From soft sensing to anomaly detection in combined sewer systems

Abstract

Climate change, intensified urbanization and stricter environmental regulations are straining the existing combined sewer systems, and large investments are needed for futureproofing. With a combination of models and data, utility companies can maximize the use of the available capacity to reduce the risk of flooding and pollution. Hardware sensors are typically used to monitor the state of sewer systems, but they are vulnerable to instrument faults, communication errors and cyberattacks. If models of the system are accurate enough, they can act as soft(ware) sensors, working alongside hardware sensors or replacing them for periods of time. Model predictions can also be used to validate available observations and detect anomalous behavior. This thesis investigated the potential of two very different models for soft sensing and anomaly detection in combined sewer systems: 1) updated 1D hydrodynamic models, which describe all the main hydrological and hydraulic processes occurring in a sewer system and are dynamically updated with sensor observations; 2) Long Short-Term Memory neural networks, which can learn repetitive patterns in water depth observations from combined sewers and predict the behavior of the system in response to a given set of inputs. Both methodologies were applied to real-world case studies, and specific advantages and disadvantages were discussed. By investigating and testing updated 1D hydrodynamic models and LSTM neural networks, this thesis demonstrates their concrete potential for soft sensing and anomaly detection applications and promotes their integration in the monitoring and control workflows of modern utility companies.