Use of hybrid models in wastewater systems

Incomplete first-principles knowledge of a system can obstruct the development of models for investigating the system's dynamic behavior. One approach proposed to overcome this problem involves using hybrid models. Hybrid models build inasmuch prior knowledge as available and then use empirical components, such as neural networks, to complete the description. In this paper hybrid modeling techniques are applied to two wastewater systems. The first system involves data from a Danish wastewater plant, and the second involves using a hybrid model for control. Potential problems involving the use of hybrid models are discussed.