Smart Control and Energy Recovery of Pump-As-Turbines in Water Networks using Model Predictive Control

Abstract

This masters dissertation explores the potential of maximizing power production in Pump As Turbines using Model Predictive Control. Pump As Turbines are pumps operating in reverse mode. They are proposed to be installed in water networks where an excess head is present which is otherwise dissipated when a Pressure Reducing Valve is installed. Therefore, this work proposes to maximize that potential using Model Predictive Control which is a control algorithm that uses a model and constraints to predict future changes in a system. In the context of water networks, the system is the water network, therefore, hydraulics of the network is represented using mathematical equations. Moreover, water networks are highly non-linear systems therefore this work also presents a way to model the non-linearity that occurs due to head loss equations. This is therefore embedded into the algorithm to accurately represent a water network. Additionally, the Pump As Turbine operation is included into the Model Predictive Control algorithm using a logic operation. Finally, to observe if there is an advantage of using the PAT operation inside the algorithm or outside (Linear MPC), two different controllers are compared. The results showed that using the logic operation inside the algorithm has positive effects when maximizing multiple PATs simultaneously. Alternatively, it also proved that the controller which has the logic embedded (Hybrid MPC) showed better performance in predicting the output. Lastly, this work also showed that tuning the weights in the objective function has effects in increasing or decreasing the power production.