Diffusion-based Time Series Forecasting for Sewerage Systems

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.