents a (non-linear) adaptive model predictive control methodology, for water distribution networks, combining two separate modes of control: 1) Normal Control, when the network operates under normal conditions and the objective is to minimize energy costs and average zonal pressure. 2) Fire Control, when a fire incident occurs and fire-flows need to be delivered at hydrant nodes while, to the furthest possible extent, resume delivery of customer demand without over pressuring the network. The proposed methodology is applied on an operational network from UK.