Veranstaltungsprogramm

Urban drainage networks are critical to address the exacerbated flooding in the cities due to climate change and rapid urbanization. Badalona, a city in Spain has been facing recurrent pluvial flooding due to high-intensity and short-duration rainfall events driven by the Mediterranean climate. Although the city has its combined sewer networks modeled in SWMM, MOUSE, and Info works, it is supported through manual calibration methods. This approach is highly time-consuming for such a large network, and is subjective, depending on the modeler which can lead to suboptimal parameter selection. This research aims to address this limitation, by configuring a hybrid algorithm leveraging Non-Dominated Sorting Genetic Algorithm (NSGA-II) and SWMM to automatize the calibration process comparing single and multi-objective optimization strategies. Results demonstrate that the multi-objective optimization strategy offers a more holistic approach with a balance between various objective criteria effectively. With this methodology integrated into urban drainage management, an effective and comprehensive framework can be provided for sustainable water infrastructure that helps in achieving improved water quality, model performance, system resilience, and flood prevention.

Deploying sensors network and collecting and using sensor data is a backbone of Digital Twins (DTs) for engineering systems, such as Urban Drainage Systems (UDS). Such data often exhibit missing values and anomalous readings due to many factors (e.g. sensors malfunction, hardware limitations, weather and site conditions). System analytics in DTs rely on these data and requires postprocessing algorithms capable to detect and reduce problems in collected data. This research aims to develop an advanced ML-powered algorithm for automated data anomaly detection (data validation) and estimation of missing data. This algorithm utilizes an ensemble of ML models to address data quality issues. The algorithm is tested on a synthetic dataset for a part of Belgrade stormwater system.

In light of the impacts of climate change and urbanisation, data-driven urban water management (UWM) has been proposed as a means of transforming sewerage and drainage systems, aiming to achieve resilient and sustainable served territories. The increased collection and utilization of data offers significant potential benefits; however, missing data and data quality pose challenges for their subsequent use in calibrating and validating computational models, evaluating infrastructures performance and implementing automatic control solutions in the field.

We introduce a novel deep learning approach that harnesses the power of generative artificial intelligence to enhance the accuracy of contextual forecasting in sewerage systems. By developing a diffusion-based model that processes multivariate time series data, our system excels at capturing complex correlations across diverse environmental variables, enabling robust predictions even during wet weather periods. To strengthen the model’s reliability, we further calibrate its predictions with a conformal inference technique, tailored for probabilistic time series data, ensuring that the resulting prediction intervals are statistically reliable and cover the true target values with a desired confidence level. Our empirical tests on real sewerage system data confirm the model’s capability to deliver reliable contextual predictions, maintaining accuracy even under severe weather conditions.

This study introduces a generic data-driven model for urban hydraulic modelling, designed to estimate flow rates in catchment areas. The model employs a machine learning architecture trained on historical observational data from 30 Water Resource Recovery Facility (WRRF) sites across France, covering diverse hydraulic and environmental conditions. A generic data-driven model integrates principles of machine learning with extensive, varied datasets to create adaptable tools that generalize across different locations and conditions without requiring site-specific recalibration. In this case, the model predicts peak flow curves, enabling the estimation of peak and nominal flow rates over time intervals ranging from 1 hour to 6 months. The database was constructed from observational data measured in the sewage networks and at the inlets of wastewater treatment plants managed by Suez Eau France, supplemented with data on land cover, soil type and rainfall from Open Data sources. The model’s effectiveness was evaluated on a pilot site, validating its versatility. This tool serves as a valuable decision-support resource for engineers and consultants in urban water management. By leveraging machine learning and a robust, diverse dataset, this approach enhances reliability, adaptability, and efficiency in addressing complex urban hydraulic challenges.

Reducing sewer overflows and flooding is vital for urban drainage systems. Traditional real-time control (RTC) methods often lack efficiency, leading to the exploration of new techniques like deep reinforcement learning (DRL). This study assesses RTC effectiveness using a multi-agent DRL approach, with a framework evaluating control objectives, decision time, robustness, and adaptability. A case study in Suzhou, China, involving 31 rainfall events, shows DRL reduces flooding and overflow risks by 15.1% to 43.5% compared to traditional methods. However, this benefit came at the cost of higher energy use (10.3% and 7.7%) and increased pump switches (4.3 and 2.2 times), reflecting a trade-off shaped by the objective weightings (80% environmental, 10% energy, 10% switching). DRL also offers superior efficiency, robustness, and adaptability, highlighting its potential in urban drainage management and infrastructure resilience.

This article presents a novel hybrid model combining data and mechanisms for high-resolution water level prediction in pipeline networks. The model utilizes graph convolutional neural networks to integrate network topology information for precise predictions and incorporates de Saint-Venant system equations through physics-driven neural networks. Compared to traditional data-driven, mechanism-driven, and hybrid methods, this model achieves 5-minute, 1-centimeter resolution predictions while maintaining computational efficiency and high accuracy. In an experimental drainage system in Suzhou, China, the model's RMSE for predicting water levels and pipeline flow rates is 0.014 and 0.012, respectively, with NSE values of 0.802 and 0.883. The model's computation time for 24 hours of data at 5-minute intervals is 0.981 seconds.