Exploring data flows and system links in real-time urban drainage modelling in the wake of digitalisation

Abstract

An increased number of combined sewer overflows (CSOs), as well as flooding of streets and basements, drives an on-going transformation of the urban drainage sector. Digitalisation is seen as a means of mitigating these events and requires the collection of system attribute data, input data and in-sewer observations. This data needs to be handled and analysed before it can be applied to decision-making, and this process is often dependent on models. Real-time models can be divided into monitoring, forecast and control models for system replication, forecasting of the system state and decision-making, respectively. The adequateness of these models may be evaluated by assessing how much information is embedded in each modelling aspect (the ‘granularity’) and comparing this with the granularity of the quantity used to evaluate the performance of the modelled solution (the ‘indicator’). Some of the underlying methods for the models have been around for decades, but technical, human and linguistic barriers have impeded their uptake in practice. Furthermore, the urban drainage system is physically linked to other natural and man-made sub-systems through the flow of water; however, this has traditionally been either ignored or not fully exploited in real-time urban drainage models. The current challenges may require that we explore these system links further. The hypothesis of this thesis is that digitalisation can facilitate the exploitation of existing and emerging data flows as well as system links in real-time models to obtain better solutions. This hypothesis is investigated through three digitalisation examples including a monitoring, a forecast and a control model. Digitalisation of the water distribution system has led to an increased number of smart meters measuring the water consumption at each house with a high temporal resolution. Linking water supply and sewer systems by expanding the observational space to include smart meter data may be used to obtain wastewater inputs distributed in both space and time. In this thesis, smart meter consumption data was therefore coupled to an urban drainage monitoring model and the results compared with flow measurements from the sewer system. This was not straightforward since the required data was hidden in different data silos and thus difficult to access. Moreover, the simulated wastewater flow did not always match the measurements, most likely due to erroneous in-sewer observations. Smart meter data may thus also be used as an independent data source to validate and question in-sewer observations. Digitalisation of urban drainage systems is expected to lead to increased in-sewer sensing. This makes it possible to expand the application area of traditional in-sewer observations to include real-time model updating, hereby keeping the models in touch with reality. In this thesis, the updating of a surrogate forecast model with level and flow observations improved the flow and overflow forecast performance two hours into the future. Overall, level observations improved the model the most. The forecasts were also improved by using a distributed representation of the dry weather flow compared to adding the dry weather flow profile to the downstream model output. Obtaining such a distributed dry weather flow through in-sewer sensing requires a large number of sensors. The use of smart meter data may thus also be relevant for forecast models. Digitalisation is expected to promote the development of new types of actuators to be regulated by more advanced control schemes, hereunder model predictive control (MPC). This allows for the activation of otherwise passive links between the sewer system, stormwater control measures and the urban landscape. This control can be used to minimise CSO, bypass, energy costs, etc. with lower environmental impacts than by constructing new grey infrastructure. Having water visible on the surface in a controlled manner is expected also to increase amenity values and the public’s resilience. Nuisances caused by the control may include an increased risk of unintended flooding and disturbance of people’s daily lives. In this thesis, the control was implemented by expanding the model boundaries of an internal MPC model (a coarsely grained model that simulates the sewage flow) to cover both the underground and above-ground systems. The control concept was close to optimal in respect to CSO minimisation when the prediction horizon was near the transport time through the system. Input forecast models such as the updated surrogate model may facilitate such a horizon. Silo thinking in respect to data, systems and institutions, low data quality, a conservative and fragmented sector and linguistic uncertainty pose limiting factors for reaching the full potential of using digitalisation to exploit traditional and emerging data flows, as well as the interlinked nature of the urban drainage system in real time. The thesis provides several structures to diminish linguistic uncertainty and shows that it is possible to expand observational spaces, application areas and model boundaries, not only to obtain a more efficient urban drainage management, but also to tap into new areas such as sustainability, resilience and liveability.