Hybrid nonlinear model predictive control of a cooling water network

Abstract

A Hybrid Nonlinear Model Predictive Control (HNMPC) strategy is developed for temperature control and power consumption minimisation of a cooling water network. The HNMPC uses a gradient descent optimisation algorithm for the continuous manipulated variables, and an enumerated tree traversal algorithm to control and optimise the Boolean manipulated variables. The HNMPC is subjected to disturbances similar to those experienced on a real plant, and its performance compared to a continuous Nonlinear Model Predictive Control (NMPC) and two base case scenarios. Power consumption is minimised, and process temperature disturbances are successfully rejected. Monetary benefits of the HNMPC control strategy are estimated.