Modelling and predictive control of a recirculating cooling water system for an industrial plant

tThis paper deals with the design and implementation of a Model Predictive Control (MPC) algorithm foran industrial recirculating cooling water system. The plant is composed of three large reservoirs, fourpumping units, three cooling towers with modulating fans, and a piezometric tower. A physical modelof the system has been derived from balance equations and has been validated with experimental data.Then, a nonlinear state observer has been developed to estimate unmeasured quantities, such as thewater flow rates and temperatures recirculating from the utilities. Finally, an MPC algorithm has beendesigned by reformulating the original nonlinear system in terms of a linear time-varying one to improvethe computational efficiency of the method. The presence of on/off actuators, pumps, and fans, makes theresulting optimization problem a Mixed Integer Quadratic Programming one. Extensive simulation resultsand comparisons with the relay-based control scheme currently implemented witness the potentialitiesof the proposed solution in terms of energy savings, better control, and lifetime of the equipments. Theproposed MPC algorithm is currently being tested on the real plant, its very satisfactory performanceconfirm the results of the simulation experiments.