Load Forecasting Technology for Wastewater Treatment Plants: From Numerical Modeling and Simulation to the Digital Twin

Kurzfassung

Wastewater treatment plants (WWTPs) play a vital role in public health and environmental protection but face ongoing operational challenges due to fluctuating wastewater inflows and composition, as well as unpredictable energy costs and supply risks. To improve resilience and efficiency, integrating real-time data and predictive analytics through a digital twin is crucial. This work introduces a grey-box probabilistic model within a digital twin framework that forecasts both wastewater inflow and chemical oxygen demand (COD), a key indicator of pollution. The model, based on Generalized Additive Models for Location, Scale, and Shape (GAMLSS), uses weather data, real-time plant measurements, and network-wide monitoring to predict future conditions with uncertainty quantified through confidence intervals derived from a Johnson SU distribution. It captures complex non-linear patterns, such as daily, weekly, and seasonal trends, using quadratic splines and ReLU activation functions, ensuring both accuracy and interpretability. Unlike purely data-driven black-box models, this approach maintains transparency while enabling scalable deployment across larger sewage networks. By combining probabilistic forecasting, feature selection across multiple monitoring points, and integration into an optimization-ready digital twin, the model supports smarter, adaptive energy use and enhances the overall reliability and sustainability of water infrastructure.