Fresh water supply and flood protection are two central issues in water management. Society needs more and more fresh water and a safe water system to guarantee a better life. A more severe climate will result in more droughts and extreme storms. As a consequence, salt water intrusion will increase. Therefore, clean and fresh water is becoming scarce. Potentially, there lies a severe conflict between people's demands and what nature can provide. In practice, water systems are complex. Both water quantity and quality criteria must be served. Moreover, water is normally used as a multi-functional resource. For example, water in a reservoir is used for irrigation, power generation, flood protection and reclamation. These objectives are usually in conflict most of the time and it is not easy for people to cope with these contradictions. Smart regulation of water systems is essential not only from the world-wide water issue perspective, but also from the specific water problem aspect. Real-time control is a powerful tool to help people with accurate regulation of water systems. In practice, water quantity control is extensively studied, but fully integrated water quantity and quality control has hardly been touched. Moreover, in order to deal with multi-objectives in a water system, advanced control techniques, such as model predictive control (MPC), are often required which require extensive computational resources. This brings forward two research questions: 1: What is the possibility of controlling both water quantity and quality in a water system? 2: In MPC, what is the possibility to reduce the computational burden in order to make the control implementation possible? In this PhD thesis, a case of polder flushing in real-time is selected for the first research question, which includes both water quantity and quality problems.The task is to flush polluted water out of the polder with clean water while keeping water levels close to the setpoints. Instead of manual operation which is often applied in practice, control systems were designed with feedback control and MPC. In MPC, different types of internal models were applied ranging from a linear reservoir model to hydrodynamic models. The different control performance of the two controllers were compared. We conclude that real-time control is possible to maintain both water quantity and quality at the same time in a one dimensional water system model. Furthermore, MPC performs much better than the classic feedback control in controlling the water quality when operational limits are very strict. In MPC, using different internal models will also result in different control performance, affecting both control effectiveness and computation time. Being an advanced control technique, MPC is playing a more and more important role in controlling water systems. The computational burden is the main barrier for MPC implementation. In this PhD thesis, we propose a control procedure of MPC with a model reduction technique, Proper Orthogonal Decomposition (POD), in order to speed up the computation. POD is able to reduce the order of states and disturbances, and speed up the matrix operation in MPC. In a test case, we concluded that MPC using the reduced model is a good trade-off between control effectiveness and computation time. Therefore, the proposed MPC procedure is considered as a successful method for MPC implementation.