A methodological approach to designing sewer system control

When designing sewer system control, there is a lack of methodology and tools that can aid in the design process. In 2004 the PASST1 framework was presented that focuses on determining the potential for control in sewer sys-tem operation. However, for the actual design of control systems urban drain-age planners still have to rely on their operational knowledge combined with model simulations and trial and error. This is an inefficient process where the final design largely depends on the urban drainage planner’s knowledge about the system dynamics and control in general. The motivation for this thesis was therefore the wish for a methodological approach to sewer system control design. Using a case study the following research hypothesis was tested in this thesis:
Using classical and modern control theory, a methodological approach can be derived for designing sewer system control. This can aid urban drainage planner and other professionals in the planning phase of sewer system con-trol design and effectively contribute to find novel control solution.
It was investigated if the established methodology used in classic control the-ory for process control design can be applied meaningfully to the sewer sys-tem. As the methodology takes its basis in a hierarchical decomposition of the control problem based on time-scale, it was also investigated if sewer sys-tem control can be decomposed in a similar manner.From a review of existing control systems for sewer systems in Europe, it was concluded that sewer system control can also be decomposed in a hierarchical manner based on differences in time-scale. The proposed time-scale dependent hierarchy for sewer system control contains four layers that each handles their own dedicated task. From the bottom and up they are: 1) the regulatory control layer, 2) the coordinating control layer, 3) the optimisation layer and 4) the management of objectives layer.
The time-scale dependent hierarchy for sewer system control is put into a framework that also contains a terminology related to control. In this way the1 Planning aid for sewer system real time controlvframework can help to compare different control system solutions and facili-tate a clear communication between different professions and disciplines working together in sewer system control design.
Starting from the hierarchical decomposition of sewer system control in lay-ers, a stepwise approach to design sewer system control was proposed and followed. The individual layers of the hierarchy were designed one by one for a case study in Copenhagen, with the methods and tools taken from both clas-sical and modern control theory.
The tools of classical control theory are developed for systems that can be approximated by linear models. The main challenge of using classical control theory on the case study was therefore the transient nature and the non-linearity of the sewer system dynamics. The methodology was adapted, by linearizing the sewer system model at various points in time, creating a step-wise linear model. The results of the linearization showed that the sewer sys-tem dynamics could be divided into four phases, characterised by the follow-ing operation modes: dry weather, filling, saturation and emptying. Having obtained a piece-wise linear model for each of the operational modes, the tools from classical control theory, such as the calculation of the condition number and the relative gain array, could be successfully applied to the sewer system. Based on the results a pairing between the measurements variables and the actuators could be suggested.
Having proposed to decompose the sewer system control in a hierarchical manner, it became necessary to investigate the role of the lowest layer in the hierarchy, which is the regulatory control layer. Traditionally the role of the regulatory layer is to reject disturbances and track the setpoints, and the sim-plest form of regulatory control has just constant setpoints. However, in a transient system like the sewer system, the setpoints may change dramatically and rapidly. Therefore the regulatory control layer may not have the same functionality when designed for the sewer system. From the application of the classical control theory it was found that the system dynamics could be de-scribed by four operational modes, and instead of a fixed setpoint the regula-tory control layer needs changing setpoints, according to the operational modes. These can either be fed from a coordinating control layer or from an online optimisation.
To design an optimisation to feed setpoints to the regulatory control layer, modern control theory was applied to the case study. The optimisation was tested when it acted directly on the actuators and when it acted on the regula-vitory control layer. The two optimisation based control structures were evalu-ated from a one year simulation and the results showed that there was little difference in the performance. The optimisation based control structures were also compared to the existing control and the regulatory control with set-points coming from the coordinating control layer, and here the latter showed the best performance. This was not unexpected, since the true potential of having optimisation arises, when a system has many control loops with limit-ing constraints and/or changing prioritisation between them. The results showed that for small sewer systems, where the complexity is limited, it is not necessarily the best option to implement advanced optimisation based control systems. Therefore it is also advisable to approach the design of a control system in a methodological manner, where the design and evaluation can be done step by step.
Based on the experiences gained from designing sewer system control sys-tems for the case study, a systematic methodology for designing sewer sys-tem control is proposed that combined the steps, control and optimisation tools and methods used throughout the thesis. The proposed methodology provides a basis for gathering experiences with sewer system control design and knowledge sharing; and will help generate control systems of the future that are more robust, more structured, have a better performance and are easi-er to maintain.