Mechanistic Modelling Within a Hybrid Controller

The Water Research Foundation project 5121, Development of Innovative Predictive Control Strategies for Nutrient Removal aims at developing full-scale hybrid nutrient management controllers by employing a hybrid approach that combines both machine learning and mechanistic models. This paper specifically examines the role of mechanistic models within the context of hybrid modeling, focusing on their integration and interactions with machine learning components. One crucial aspect of the hybrid optimizer is its ability to automate the simulation and calibration of mechanistic models, which transforms mechanistic models into soft sensors. The soft sensor provides benefits facilitated by the mechanistic models including monitoring the state of every process, enabling plant-wide control, generating adequate data for machine learning models and etc. Integrating the mechanistic model into the hybrid optimizer has resulted in valuable practical knowledge and experience, which significantly contribute to the ongoing discussions on hybrid modeling practice.