Theory & Applications of Stochastic State Space Systems: with examples from Wastewater Treatment and District Heating

The global climate aspirations of a sustainable future place increasingly larger demands on the energy efficiency and flexibility of all our systems. These ambitious ideas can arguably only be realized by exploiting the capacities of our systems fully, obtaining more with fewer resources. A central aspect to this end, is the use of data-driven mathematical models and machine-learning algorithms, why the continual pursuit of new applications in these fields is crucial.

This thesis is concerned with the use of stochastic state-space systems in real-world applications, for modelling, prediction and control purposes, and it introduces the necessary theoretical methods that enable such applications. A central outcome of the thesis is a statistical package written for the R programming language which significantly eases the effort required to work with these systems in practice.

The thesis furthermore explores two specific modelling use cases within the fields of wastewater and district heating, specifically for estimating district heating network temperatures, and for sludge blanket prediction in secondary clarifiers at wastewater treatment plants. The results in the former case show that network temperatures can be robustly and precisely estimated, and the developed model simultaneously yields house-specific thermal properties useful for operators. The results in the latter case show that the developed model is able to accurately predict the sludge blanket height, but the accuracy relies in turn on precise input forecasts, in particular of the plant inflow rate. In addition, treatment plants must have mechanisms that secure well-controlled clarifier flow-distribution in order to be able to obtain the best model performance. The thesis goes on to propose using the developed model in a predictive control algorithm operating on inflow bypass or clarifier recycle flow, for better exploitation of hydraulic capacities and to provide embedded flexibility.